Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

MicMac – a free, open-source solution for photogrammetry

MicMac – a free, open-source solution for photogrammetry The publication familiarizes the reader with MicMac - a free, open-source photogrammetric software for 3D reconstruction. A brief history of the tool, its organisation and unique features vis-à-vis other software tools are in the highlight. The essential algorithmic aspects of the structure from motion and image dense matching problems are discussed from the implementation and the user’s viewpoints. Keywords: Free, Open-source, Photogrammetry, Structure from motion, Bundle block adjustment, Semi-global image dense matching Background or free/open-source. MicMac – together with Bundler, Photogrammetry is the art, science, and technology of PMVS, VisualSfM, openMVG, OpenCV and others – obtaining geometric information on the 3-dimensional belongs to the open-source solutions . This publication shape and orientation of objects from images and other aims at familiarizing the reader with the philosophy imaging sensors. It is a cheap measurement method- behind MicMac, some of its crucial algorithmic aspects, ology as it can be executed by practically any digital the software architecture and the pool of available tools. camera of a decent make. It is instantaneous as it cap- tures the observed phenomena at once and in a split Historical notes second, and highly automated, therefore accessible to MicMac has been developed at the National Institute non-expert users. Thanks to the field of computer vision, of Geographic and Forestry Information (IGN) and the photogrammetry rejuvenated and today places among National School of Geographic Sciences (ENSG), since other competitive remote sensing techniques (e.g. Light 2003. Initially, the software tools were developed having in Detection and Ranging LiDAR) [1]. The several mile- mind exclusively the IGN’s cartographic production. The stones leading to this progress are the automated interest independent tools were interfaced in 2005 via an XML points detection [2], the Structure from Motion (SfM) framework, allowing the user to freely parametrize the algorithms capable of reconstructing scenes from sets of calculations at all processing stages. In 2007, IGN began to unordered image collections [3, 4], and the dense image freely distribute MicMac under the CECILL-B license that matching techniques delivering surface models of resolu- is a version of the L-GPL license adapted to the french law. tion equal to the pixel size on the ground [5, 6]. Until 2008, the dense image matching of already ori- All this contributes to an ever growing visibility of ented images was possible only with the IGN’s internal the photogrammetric tools across various fields of sci- image file format. In the same year the Apero tool was ence and engineering, the growing market interest, and added to the sofware kernel, offering from now on the subsequently a multitude of photogrammetric/computer possibility to estimate camera exterior and interior orien- vision libraries and software solutions, be it commercial tations, with no restriction on the image format. In 2010, the XML interface was replaced by a simplified command *Correspondence: erupnik@gmail.com line. This evolution contributed to an improved accessi- ENSG, École Nationale des Sciences Géographiques, Champs-sur-Marne, bility, diffusion and subsequently a better visibility of the France 2 software in the scientific communities, and the general IPGP – Institut de Physique du Globe de Paris, Sorbonne Paris Cité, Univ Paris Diderot, CNRS, Paris, France public. Full list of author information is available at the end of the article © The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 2 of 9 Since 2010 MicMac has been undergoing a significant access to intermediary results in open data formats evolution due to its implication in many french and euro- allowing to interact with the processing chain at any pean projects. Besides the contributions to the source desired stage, code, new software distributions under Windows, Mac qualitative evaluation of the results via quality OSX, as well as the GPU processing for certain tasks indicators (e.g. bundle block adjustment residuals, became available. covariance matrices, parameter’s sensitivity, correlation images), MicMac vis-à-vis other tools a wide range of camera calibration models (e.g. the From the architecture point of view, the unique feature models adapted to consumer grade cameras, of MicMac with respect to its alternatives is that the large-frame aerial cameras, cameras with very long user can operate on different software levels, cf. Fig. 1. focal lengths, fish-eye and spherical cameras), The least experienced user can steer the processing by two-dimensional dense image matching for a simple command line where little or no parameters deformation studies, must be set. An experienced user can operate from the processing of frame camera and pushbroom sensor same command line by adapting the parameters typically images, filled with default values. An expert in photogramme- processing of scanned analogue images, try can access any parameter at any processing step via architecture adapted to big datasets. a set of XML files (cf. Fig. 1). On top of that, develop- ers and scientists can use MicMac as a library, implement their own algorithms, and share them for the benefit Software organisation of photogrammetric community. This said, MicMac was MicMac is organised in a number of modules, accessi- designed primarily as a tool for professionals and con- ble through a common command mm3d.Execution with a sequently is far from a model design in the sense of unique command has a twofold advantage: (a) it allows to HCI . factorize certain developments while keeping the binary MicMac, as a photogrammetric tool, stresses the files compact, and from the user point of view, (b) it is easy aspects of accuracy, reliability, and provides tools typically to remember and gives instantaneous access to all mod- unavailable in existing software alternatives, for instance: ules (cf. Fig. 2). Invoking a selected module is followed Fig. 1 The simplified architecture of the core MicMac modules. a the low-level and high-level modules dependencies, b the processing workflow. Marked in red are the outcome products, in green are the direct georeferencing inputs Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 3 of 9 Structure from motion Recovery of the structure (i.e. 3D coordinates of the observed scene) and the model. It is well known that this model (i.e. the transformation from 3D to 2D space and vice-versa also referred to as collinearity equation) is not linear, hence requires linearization. Moreover, there exists no direct algorithm able to compute orientation parameters globally consistent with a number of images (generally n > 3). To overcome this gap, bootstrap solu- tions were proposed. Using direct algorithms for a single image, a pair or triplet of images, the global orientation is deduced sequentially [3] or hierarchically [7] starting from Fig. 2 Each module is executed with some obligatory and optional a seed image pair. The so-obtained parameters serve as parameters. Adding -help at the end will print out the expected input to a system of equations composed of the linearized input arguments collinearity equations, where their optimal values – in the stochastic sense – are found iteratively. The observa- tions (e.g. tie points) are redundant therefore the solution by a number of parameters. The obligatory parameters is deduced with the least squares method by minimizing must be specified at first, and since they do not carry an objective function. The typical function is the squared names, their order of occurrence is important. In the differences between the nominal observations and those example shown in Fig. 2, the obligatory parameter would predicted from the estimated model, possibly subject to be the format specification, while the change coordinate constraints. The latter stage is also known as the bundle file, ChSys,servesasanoption. block adjustment (BBA). Implementation MicMac implementation. MicMac solves the BBA with The algorithmic aspects the Levenberg-Marquardt (L-M) method. The L - M is The photogrammetric workflow encompasses the passage in essence the Gauss-Newton method enriched with a from images, through the estimation of the orientation damping factor to handle rank-deficient (or nearly so) parameters, finalizing with a 3D surface model. In other Jacobian matrices [8]. A range of different observations words, it is a passage from a 2D representation of the equations, constraints and unknowns (i.e. parameters) are world captured by a camera, through inference of the posi- allowed in the adjustment routine, among them: tion and rotation of that camera at the moment of the image taking, towards a 3D restitution of the lost dimen- tie points, sion. As long as the quality of the final output depends lines, on the skillful engineering of many small processing steps, perspective centers derived from direct georeferencing (i.e. GNSS), estimation of the camera orientation parameters and the matching algorithms constitute the heart of the pipeline. ground control points (GCP), Respectively, the coming sections concentrate on these the lever-arm vector separating the GNSS antenna two aspects, report on the adopted methods and give a and the perspective center, global look on what is possible in MicMac. rigid relative orientation between cameras. Fig. 3 Three restitution geometries in MicMac. a ground, euclidean space, b image geometry discretised along the ray, no resampling c image space resampled to epipolar geometry. For a range of the potential Z-coordinates/depths/disparities the similarity measure is computed and further passed to the energy minimization solver Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 4 of 9 On top of that, a broad spectrum of camera calibration Multi-view stereo image matching (MVSM) models are available. Both argumentation are supported, Given several overlapping images with known orienta- the physical and the phenomenological. The former mod- tion parameters, MVSM is the process of reconstructing els any imaging error attributed to a phenomenon of a a complete 3D model of the scene by finding correspond- physical nature, while the latter models solely the effects ing pixels (i.e. their disparities or depths, respectively) of the lens imperfections, without interrogating its causes. between image pairs, and triangulating them in space. The The total distortion is defined as a composition of pre- generic algorithm is defined as an energy minimization defined elementary distortions, such as radial, decentric problem that searches for a disparity map minimizing the or a generic polynomial. Typically, the major distortion energy. The minimization can be solved with local, semi- are modeled by the physical model, while the generic global and global algorithms. Today the golden standard polynomial models remove the less significant remaining method for producing dense 3D models from images is systematism. the semi-global matching [6]. The critical element of the BBA’s mathematical model is Image matching breaks down into three processing the observation weighting, known as the stochastic model. stages: computation of the pixel matching cost, cost aggre- MicMac specifies three different weighting strategies. gation and disparity calculation. The matching cost is The first strategy corresponds to the weighting in the a measure of dissimilarity, i.e. describes the unlikeli- classical model known in photogrammetry – the Gauss- hood that two pixels belong to a unique point in 3D. Markov – where the observations are weighted by their As the matching can be ambiguous, and the minimum true standard deviations known a priori. The second strat- cost insignificant, an additional a priori is imposed on egy controls the influence of a particular category of the energy function that penalizes disparity changes at observations so as to avoid solutions that are driven by a neighboring pixels (the case of local and semi-global single category only for the sheer reason of its abundant algorithms). This aggregation takes place within a win- observations. The last strategy addresses the robustness. dow, a 1D path (locally) or multiple paths (semi-globally). The weight is a function of the observation’s residual in In the latter case, the cost for a given disparity is an accu- the BBA, therefore giving more credibility to observa- mulated cost along many paths that end in the pixel in tions that are close to the estimated model, and contrary, question [9]. limiting the influence of observations with high residuals. User’s viewpoint. There are two principal modules that MicMac implementation Image dense matching in handle the camera orientation recovery with the simpli- MicMac is possible with a version of semi-global and fied interface – Tapas, Campari – and both of the global algorithms. The objective of matching is to find modules call the parent orientation tool Apero (cf. Fig. 1). a mapping F : τ ⊗ ε . Depending on the restitution px px Tapas calculates the purely relative orientation of images, geometry, the τ can be the 3D euclidean space or the using observed tie points as the only input. Since at this image space of the master image, hence it is the space point there is no a priori on positions and rotations of the where the surface model is produced. The ε can be px cameras, Tapas also entails the initialization step where either the depth of the euclidean Z or the image disparity it progressively reconstructs the cameras with the help of (cf. Eq.1andFig.3). the direct orientation algorithms, and intertwines it with The motivation to distinguish between these two resti- the BBA routine. tution geometries – the ground and the image – is Unlike Tapas, Campari is a BBA module adapted twofold. On the one hand, a Digital Surface Model (DSM) to handle heterogeneous observations. Besides the tie produced from aerial images is normally defined in some points, it works with GCPs, the GNNS data and can reference CS, therefore it is more intuitive to perform the impose various constraints on the solution. This module reconstruction directly in the target CS (i.e. ground geom- is typically executed once a good relative orientation is etry), where the disparity space ε is the Z coordinate px established, and the cameras are moved two a coordinate (cf. Fig. 3a). On the other hand, it is known that match- system (CS) consistent with that of auxiliary observations, ing in image space with a master and a set of slaves images e.g. the GCPs. The latter is performed with any Bascule is more reliable, especially in close-range photogramme- tool, and it is a spatial similarity transformation (cf. Fig. 1). try applications and for small datasets. The disparity space Both Campari and all variations of the Bascule can be ε in the image geometry is then either the depth along px regarded as georeferencing tools. the ray or the respective disparity in image space (i.e. the Since recently, MicMac provides the user with the inter- image geometry; cf. Fig. 3b and c). It is up to the user nal accuracy estimates of the BBA parameters (their stan- which geometry to employ. dard deviations and correlations) derived from covariance MicMac offers a very flexible formulation of the match- matrices. ing cost function as far as its terms are concerned, the Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 5 of 9 Fig. 4 Bas Relief stone in Louvre: the image and the 3D model rendered in shading and depth map reg optimization algorithm and the dimension of the match- The ∇(F ) controls the a priori on the disparity px ing are regarded (cf. Eq. 1). space in a way that: if the desired model is smooth, a convex function F is reg E (F ) = A(x, y, F (x, y)) +∇(F ) (1) px px px adequate (it’s better to climb up a given jump by regular steps), reg 1 2 ∇(F ) = α ∗|∇ F |+ α ∗|∇ F | (2) • if the desired model has many discontinuities, a px 1 px 2 px where A(x, y, F (x, y)) is the measure of similarity concave function F is adequate (it’s better to climb px between pixels with A = 0 when they are identical; up a given jump in a single step), reg ∇(F ) a norm on the gradient, which is used as a when there is no strong prior, the default choice is to px regularity criteria (it penalizes the variation of F ); α have F linear, px 1 is the regularization on first component of disparity and if a priori knowledge on scene slope exists, it can α the regularization on the second component (equiva- impose an allowable maximum scene slope, lent of matching in the direction transverse to the epipolar for 2D matching, non-isotropic smoothing factors line). can be set. Fig. 5 Under induced force the concrete beam deformed, while a still camera measured the displacements in during the breaking phase. Upper left: the full view image, upper right: a close-up of the concrete structure, bottom left: displacement in the x-coordinate (max amplitude pixel), bottom right: displacement in the y-coordinate (max amplitude pixel) 4 Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 6 of 9 Fig. 6 Munich dataset, acquired with a DMC camera at GSD of 10 cm. Left: the depth map, right: the y-parallax generated with MMTestOrient. The amplitude of transverse parallax is ± 0.1 pixel The actual similarity measure A(x, y, F (x, y)) is calcu- User’s viewpoint DSM creation in MicMac takes place px lated from the the normalized cross correlation coefficient via several tools that are semi- or fully-automated. In (1 − Cor), defined as a function of multiple images, rather both cases the parent tool to handle the low-level dense than a single pair. The coefficient can privilege an image matching is the MicMac, which an expert user can access out of a set of images (e.g. a master image), or consider via an XML file. the set as “symmetrical”. Varying cross correlation window Malt, the semi-automated tool, operates in three pre- sizes and weighted windows are also possible. defined modes, i.e. Ortho, UrbanMNE and GeomImage. To find the disparity map that minimizes the energy, The modes are adapted to producing orthophotography, MicMac by default employs a multi-directional variant of DSMs in ground and image geometries. Accordingly, the dynamic programming algorithm. The results along Ortho and UrbanMNE would be selected for flat ter- independent directions are aggregated by taking a mean rains or planar surfaces (i.e. 2.5D; cf. Figs. 4, 5, 6) and of all results or simply the best result. Optionally, a global GeomImage fortrue3Dsurfaces(cf.Fig.7). minimization by the Min-Cut/Max-Flow algorithm [10] The inconvenience of the Malt GeomImage is that can be enforced. for a complete 3D restitution of a surface, one must In order to limit the disparity search-space, speed indicate several master and their relevant slave images. up the calculation and to reduce the noise, MicMac As long as for simple object geometries it is feasible, adopts a multi-resolution matching approach where it becomes inefficient for more complex and exten- coarse resolution disparity maps serve as predictors of the sive objects. The fully automated tools overcome this fine resolution output. impediment by making this choice for the user, while Fig. 7 Indoor architecture: Chapelle imperiale Ajaccio, with 100 fish-eye images; left: retrieved camera poses, right:3Dmodel Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 7 of 9 Fig. 8 Forteresse de Salses, photo acquired from a drone, in collaboration with Map-CNRS/ Left: hypsometry and shading maps overlapped, right:an orthophotography prioritising completeness and precision of the output. The a 2D matching strategy. This strategy is useful in two automated tools are C3DC and PIMs which furnish 2D deformation studies between rectified images (cf. a point cloud and a set of depth maps, respectively. To pro- Fig. 5; also MM2DPosSism), as an orientation qual- ceed with the orthophotomap generation, the PIMs2MNT ity estimation tool to asses the remaining y-parallax and subsequently the Tawny shall be executed (an exem- (MMTestOrient; cf. Fig. 6), or in cases when orientation plar orthophoto in Fig. 8). The latter collects the per- parameters are unknown or known with poor precision image orthophotos and merges it to a single mosaic. (executed with XML file). Because the expected dispari- The individual orthophotos can be created either by ties pertinent to geometry are of higher frequencies than PIMs2MNT tool or Malt;cf. Fig. 1. those along the y-parallax, the regularization of the energy Besides the classical 1D matching (along an image function in the two direction is managed separately; cf. ray, an image row or the Z-coordinate) for recon- Eq. (2). struction of object geometry, MicMac also implements Lastly, in the event where all camera optical centers are located on a 3D line, parallel to the surveyed object, and the reconstruction is defined in object space, the match- ing becomes ambiguous along the epipolar lines. To avoid the ambiguities and the resultant noise, MicMac calcu- lates the matching in ortho-cylindrical coordinates (see RepLocBascule). Interactive tools Although MicMac by default is based on simplified com- mand lines, for some processes – i.e. image measure- ments, visualization of the results – a graphical user interface (GUI) is available. Besides this, for multiple com- mands there exists a GUI that replaces the standard com- mand line. To run it, the user should simply type the name of the command preceded by the letter ’v’. For example, in the GrShade command, the GUI is called by typing vGrShade (cf. Fig 9). All results produced by MicMac are exclusively in open data formats (tif, xml, ply) and as such it opens up the possibility to interact with the processing chain at any desirable stage. For instance, the user can import interior and exterior image orientation data and proceed with the dense matching and orthophoto generation. Or vice versa, the orientation parameters or epipolar images can Fig. 9 Graphical interface to launch commands be generated to proceed with an alternative dense image Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 8 of 9 Fig. 10 Left: drawing of a 2D mask on an image; right: and a 3D mask in a point cloud matcher. The reader is advised to refer to the software Fig. 10). Masks can be used to, e.g., remove tie points (cf. documentation [11] for more details on the conversion HomolFilterMasq), or to limit the image matching to a procedures. selected part. The SaisieAppuisInitQT/SaisieAppuisPred- Image measurement tools icQT tools allow for GCPs’ image measurements (cf. In the following, several tools for image measurements Fig 11). Provided that the camera orientation parameters and the use of 2D/3D masks are discussed. are known with a reasonable accuracy and expressed Many of the interactive tools are available in a ver- in a CS coherent with the GPCs, the latter tool serves sion requiring the installation of the QT library. For as a point position prediction tool. Measurements nec- example, for the SaisieMasq, the corresponding QT essary to perform a spatial similarity transformation, tool is the SaisieMasqQT. In either case, the mask- i.e. the origin, the axes directions and distances, can ing works by creating a binary image mask from a be collected within another georeferencing tool – the 2D polygon, or a 3D cylinder from a 3D mask (cf. SaisieBascQT. These tools normally precede the Fig. 11 Interface for measuring GCPs in images Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 9 of 9 georeferencing operations with GCPBascule and/or Funding The project is funded by : Campari. Institut National de l’Information Geographique et Forestiere (IGN main Visualization tools funder since 2003) The French FUI Project “Culture 3D Cloud” To visualize the intermediary and final results MicMac The French ANR Project “MONUMENTUM” offers a range of different tools. Centre national d’études spatiales (CNES) via the TOSCA programme The SEL tool visualizes the tie points between a stereo Authors’ contributions pair. The AperiCloud tool converts the result of the As long as all authors equally contributed to this article, the software itself was BBA, that is the poses and 3D tie points, to the point cloud in major part developped by MPD. All authors read and approved the final ply format. manuscript. The Vino tool is a simple image viewer adapted to Competing interests visualization of very large images (i.e. in size of a few The authors declare that they have no competing interests. gigabytes), irrespective of their type (i.e. integer, float or >8Bit). Moreover, it lets the user to modify the histogram Publisher’s Note of the image or generate an image crop. Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Other tools allow a more intuitive visualization of depth maps. For instance, the to8bits tool converts 32-bit or Author details 16-bit images to 8-bit images, while the GrShade tool ENSG, École Nationale des Sciences Géographiques, Champs-sur-Marne, France. IPGP – Institut de Physique du Globe de Paris, Sorbonne Paris Cité, computes shading from a depth map image (cf. Fig 4). Univ Paris Diderot, CNRS, Paris, France. Vinci-Construction-Terrassement, Lastly, the Nuage2Ply transforms a depth map into a Sixense Mapping, Morangis, France. point cloud ply format (cf. Fig 7). Received: 20 December 2016 Accepted: 1 May 2017 Discussion MicMac is an open-source project for photogrammetry. References As a research implementation platform, it serves the IGN’s 1. Granshaw SI, Fraser CS. Editorial: Computer vision and photogrammetry: internal needs as well as aims at bringing the state-of- Interaction or introspection? Photogrammetric Rec. 2015;30(149):3–7. 2. Lowe DG. Distinctive image features from scale-invariant keypoints. Int J the-art 3D image processing algorithms to a wider audi- Comput Vis. 2004;60(2):91–110. doi:10.1023/B:VISI.0000029664.99615.94. ence, be it universities, public institutions or enterprises. 3. Snavely N, Seitz SM, Szeliski R. Photo tourism: exploring photo Like any other open-source project, MicMac authorizes collections in 3d. In: ACM Transactions on Graphics (TOG), vol. 25. ACM; 2006. p. 835–46. http://dl.acm.org/citation.cfm?id=1141964. the volunteers to contribute or change the source code 4. Pierrot-Deseilligny M, Clery I. Apero, an open source bundle adjustment according to their needs. For more information please software for automatic calibration and orientation of set of images. refer to: 2011;38(5):269–76. 5. Pierrot-Deseilligny M, Paparoditis N. A multiresolution and optimization-based image matching approach: An application to surface http://micmac.ensg.eu, as the software’s reference reconstruction from spot5-hrs stereo imagery. Arch Photogramm Remote webpage; Sens Spat Inf Sci. 2006;36(1/W41). • 6. Hirschmuller H. Stereo processing by semiglobal matching and mutual http://forum-micmac.forumprod.com/, for technical information. IEEE Trans Pattern Anal Mach Intell. 2008;30(2):328–41. support; 7. Toldo R, Gherardi R, Farenzena M, Fusiello A. Hierarchical https://github.com/micmacIGN/, for the source code. structure-and-motion recovery from uncalibrated images. Comput Vis Image Underst. 2015;140:127–43. 8. Nocedal J, Wright S. Numerical Optimization: Springer; 2006. http://link. The high priority on-going developments and short- springer.com/book/10.1007/978-0-387-40065-5. term perspectives are those concerning effective process- 9. Szeliski R. Computer Vision: Algorithms and Applications: Springer; 2010. http://dl.acm.org/citation.cfm?id=1941882. ing of big datasets. The developments pertain to (a) the 10. Roy S, Cox IJ. A maximum-flow formulation of the n-camera stereo tie points extraction, (b) the SfM,aswellas(c) thestorage correspondence problem. In: Computer Vision, 1998. Sixth International of 3D data. Regarding the tie points, very precise tie point Conference On. IEEE; 1998. p. 492–9. http://ieeexplore.ieee.org/abstract/ document/710763/. detectors and detectors invariant to affine distortions are 11. MicMac, Apero, Pastis and Other Beverages in a Nutshell. https://github. being developed. Within the SfM, global, structure-less com/micmacIGN/Documentation/blob/master/DocMicMac.pdf. methods [12–14] are under development. As a further Accessed 11 May 2017. 12. Enqvist O, Kahl F, Olsson C. Non-sequential structure from motion. In: perspective, adequate methods for storing very large 3D Computer Vision Workshops (ICCV Workshops), 2011 IEEE International point clouds will be conceived. Lastly, as there has been Conference On. IEEE; 2011. p. 264–71. http://ieeexplore.ieee.org/ numerous demands to create a GUI, computer program- abstract/document/6130252/. 13. Moulon P, Monasse P, Marlet R. Global fusion of relative motions for mers willing to contribute are strongly encouraged to robust, accurate and scalable structure from motion. In: Proceedings of contact the team of developers. A GUI in form of a the IEEE International Conference on Computer Vision; 2013. p. 3248–255. stand-alone application or a GUI integrated with other 14. Reich M, Yang MY, Heipke C. Global robust image rotation from combined open-source software tools, e.g. QGIS, are possible. weighted averaging. ISPRS J Photogrammetry Remote Sens. 2017. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Open Geospatial Data, Software and Standards Springer Journals

MicMac – a free, open-source solution for photogrammetry

Loading next page...
 
/lp/springer-journals/micmac-a-free-open-source-solution-for-photogrammetry-9WN8lXHxXr

References (15)

Publisher
Springer Journals
Copyright
Copyright © 2017 by The Author(s)
Subject
Geography; Geographical Information Systems/Cartography; Information Systems Applications (incl.Internet); Communications Engineering, Networks
eISSN
2363-7501
DOI
10.1186/s40965-017-0027-2
Publisher site
See Article on Publisher Site

Abstract

The publication familiarizes the reader with MicMac - a free, open-source photogrammetric software for 3D reconstruction. A brief history of the tool, its organisation and unique features vis-à-vis other software tools are in the highlight. The essential algorithmic aspects of the structure from motion and image dense matching problems are discussed from the implementation and the user’s viewpoints. Keywords: Free, Open-source, Photogrammetry, Structure from motion, Bundle block adjustment, Semi-global image dense matching Background or free/open-source. MicMac – together with Bundler, Photogrammetry is the art, science, and technology of PMVS, VisualSfM, openMVG, OpenCV and others – obtaining geometric information on the 3-dimensional belongs to the open-source solutions . This publication shape and orientation of objects from images and other aims at familiarizing the reader with the philosophy imaging sensors. It is a cheap measurement method- behind MicMac, some of its crucial algorithmic aspects, ology as it can be executed by practically any digital the software architecture and the pool of available tools. camera of a decent make. It is instantaneous as it cap- tures the observed phenomena at once and in a split Historical notes second, and highly automated, therefore accessible to MicMac has been developed at the National Institute non-expert users. Thanks to the field of computer vision, of Geographic and Forestry Information (IGN) and the photogrammetry rejuvenated and today places among National School of Geographic Sciences (ENSG), since other competitive remote sensing techniques (e.g. Light 2003. Initially, the software tools were developed having in Detection and Ranging LiDAR) [1]. The several mile- mind exclusively the IGN’s cartographic production. The stones leading to this progress are the automated interest independent tools were interfaced in 2005 via an XML points detection [2], the Structure from Motion (SfM) framework, allowing the user to freely parametrize the algorithms capable of reconstructing scenes from sets of calculations at all processing stages. In 2007, IGN began to unordered image collections [3, 4], and the dense image freely distribute MicMac under the CECILL-B license that matching techniques delivering surface models of resolu- is a version of the L-GPL license adapted to the french law. tion equal to the pixel size on the ground [5, 6]. Until 2008, the dense image matching of already ori- All this contributes to an ever growing visibility of ented images was possible only with the IGN’s internal the photogrammetric tools across various fields of sci- image file format. In the same year the Apero tool was ence and engineering, the growing market interest, and added to the sofware kernel, offering from now on the subsequently a multitude of photogrammetric/computer possibility to estimate camera exterior and interior orien- vision libraries and software solutions, be it commercial tations, with no restriction on the image format. In 2010, the XML interface was replaced by a simplified command *Correspondence: erupnik@gmail.com line. This evolution contributed to an improved accessi- ENSG, École Nationale des Sciences Géographiques, Champs-sur-Marne, bility, diffusion and subsequently a better visibility of the France 2 software in the scientific communities, and the general IPGP – Institut de Physique du Globe de Paris, Sorbonne Paris Cité, Univ Paris Diderot, CNRS, Paris, France public. Full list of author information is available at the end of the article © The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 2 of 9 Since 2010 MicMac has been undergoing a significant access to intermediary results in open data formats evolution due to its implication in many french and euro- allowing to interact with the processing chain at any pean projects. Besides the contributions to the source desired stage, code, new software distributions under Windows, Mac qualitative evaluation of the results via quality OSX, as well as the GPU processing for certain tasks indicators (e.g. bundle block adjustment residuals, became available. covariance matrices, parameter’s sensitivity, correlation images), MicMac vis-à-vis other tools a wide range of camera calibration models (e.g. the From the architecture point of view, the unique feature models adapted to consumer grade cameras, of MicMac with respect to its alternatives is that the large-frame aerial cameras, cameras with very long user can operate on different software levels, cf. Fig. 1. focal lengths, fish-eye and spherical cameras), The least experienced user can steer the processing by two-dimensional dense image matching for a simple command line where little or no parameters deformation studies, must be set. An experienced user can operate from the processing of frame camera and pushbroom sensor same command line by adapting the parameters typically images, filled with default values. An expert in photogramme- processing of scanned analogue images, try can access any parameter at any processing step via architecture adapted to big datasets. a set of XML files (cf. Fig. 1). On top of that, develop- ers and scientists can use MicMac as a library, implement their own algorithms, and share them for the benefit Software organisation of photogrammetric community. This said, MicMac was MicMac is organised in a number of modules, accessi- designed primarily as a tool for professionals and con- ble through a common command mm3d.Execution with a sequently is far from a model design in the sense of unique command has a twofold advantage: (a) it allows to HCI . factorize certain developments while keeping the binary MicMac, as a photogrammetric tool, stresses the files compact, and from the user point of view, (b) it is easy aspects of accuracy, reliability, and provides tools typically to remember and gives instantaneous access to all mod- unavailable in existing software alternatives, for instance: ules (cf. Fig. 2). Invoking a selected module is followed Fig. 1 The simplified architecture of the core MicMac modules. a the low-level and high-level modules dependencies, b the processing workflow. Marked in red are the outcome products, in green are the direct georeferencing inputs Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 3 of 9 Structure from motion Recovery of the structure (i.e. 3D coordinates of the observed scene) and the model. It is well known that this model (i.e. the transformation from 3D to 2D space and vice-versa also referred to as collinearity equation) is not linear, hence requires linearization. Moreover, there exists no direct algorithm able to compute orientation parameters globally consistent with a number of images (generally n > 3). To overcome this gap, bootstrap solu- tions were proposed. Using direct algorithms for a single image, a pair or triplet of images, the global orientation is deduced sequentially [3] or hierarchically [7] starting from Fig. 2 Each module is executed with some obligatory and optional a seed image pair. The so-obtained parameters serve as parameters. Adding -help at the end will print out the expected input to a system of equations composed of the linearized input arguments collinearity equations, where their optimal values – in the stochastic sense – are found iteratively. The observa- tions (e.g. tie points) are redundant therefore the solution by a number of parameters. The obligatory parameters is deduced with the least squares method by minimizing must be specified at first, and since they do not carry an objective function. The typical function is the squared names, their order of occurrence is important. In the differences between the nominal observations and those example shown in Fig. 2, the obligatory parameter would predicted from the estimated model, possibly subject to be the format specification, while the change coordinate constraints. The latter stage is also known as the bundle file, ChSys,servesasanoption. block adjustment (BBA). Implementation MicMac implementation. MicMac solves the BBA with The algorithmic aspects the Levenberg-Marquardt (L-M) method. The L - M is The photogrammetric workflow encompasses the passage in essence the Gauss-Newton method enriched with a from images, through the estimation of the orientation damping factor to handle rank-deficient (or nearly so) parameters, finalizing with a 3D surface model. In other Jacobian matrices [8]. A range of different observations words, it is a passage from a 2D representation of the equations, constraints and unknowns (i.e. parameters) are world captured by a camera, through inference of the posi- allowed in the adjustment routine, among them: tion and rotation of that camera at the moment of the image taking, towards a 3D restitution of the lost dimen- tie points, sion. As long as the quality of the final output depends lines, on the skillful engineering of many small processing steps, perspective centers derived from direct georeferencing (i.e. GNSS), estimation of the camera orientation parameters and the matching algorithms constitute the heart of the pipeline. ground control points (GCP), Respectively, the coming sections concentrate on these the lever-arm vector separating the GNSS antenna two aspects, report on the adopted methods and give a and the perspective center, global look on what is possible in MicMac. rigid relative orientation between cameras. Fig. 3 Three restitution geometries in MicMac. a ground, euclidean space, b image geometry discretised along the ray, no resampling c image space resampled to epipolar geometry. For a range of the potential Z-coordinates/depths/disparities the similarity measure is computed and further passed to the energy minimization solver Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 4 of 9 On top of that, a broad spectrum of camera calibration Multi-view stereo image matching (MVSM) models are available. Both argumentation are supported, Given several overlapping images with known orienta- the physical and the phenomenological. The former mod- tion parameters, MVSM is the process of reconstructing els any imaging error attributed to a phenomenon of a a complete 3D model of the scene by finding correspond- physical nature, while the latter models solely the effects ing pixels (i.e. their disparities or depths, respectively) of the lens imperfections, without interrogating its causes. between image pairs, and triangulating them in space. The The total distortion is defined as a composition of pre- generic algorithm is defined as an energy minimization defined elementary distortions, such as radial, decentric problem that searches for a disparity map minimizing the or a generic polynomial. Typically, the major distortion energy. The minimization can be solved with local, semi- are modeled by the physical model, while the generic global and global algorithms. Today the golden standard polynomial models remove the less significant remaining method for producing dense 3D models from images is systematism. the semi-global matching [6]. The critical element of the BBA’s mathematical model is Image matching breaks down into three processing the observation weighting, known as the stochastic model. stages: computation of the pixel matching cost, cost aggre- MicMac specifies three different weighting strategies. gation and disparity calculation. The matching cost is The first strategy corresponds to the weighting in the a measure of dissimilarity, i.e. describes the unlikeli- classical model known in photogrammetry – the Gauss- hood that two pixels belong to a unique point in 3D. Markov – where the observations are weighted by their As the matching can be ambiguous, and the minimum true standard deviations known a priori. The second strat- cost insignificant, an additional a priori is imposed on egy controls the influence of a particular category of the energy function that penalizes disparity changes at observations so as to avoid solutions that are driven by a neighboring pixels (the case of local and semi-global single category only for the sheer reason of its abundant algorithms). This aggregation takes place within a win- observations. The last strategy addresses the robustness. dow, a 1D path (locally) or multiple paths (semi-globally). The weight is a function of the observation’s residual in In the latter case, the cost for a given disparity is an accu- the BBA, therefore giving more credibility to observa- mulated cost along many paths that end in the pixel in tions that are close to the estimated model, and contrary, question [9]. limiting the influence of observations with high residuals. User’s viewpoint. There are two principal modules that MicMac implementation Image dense matching in handle the camera orientation recovery with the simpli- MicMac is possible with a version of semi-global and fied interface – Tapas, Campari – and both of the global algorithms. The objective of matching is to find modules call the parent orientation tool Apero (cf. Fig. 1). a mapping F : τ ⊗ ε . Depending on the restitution px px Tapas calculates the purely relative orientation of images, geometry, the τ can be the 3D euclidean space or the using observed tie points as the only input. Since at this image space of the master image, hence it is the space point there is no a priori on positions and rotations of the where the surface model is produced. The ε can be px cameras, Tapas also entails the initialization step where either the depth of the euclidean Z or the image disparity it progressively reconstructs the cameras with the help of (cf. Eq.1andFig.3). the direct orientation algorithms, and intertwines it with The motivation to distinguish between these two resti- the BBA routine. tution geometries – the ground and the image – is Unlike Tapas, Campari is a BBA module adapted twofold. On the one hand, a Digital Surface Model (DSM) to handle heterogeneous observations. Besides the tie produced from aerial images is normally defined in some points, it works with GCPs, the GNNS data and can reference CS, therefore it is more intuitive to perform the impose various constraints on the solution. This module reconstruction directly in the target CS (i.e. ground geom- is typically executed once a good relative orientation is etry), where the disparity space ε is the Z coordinate px established, and the cameras are moved two a coordinate (cf. Fig. 3a). On the other hand, it is known that match- system (CS) consistent with that of auxiliary observations, ing in image space with a master and a set of slaves images e.g. the GCPs. The latter is performed with any Bascule is more reliable, especially in close-range photogramme- tool, and it is a spatial similarity transformation (cf. Fig. 1). try applications and for small datasets. The disparity space Both Campari and all variations of the Bascule can be ε in the image geometry is then either the depth along px regarded as georeferencing tools. the ray or the respective disparity in image space (i.e. the Since recently, MicMac provides the user with the inter- image geometry; cf. Fig. 3b and c). It is up to the user nal accuracy estimates of the BBA parameters (their stan- which geometry to employ. dard deviations and correlations) derived from covariance MicMac offers a very flexible formulation of the match- matrices. ing cost function as far as its terms are concerned, the Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 5 of 9 Fig. 4 Bas Relief stone in Louvre: the image and the 3D model rendered in shading and depth map reg optimization algorithm and the dimension of the match- The ∇(F ) controls the a priori on the disparity px ing are regarded (cf. Eq. 1). space in a way that: if the desired model is smooth, a convex function F is reg E (F ) = A(x, y, F (x, y)) +∇(F ) (1) px px px adequate (it’s better to climb up a given jump by regular steps), reg 1 2 ∇(F ) = α ∗|∇ F |+ α ∗|∇ F | (2) • if the desired model has many discontinuities, a px 1 px 2 px where A(x, y, F (x, y)) is the measure of similarity concave function F is adequate (it’s better to climb px between pixels with A = 0 when they are identical; up a given jump in a single step), reg ∇(F ) a norm on the gradient, which is used as a when there is no strong prior, the default choice is to px regularity criteria (it penalizes the variation of F ); α have F linear, px 1 is the regularization on first component of disparity and if a priori knowledge on scene slope exists, it can α the regularization on the second component (equiva- impose an allowable maximum scene slope, lent of matching in the direction transverse to the epipolar for 2D matching, non-isotropic smoothing factors line). can be set. Fig. 5 Under induced force the concrete beam deformed, while a still camera measured the displacements in during the breaking phase. Upper left: the full view image, upper right: a close-up of the concrete structure, bottom left: displacement in the x-coordinate (max amplitude pixel), bottom right: displacement in the y-coordinate (max amplitude pixel) 4 Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 6 of 9 Fig. 6 Munich dataset, acquired with a DMC camera at GSD of 10 cm. Left: the depth map, right: the y-parallax generated with MMTestOrient. The amplitude of transverse parallax is ± 0.1 pixel The actual similarity measure A(x, y, F (x, y)) is calcu- User’s viewpoint DSM creation in MicMac takes place px lated from the the normalized cross correlation coefficient via several tools that are semi- or fully-automated. In (1 − Cor), defined as a function of multiple images, rather both cases the parent tool to handle the low-level dense than a single pair. The coefficient can privilege an image matching is the MicMac, which an expert user can access out of a set of images (e.g. a master image), or consider via an XML file. the set as “symmetrical”. Varying cross correlation window Malt, the semi-automated tool, operates in three pre- sizes and weighted windows are also possible. defined modes, i.e. Ortho, UrbanMNE and GeomImage. To find the disparity map that minimizes the energy, The modes are adapted to producing orthophotography, MicMac by default employs a multi-directional variant of DSMs in ground and image geometries. Accordingly, the dynamic programming algorithm. The results along Ortho and UrbanMNE would be selected for flat ter- independent directions are aggregated by taking a mean rains or planar surfaces (i.e. 2.5D; cf. Figs. 4, 5, 6) and of all results or simply the best result. Optionally, a global GeomImage fortrue3Dsurfaces(cf.Fig.7). minimization by the Min-Cut/Max-Flow algorithm [10] The inconvenience of the Malt GeomImage is that can be enforced. for a complete 3D restitution of a surface, one must In order to limit the disparity search-space, speed indicate several master and their relevant slave images. up the calculation and to reduce the noise, MicMac As long as for simple object geometries it is feasible, adopts a multi-resolution matching approach where it becomes inefficient for more complex and exten- coarse resolution disparity maps serve as predictors of the sive objects. The fully automated tools overcome this fine resolution output. impediment by making this choice for the user, while Fig. 7 Indoor architecture: Chapelle imperiale Ajaccio, with 100 fish-eye images; left: retrieved camera poses, right:3Dmodel Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 7 of 9 Fig. 8 Forteresse de Salses, photo acquired from a drone, in collaboration with Map-CNRS/ Left: hypsometry and shading maps overlapped, right:an orthophotography prioritising completeness and precision of the output. The a 2D matching strategy. This strategy is useful in two automated tools are C3DC and PIMs which furnish 2D deformation studies between rectified images (cf. a point cloud and a set of depth maps, respectively. To pro- Fig. 5; also MM2DPosSism), as an orientation qual- ceed with the orthophotomap generation, the PIMs2MNT ity estimation tool to asses the remaining y-parallax and subsequently the Tawny shall be executed (an exem- (MMTestOrient; cf. Fig. 6), or in cases when orientation plar orthophoto in Fig. 8). The latter collects the per- parameters are unknown or known with poor precision image orthophotos and merges it to a single mosaic. (executed with XML file). Because the expected dispari- The individual orthophotos can be created either by ties pertinent to geometry are of higher frequencies than PIMs2MNT tool or Malt;cf. Fig. 1. those along the y-parallax, the regularization of the energy Besides the classical 1D matching (along an image function in the two direction is managed separately; cf. ray, an image row or the Z-coordinate) for recon- Eq. (2). struction of object geometry, MicMac also implements Lastly, in the event where all camera optical centers are located on a 3D line, parallel to the surveyed object, and the reconstruction is defined in object space, the match- ing becomes ambiguous along the epipolar lines. To avoid the ambiguities and the resultant noise, MicMac calcu- lates the matching in ortho-cylindrical coordinates (see RepLocBascule). Interactive tools Although MicMac by default is based on simplified com- mand lines, for some processes – i.e. image measure- ments, visualization of the results – a graphical user interface (GUI) is available. Besides this, for multiple com- mands there exists a GUI that replaces the standard com- mand line. To run it, the user should simply type the name of the command preceded by the letter ’v’. For example, in the GrShade command, the GUI is called by typing vGrShade (cf. Fig 9). All results produced by MicMac are exclusively in open data formats (tif, xml, ply) and as such it opens up the possibility to interact with the processing chain at any desirable stage. For instance, the user can import interior and exterior image orientation data and proceed with the dense matching and orthophoto generation. Or vice versa, the orientation parameters or epipolar images can Fig. 9 Graphical interface to launch commands be generated to proceed with an alternative dense image Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 8 of 9 Fig. 10 Left: drawing of a 2D mask on an image; right: and a 3D mask in a point cloud matcher. The reader is advised to refer to the software Fig. 10). Masks can be used to, e.g., remove tie points (cf. documentation [11] for more details on the conversion HomolFilterMasq), or to limit the image matching to a procedures. selected part. The SaisieAppuisInitQT/SaisieAppuisPred- Image measurement tools icQT tools allow for GCPs’ image measurements (cf. In the following, several tools for image measurements Fig 11). Provided that the camera orientation parameters and the use of 2D/3D masks are discussed. are known with a reasonable accuracy and expressed Many of the interactive tools are available in a ver- in a CS coherent with the GPCs, the latter tool serves sion requiring the installation of the QT library. For as a point position prediction tool. Measurements nec- example, for the SaisieMasq, the corresponding QT essary to perform a spatial similarity transformation, tool is the SaisieMasqQT. In either case, the mask- i.e. the origin, the axes directions and distances, can ing works by creating a binary image mask from a be collected within another georeferencing tool – the 2D polygon, or a 3D cylinder from a 3D mask (cf. SaisieBascQT. These tools normally precede the Fig. 11 Interface for measuring GCPs in images Rupnik et al. Open Geospatial Data, Software and Standards (2017) 2:14 Page 9 of 9 georeferencing operations with GCPBascule and/or Funding The project is funded by : Campari. Institut National de l’Information Geographique et Forestiere (IGN main Visualization tools funder since 2003) The French FUI Project “Culture 3D Cloud” To visualize the intermediary and final results MicMac The French ANR Project “MONUMENTUM” offers a range of different tools. Centre national d’études spatiales (CNES) via the TOSCA programme The SEL tool visualizes the tie points between a stereo Authors’ contributions pair. The AperiCloud tool converts the result of the As long as all authors equally contributed to this article, the software itself was BBA, that is the poses and 3D tie points, to the point cloud in major part developped by MPD. All authors read and approved the final ply format. manuscript. The Vino tool is a simple image viewer adapted to Competing interests visualization of very large images (i.e. in size of a few The authors declare that they have no competing interests. gigabytes), irrespective of their type (i.e. integer, float or >8Bit). Moreover, it lets the user to modify the histogram Publisher’s Note of the image or generate an image crop. Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Other tools allow a more intuitive visualization of depth maps. For instance, the to8bits tool converts 32-bit or Author details 16-bit images to 8-bit images, while the GrShade tool ENSG, École Nationale des Sciences Géographiques, Champs-sur-Marne, France. IPGP – Institut de Physique du Globe de Paris, Sorbonne Paris Cité, computes shading from a depth map image (cf. Fig 4). Univ Paris Diderot, CNRS, Paris, France. Vinci-Construction-Terrassement, Lastly, the Nuage2Ply transforms a depth map into a Sixense Mapping, Morangis, France. point cloud ply format (cf. Fig 7). Received: 20 December 2016 Accepted: 1 May 2017 Discussion MicMac is an open-source project for photogrammetry. References As a research implementation platform, it serves the IGN’s 1. Granshaw SI, Fraser CS. Editorial: Computer vision and photogrammetry: internal needs as well as aims at bringing the state-of- Interaction or introspection? Photogrammetric Rec. 2015;30(149):3–7. 2. Lowe DG. Distinctive image features from scale-invariant keypoints. Int J the-art 3D image processing algorithms to a wider audi- Comput Vis. 2004;60(2):91–110. doi:10.1023/B:VISI.0000029664.99615.94. ence, be it universities, public institutions or enterprises. 3. Snavely N, Seitz SM, Szeliski R. Photo tourism: exploring photo Like any other open-source project, MicMac authorizes collections in 3d. In: ACM Transactions on Graphics (TOG), vol. 25. ACM; 2006. p. 835–46. http://dl.acm.org/citation.cfm?id=1141964. the volunteers to contribute or change the source code 4. Pierrot-Deseilligny M, Clery I. Apero, an open source bundle adjustment according to their needs. For more information please software for automatic calibration and orientation of set of images. refer to: 2011;38(5):269–76. 5. Pierrot-Deseilligny M, Paparoditis N. A multiresolution and optimization-based image matching approach: An application to surface http://micmac.ensg.eu, as the software’s reference reconstruction from spot5-hrs stereo imagery. Arch Photogramm Remote webpage; Sens Spat Inf Sci. 2006;36(1/W41). • 6. Hirschmuller H. Stereo processing by semiglobal matching and mutual http://forum-micmac.forumprod.com/, for technical information. IEEE Trans Pattern Anal Mach Intell. 2008;30(2):328–41. support; 7. Toldo R, Gherardi R, Farenzena M, Fusiello A. Hierarchical https://github.com/micmacIGN/, for the source code. structure-and-motion recovery from uncalibrated images. Comput Vis Image Underst. 2015;140:127–43. 8. Nocedal J, Wright S. Numerical Optimization: Springer; 2006. http://link. The high priority on-going developments and short- springer.com/book/10.1007/978-0-387-40065-5. term perspectives are those concerning effective process- 9. Szeliski R. Computer Vision: Algorithms and Applications: Springer; 2010. http://dl.acm.org/citation.cfm?id=1941882. ing of big datasets. The developments pertain to (a) the 10. Roy S, Cox IJ. A maximum-flow formulation of the n-camera stereo tie points extraction, (b) the SfM,aswellas(c) thestorage correspondence problem. In: Computer Vision, 1998. Sixth International of 3D data. Regarding the tie points, very precise tie point Conference On. IEEE; 1998. p. 492–9. http://ieeexplore.ieee.org/abstract/ document/710763/. detectors and detectors invariant to affine distortions are 11. MicMac, Apero, Pastis and Other Beverages in a Nutshell. https://github. being developed. Within the SfM, global, structure-less com/micmacIGN/Documentation/blob/master/DocMicMac.pdf. methods [12–14] are under development. As a further Accessed 11 May 2017. 12. Enqvist O, Kahl F, Olsson C. Non-sequential structure from motion. In: perspective, adequate methods for storing very large 3D Computer Vision Workshops (ICCV Workshops), 2011 IEEE International point clouds will be conceived. Lastly, as there has been Conference On. IEEE; 2011. p. 264–71. http://ieeexplore.ieee.org/ numerous demands to create a GUI, computer program- abstract/document/6130252/. 13. Moulon P, Monasse P, Marlet R. Global fusion of relative motions for mers willing to contribute are strongly encouraged to robust, accurate and scalable structure from motion. In: Proceedings of contact the team of developers. A GUI in form of a the IEEE International Conference on Computer Vision; 2013. p. 3248–255. stand-alone application or a GUI integrated with other 14. Reich M, Yang MY, Heipke C. Global robust image rotation from combined open-source software tools, e.g. QGIS, are possible. weighted averaging. ISPRS J Photogrammetry Remote Sens. 2017.

Journal

Open Geospatial Data, Software and StandardsSpringer Journals

Published: Jun 5, 2017

There are no references for this article.