Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/hindawi-publishing-corporation/unmanned-aerial-vehicle-surveying-and-mapping-trajectory-scheduling-EWT3IFz0G7
References
References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.
- Publisher
- Hindawi Publishing Corporation
- Copyright
- Copyright © 2022 Shifang Liao et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
- ISSN
- 1687-9600
- eISSN
- 1687-9619
- DOI
- 10.1155/2022/2365006
- Publisher site
- See Article on Publisher Site
Abstract
Hindawi Journal of Robotics Volume 2022, Article ID 2365006, 13 pages https://doi.org/10.1155/2022/2365006 Research Article Unmanned Aerial Vehicle Surveying and Mapping Trajectory Scheduling and Autonomous Control for Landslide Monitoring 1 2 3 4 Shifang Liao, Manzhu Ye, Rongcai Yuan, and Wanzhi Ma School of Resource & Environment and Historical Culture, Xianyang Normal University, Xianyang, Shaanxi 712000, China School of Surveying & Testing, Shaanxi Railway Institute, Weinan, Shaanxi 714000, China Xi’an Zhongke Xingyun Space Information Research Institute Co., Ltd.,, Xi’an, Shannxi 710000, China School of Educational Science, Ningxia Normal University, Guyuan, Ningxia 756000, China Correspondence should be addressed to Wanzhi Ma; mawanzhi79@nxnu.edu.cn Received 29 December 2021; Revised 8 February 2022; Accepted 9 February 2022; Published 24 March 2022 Academic Editor: Shan Zhong Copyright © 2022 Shifang Liao et al. .is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Real-time and efficient monitoring of geological disasters has received extensive attention in the application of UAV surveying and mapping control technology. .e application of traditional landslide monitoring methods lacks the accuracy of control algo- rithms, which has become a hot issue currently facing. Based on the landslide surface subsidence monitoring method, this article designs the UAV trajectory scheduling subsidence monitoring software, which can monitor the UAV’s flight status and navigation information, and draw the flight trajectory in real time. At the same time, the model solves the problem of storage and management of landslide inspection results by the landslide inspection management system, and realizes the functions of entering and querying landslide information, viewing inspection results, landslide safety judgment, generating reports, and autonomous control. .e simulation results show that the global accuracy reaches 0.975, and the algorithm recognition degree reaches 99.8%, which promotes the reliability of the landslide monitoring data for the identification of the surveying and mapping trajectory, and provides a decision-making basis for landslide treatment. consideration of many aspects, highway construction cannot 1. Introduction avoid the use of high filling and deep excavation methods, Due to the subsidence and deformation of the ground surface which will form a high landslide structure, because such high caused by landslide activities, the methods of subsidence landslides are all reconstructed from the original landform. monitoring have undergone a long-term process, ranging from Breaking the original balance relationship and balance con- traditional leveling traverse measurement and GPS monitoring ditions of the landform, the stress field will change accordingly, to the more commonly used three-dimensional laser scanning and the open-air slope will have cracks and landslide settlement monitoring and D-InSAR monitoring. .ese monitoring and sliding due to the local lithology and structure, surface methods range from full-field to few-field measurements, and water, ground runoff, etc. [2–5]. At present, with the continuous development and the observation efficiency is greatly improved, but their monitoring range, accuracy, and efficiency still cannot meet the progress of drone technology, autonomous drone navigation needs of comprehensive monitoring of regional subsidence [1]. technology has gradually been applied to all walks of life in .e emergence of UAV low-altitude photogrammetry at this society, providing it with high-performance technology ser- stage has greatly improved the efficiency of monitoring with its vices such as safety, convenience, intelligence, and environ- high mobility and real-time performance. Its high-precision mental protection, making people’s life and social production digital results are applied to various monitoring fields, in- have become more and more convenient and efficient [6–8]. cluding monitoring of mountain landslides, mudslides, and .e research and development of UAV autonomous navi- collapses. .e terrain conditions of highway construction in gation technology will continue to promote the innovation of mountainous areas are complex and changeable. Due to the social science and technology, which is of great significance to 2 Journal of Robotics the future development. With the vigorous development of maneuverable, flexible, and lightweight in the process of national highway traffic, highway mileage has increased year acquiring images. .e acquired images have high resolution by year. In the road operation stage, road maintenance and and play an important role in surveying and mapping management are particularly important, and landslide in- geographic information, earthquake mitigation and disaster spection is a basic but important task in road maintenance relief, agricultural production estimation, water conservancy and management. Daily regular inspections are conducive to and hydropower, and transportation and construction. .e real-time grasp of the stability and safety of landslides. Timely most widely used technology in UAV photogrammetry early warning and treatment of diseases are essential. At technology is tilt photogrammetry technology. .is tech- present, landslide inspections use manual inspections, gen- nology has become one of the current research hotspots. erally using visual inspection, tapping, and touching. .e With its high precision, high mobility, and high efficiency, it terrain conditions of landslides are complex and changeable. is suitable for operation in a variety of complex environ- When inspectors face high landslides and mountain land- ments. [12–15]. slides, manual inspections are very dangerous or even to reach At present, there are three methods for differential in- certain areas of the landslide, resulting in inspection results terferometry, namely, the two-track method of two InSAR that cannot fully reflect the true situation of the landslide. images, the three-track method of three InSAR images, and .erefore, conducting drone inspections for landslide disease the four-track method of four InSAR images. .e defor- has important practical significance for disaster reduction and mation result measured by InSAR technology is the esti- prevention and traffic safety. In traditional manual inspection, mated value of the spatial average change in a certain area. visual inspection is the main method. .erefore, the operation .is method makes up for the shortcomings of traditional mode of drones is used to replace or assist traditional in- monitoring methods for the sparse measurement points. It spection personnel’s short-distance visual inspection, which has the characteristics and advantages of large area, real makes landslide safety inspection possible [9–11]. time, and high precision, which effectively complements UAV low-altitude photogrammetry as a fast, convenient, traditional monitoring. Zanol et al. [16] used the X-band and and safe image acquisition technology has been studied in the L-band of D-InSAR to monitor the subsidence of the central field of geological disaster investigation and other fields. .e area of Utah. .e results show that the L-band can accurately purpose of this article is to study the technical methods of using monitor the subsidence value of this area, and it is less drones for landslide inspection to replace or assist manual road affected by slope changes. Although it is difficult to achieve a inspections, so as to ensure the safety of inspectors. According high-precision level for subsidence monitoring using to the relevant requirements of the landslide inspection, X-band, it can also monitor the survey area with a reasonable starting from the actual inspection work, two technical and accurate subsidence range. Figueiredo et al. [17] used methods of overall inspection and fine inspection of landslides high-resolution images to extract and monitor tunnel cracks based on unmanned low-altitude photogrammetry are pro- to provide a solid guarantee for the safe operation of the posed. In order to realize the early deformation recognition of subway. However, the uncertainty in the selection of the the landslide in the area, this article uses drone photogram- segmentation threshold directly affects the fracture extrac- metry technology to take multiple phases of the landslide to tion results. Julge et al. [18] applied it to the UAV image study its accuracy and deformation recognition methods. First, classification experiment and achieved 83% global accuracy the image of the survey area is obtained by flying the drone, and and 77.36% Kappa accuracy. .e classification results are then, the image is processed to obtain the point cloud data. highly consistent with the reference data type; in order to .rough continuous improvement of the number and location improve the extraction accuracy of ground fissures, the hit- of ground control points, the accuracy of UAV aerial pho- to-hit transform algorithm connects the broken ground tography technology has been improved, and the accuracy has fissures, and at the same time, the small pattern removal been increased from the decimeter level to the centimeter level. algorithm combining shape and area proposed in this article Before entering the test in the subsidence area, the UAV aerial is used to realize the removal of ground fissures and survey plan was designed. Combining two more classic sparse combined with ground fissure extraction results to conduct and dense deployment schemes, the accuracy of the UAV experiments to verify the effect of ground fissure fine digital results under the two schemes is verified to obtain a set treatment. of reasonable image control point schemes. .e two sets of Zhang et al. [19] studied a fault-tolerant path planning data, respectively cite the post difference technology to verify algorithm, using multisensor fusion to study fault-tolerant the digital results and verify the accuracy and superiority of the path planning and using wireless communication systems post difference technology. Under the same scheme, the DEM for positioning. Basir et al. [20] studied the path planning accuracy of tilt and vertical photography was compared, and algorithm based on a firefly under an uncertain environ- the best tilt aerial photography plan was obtained. ment, using the attraction law between fireflies to reduce the number of iterations and helping the robot to search for the optimal path in the constantly changing dynamic envi- 2. Related Work ronment according to the local static environment. At the Compared with traditional methods of monitoring the re- same time, a linear computational complexity graph search gional surface, the rapid development of drone aerial algorithm is proposed, which uses the priority queue, graph search, and the combination of a cattle to study a two-way photogrammetry technology in recent years has become more and more widely used. .e drone technology is highly sublist path planning algorithm that is faster than the Journal of Robotics 3 algorithm under global information. Domestic researchers erosion. In these locations, the concrete is easy to peel off and have also made a lot of contributions in the field of path the steel bars are exposed, causing structural damage and planning. Scholars have studied the PRM path planning loss of function: algorithm based on distance transformation and constructed lim u (cos x, sin x) � lim tan(xt − d)U. d (3) a narrow channel undirected graph based on image pro- x⟶∞ x⟶∞ cessing and random icon technology to study the density of .e protective structure is mainly subjected to rock and obstacles [21–23]. .e researcher proposed a multithreaded soil pressure. When the local structural strength is lower SA∗ path planning algorithm, adding multithreaded parallel than the structural stress, the structure will bend and fail. computing to the A-Xing algorithm and using heuristic .e magnetometer is susceptible to external interference search to improve the SA∗ algorithm to improve the speed when the drone is autonomously navigating. .erefore, the and accuracy of path calculation. Some scholars try to use magnetometer data are not used during complementary the genetic iterative algorithm to extract road cracks, but the fusion, but the value is the magnetometer output when the selection of the best parameters in the algorithm is still a system initializes the quaternion: problem worthy of in-depth study. In order to extract the cracks of concrete bricks, multiple threshold segmentation u 1 −1 −u d d ⎣ ⎦ ⎣ ⎦ ⎡ ⎤ ⎡ ⎤ u u × � × . (4) algorithms are used for experiments, although the final d q u u 1 1 q q result is optimized to the subpixel level, but the threshold segmentation algorithm cannot distinguish targets with In the safety judgment module, according to the drone similar gray values, resulting in a large amount of back- landslide inspection results, the four aspects of landslide ground noise; in order to realize the automatic detection of weathering degree, drainage system, protective structure, and cracks in concrete pipes, the researchers proposed a method other landslide phenomena are evaluated, and the corre- to segment pipe images based on morphological methods sponding options are selected in the inspection management [24–26]. system, such as weathering less intense, less intense, more intense, and intense. .e inspectors make choices based on 3. Analysis of Landslide Monitoring System the actual inspection results. .e system integrates these four options to determine the safety of the road landslide. .e 3.1. Structural Diseases of Landslides. In the landslide judgment results are as follows: stable, basically stable, under- structure, the broken rock mass of the slope top and the stable, and unstable. After the judgment, the corresponding slope surface is the broken rock mass formed by the fracture recommended treatment measures are displayed. and disintegration of the rock mass located on the top of the slope and the slope surface. At the initial stage of the for- mation of a landslide, the redistribution of stress resulted in 3.2. Classification of Landslide Safety Index. .e landslide the concentration of tensile stress on the top of the slope, safety index information needs to be properly preserved. In pressure cracking on the slope, the development of tension the information management module, the information has and cracks in the rock mass, and the fragmentation of the been stored in the database. Users can export landslide rock mass. Collapse is a phenomenon in which the rock and information and inspection results to files by generating soil on a high and steep landslide suddenly completely reports, which is convenient for users to save and can ef- separates from the parent body under the action of gravity fectively ensure the safety of landslide data. In the safety and external force, and then rolls, jumps, dumps or falls, and prewarning module, users can also send the results of accumulates to the foot of the slope or the road: landslide inspection and treatment suggestions to managers’ mailboxes to realize real-time communication and feedback lim u (t, x) − lim u (wxt + q)U � 0. (1) q q x⟶∞ x⟶∞ functions: When the quaternion is updated, the Euler angle can be ′ 1 − u sin θ cot θ ⎣ ⎦ used to obtain the attitude angle in the dispatch, but the yaw ⎡ ⎤ F(s, t, θ) � × . (5) cot θ sin θ 1 − u angle is not used as the final yaw angle of the attitude, and it needs to go through the first-order complementary calcu- Finding the extreme point (gradient position) in the lation. In the calculation, the complementary gyroscope first-order derivative function of the gray value is a trou- calculated output specific gravity is increased, while the blesome task, while finding the position where the second- magnetometer output specific gravity is decreased, and the order derivative function value is 0 is relatively easy. .e yaw angle is obtained by the fusion calculation: second-order edge detection algorithm is to detect the pixel lim u (x, t) � lim u (wxt − d)U. position in the image where the second-order derivative d d (2) x⟶∞ x⟶∞ function value of the gray value is 0, which is regarded as an .e protective structure is a force-bearing structure set edge. Similar to the first-order derivative function, the on the landslide slope surface to support the rock and soil second-order derivative function can be replaced by the pressure. .e protective structure is mainly affected by water second-order difference form of pixel gray value. erosion and rock and soil pressure. .e coagulation struc- Under the premise of ensuring accuracy, the use of drone ture has a large amount of cement. .e internal cracks in the camera measurement technology can be used to monitor the concrete are the primary breakthrough points for water deformation of the landslide group. For areas with a very 4 Journal of Robotics Table 1: Monitoring instructions for deformed areas. large number of landslides in Table 1, the deformation area has been monitored, and the identification of potential Partial deformed areas Monitoring test landslide areas has also been realized: Phase A Phase B Phase C s(s + w) 1 2.30 2.44 2.58 f(x, t) × U (s) − f(x − 1, t − 1) × U � 0. 2 2 2 1.77 1.87 1.97 w − s 3 1.24 1.30 1.36 (6) 4 0.71 0.73 0.75 .e above system errors have destroyed the central 5 0.18 0.16 0.14 6 −0.35 −0.41 −0.47 projection relationship between the objects, and the influ- 7 −0.88 −0.98 −1.08 ence of the curvature of the Earth belongs to the difference 8 −1.41 −1.55 −1.69 caused by different projection transformations. .e geoid is 9 −1.94 −2.12 −2.30 an ellipsoid. .e ground coordinate system used in map projection uses a plane as a horizontal plane. When the aerial photography range is large, this difference will affect the has proposed many methods for automatic extraction of accuracy of the aerial three encryption results, so it needs to connection points. Commonly used automatic connection be corrected. In the length of the buffer number of NS, the point extraction algorithms include scale-invariant feature historical value of the appropriate interval is taken as the transform (SIFT) and fast robust feature (SURF), Harris, estimated value of the current moment according to the MSER (the most stable extreme value region), and FAST delay time of the auxiliary sensor: (features from accelerated segment test) algorithms. In the actual field operation of tilt photogrammetry, the s(s + w) 2 2 s(s + w) × w − s − f(s, t) × � 0. (7) 2 2 image will inevitably be affected by uncertain external fac- w − s tors, such as radiation distortion and geometric distortion in If it is under high-speed operation, the delay correction the image. For external influence factors, these operators can needs to use the least square fitting; that is, the delay cor- accurately extract feature points according to their respective rection buffer (effective broadband autonomous control characteristics and appropriate measures to extract feature algorithm value) is polynomial fitting, and the curve is fitted points, and make the feature points have good uniqueness, according to the buffer area. .en, use the median value antirotation, antiscaling, and antilight variation. For the average recursive filtering to calculate the optimal estimated matched connection points, the quality of these connection value at the next moment, and refresh the buffer area. points has a greater impact on the later feature matching, In the same way, through the working principle of each which is directly related to its accuracy and accuracy: module, the data manual can calculate that the maximum ⎧ ⎪ ′ lim sin θ × cos θ � θ − 2x , sampling frequency of ultrasound in this study is 75 ms, the θ⟶∞ (9) maximum sampling frequency of GPS M8N is 100 ms, the ⎪ lim cos θ × cos θ � θ + 2x . maximum sampling frequency of the barometer is 20 ms, θ⟶∞ and the maximum sampling frequency of the optical .e monitoring personnel set the route and UAV pa- flowmeter is 8 ms. .e previous surveying and mapping rameters in the ground control system, and the aerial part of scheduling, serial port receiving buffer, digital filtering, etc. the UAV then performs the flight mission according to the will also have a period of time. instructions of the ground control system. UAV flight data are transmitted to the ground control system in real time 3.3. UAV Fine Inspection. .e development of UAV fine using wireless transmission channels, and ground moni- remote sensing and photogrammetry science has enabled toring personnel can change the flight plan or let the UAV humans to use low-altitude, high-altitude, and even outer continue to execute instructions based on the received data. space sensors to obtain various image data reflecting the When you take supplementary shots of poorly-photo- characteristics of the surface. By extracting the physical graphed areas, emergency situations, and landing, you can characteristics and information of various target objects, the switch the flight status from automatic to manual control: spatial shape, location, nature, change, and the relationship u (t) − u (t − 1) sin(wt) α α with the environment are studied. Feature points are pixels | � 1. (10) that describe the geographic location of the same infor- u (t) + u (t − 1) sin(w(t − 1)) α α mation in two or more images and are used for image On the basis of network socket communication, the TCP geometric transformation, image splicing, and three-di- or UDP protocol needs to be added. .is study takes net- mensional model generation: work communication based on the TCP protocol as an ′ ′ lim Δθ − θ − 2θ − wt − 2θ � 0. (8) example. .is protocol mainly uses the TcpClient and θ⟶∞ TcpListener classes built on socket. .e attributes of the .e extraction of feature points is an important step in the TcpClient class mainly include Available, ReceiveBufferSize, processing of oblique photogrammetry. .e relationship in Client, Connected, LingerState, NoDelay, SendTimeout, etc. Figure 1 between multiview images is constructed by con- .e TcpListener class is mainly used to monitor and receive jugate points. At present, the photogrammetry community incoming connection requests. It mainly includes five stages: Journal of Robotics 5 Feature points Automatic conection point extraction algorithm Feature points Development ∙ Altitude Image 2 Image 4 Data 1 ∙ Relationship ∙ Automatic ∙ Extraction ∙ Conjugate Photogrammetry ∙ Humans Image 3 Image 1 Data 2 Points ∙ Target Data 3 Extracting Model generation Model generation ∙ Location Data 4 Environment ∙ Objects Data 5 Constructed Sift P5 P6 Surf S5 S6 ∙ Nature Processing Data 6 P1 P4 P6 S1 S4 S7 P2 P3 Shift S2 S3 Surf Connection point Connection point Figure 1: UAV fine inspection topology. creating TCP instances, monitoring connections, receiving .e edge detection algorithm in computer vision is connection requests, sending and receiving data, and one of the earliest methods used for linear target ex- stopping services. traction, which is mainly divided into the first-order edge detection algorithm and the second-order edge detection algorithm. .e area where the local gray value changes 3.4. Image Processing Technology. .e low-altitude UAV obviously in the image can be regarded as the edge, and photogrammetry technology is mainly used for small sur- the change in the local gray value can be expressed by the veys or emergency survey missions. It is characterized by first-order derivative function. If the value of the first unmanned aircraft as a flying platform, various types of derivative function is larger, the probability of becoming sensors are used to obtain ground information, and com- an edge is higher. puter graphics are used. Image technology processes the From a qualitative point of view, the deformation shown acquired images and provides basic data for various types of by the monitoring data and the orthographic image is ba- photogrammetry applications. sically the same as the difference result; from a quantitative .e data processing part includes a data preprocessing point of view, there is a certain difference between the system and a data postprocessing system. .e data pre- amount of deformation displayed by the difference model processing system includes photogrammetric data down- and the actual amount of deformation: load, flight quality and data quality inspection, camera sin(wt) + sin[w(t + 1)] � 0 calibration, and distortion correction; data postprocessing . (12) includes air three processing, drawing production, inter- cos(wt) − cos[w(t + 1)] � 0 pretation, and result evaluation: .e amount of data in the same area has greatly in- u (t) − sin(wt) + sin(w(t + 1)) dt � 0. (11) creased, and the increase in the number of lenses has increased the amount of data, which brings a burden to .e UAV low-altitude photogrammetry system in subsequent image processing, and puts forward higher requirements on computer performance. .e clustered Table 2 consists of two parts: data acquisition and data processing. Data acquisition refers to the route planning multi-CPU and multi-GPU architecture is now used to function of ground personnel through the ground moni- handle tilt massive data obtained from photography. When the flight control board is rotated, the 3D modeling toring system or the remote control card control system of the UAV remote control. We realize the take-off and landing icon of the anonymous host computer will also rotate and the data will also change. It can be judged whether the of the UAV, and achieve the precise control of the drone dagger platform. Ground fissures are a form of existence of IMU module can communicate normally according to the consistency of the rotation direction and the beating fissures on the ground, and their spatial form is charac- terized by dark tones and strip-like linear features. amplitude of the data change: 6 Journal of Robotics Table 2: UAV low-altitude photometric system. Features step algorithm Connection point 1 Connection point 2 Connection point 3 Connection point 4 SIFT 1 2.37 0.47 0.57 2.67 SIFT 2 1.60 0.66 0.72 1.78 Algorithm accuracy SURF 1 0.83 0.85 0.87 0.89 SURF 2 0.06 0.04 0.02 0.00 MSER 1 −0.71 −0.77 −0.83 −0.89 MSER 2 −1.48 −1.58 −1.68 −1.78 z sin(wt) Signal echo 2 2 if (sin t> cos t), c.t.out fx, t|x + t � 0 � . 2468 10 zwzt (13) -32.85000000 -28.59628906 .e low-altitude indoor autonomous control module, -24.34257812 which is a combination of ultrasonic and optical flow meter, -20.08886719 is connected to the reserved UART port of the flight control -15.83515625 -11.58144531 system. You can use the serial debugging assistant to check -7.327734375 whether the data output is normal or the onboard OLED -3.074023437 screen on the flight control system to observe the data. 1.179687500 4.700000000 4. UAV Surveying and Mapping Trajectory Scheduling and Autonomous Control Model Construction for Landslide Monitoring 4.1. UAV Mapping Trajectory Scheduling. .e cameras Figure 2: UAV surveying and mapping trajectory scheduling carried by UAV photogrammetry are all nonmeasurement distribution. cameras, and the images acquired by them have a certain degree of distortion due to the time difference. .e image z sin(wt) z sin(w(t + 1)) must be removed before data processing. .rough the ad- + � 1, if cos(wt) − 1< 0 . { } justment of the position of the principal point of the image z(wt) z cos(wt) and the correction of the distortion parameters, the pro- (14) cessing of the distortion effect on the image is realized. However, the analog image taken by a traditional film In order to improve the matching speed of the image camera (i.e., an image with continuous image point and gray points with the same name and increase the density and distribution) or a digital camera equipped with a charge- accuracy of the matching points, it is necessary to create a coupled device (CCD) is limited by the storage of digital pyramid for the image and perform image matching based devices such as computers. .e captured digital image (i.e., on the image classification matching strategy that is grad- the image in which the image point and gray-scale distri- ually refined by the image pyramid, so that good reliability bution are expressed in the form of discrete data), to be and high accuracy can be obtained to extract similar feature processed in the computer, must be stored in the form of points from two adjacent images with a certain degree of discrete data. overlap in Figure 2, and then use gray-scale or feature-based matching algorithms to match points with the same name, and obtain a point set after relative orientation. 4.2. Composition of Landslide Inspection Management .e data transmission system consists of a modem and a System. .e landslide inspection management system is radio station. .e reference station modem encodes and divided into two parts: air and ground. .e air part contains modulates the relevant data, and then, the radio station electronic equipment for sending data, external antennas, transmits it. .e radio station on the rover station receives and ports for receiving data, while the ground part contains and demodulates the relevant signal through the radio electronic equipment for sending data and various antenna station. Finally, the three-dimensional coordinates of the interfaces. .ese are used for data transmission between air point are calculated and processed by the real-time dynamic cruise control equipment and ground observation stations. measurement software system. .is part of the system contains equipment that enables In order to make up for the lack of edge extraction accuracy the aircraft to take off and land. .e launch system provides caused by the randomness of the threshold selection in the first- guarantee for the accelerated take-off of the UAV, and the order edge detection algorithm, the second-order edge de- recovery part provides guarantee for the safe landing of the tection algorithm was proposed to search for the maximum UAV. Ground support equipment for air flight includes point of the local gradient value of the first-order derivative support equipment during the transportation phase and function (i.e., the value of the first-order derivative function): equipment during filming. UAV mapping trajectory scheduling value (%) Journal of Robotics 7 4.3. Autonomous Control of Safety Early Warning. Safety .e function of the transportation safety assurance equipment is to protect the shooting equipment and mission early warning needs to introduce the tilt aerial photography program into the subsidence area, obtain the two phases of equipment from damage, while the shooting safety assur- ance equipment refers to the equipment that allows the ground drone tilt data corresponding to underground drone to take pictures safely in the field. In actual research, mining activities, and use the level to measure the two phases the raw data of the accelerometer and gyroscope in Figure 3 of ground observation station data; carry out the two phases can be sent to MATLAB for FFT change, and the spectrum of drone tilt data. graph near the throttle can be analyzed, so that the frequency .e empirical parameters of the mine are used to predict range of noise can be observed, and the MATLAB filter the sinking basins, and the sinking basins are predicted to form. design toolbox can be used to study the filter parameter .e sinking basins obtained by tilt photography are quanti- design to determine various parameters of Butterworth low- tatively analyzed using the measured observation line data, and Table 4 is used to predict the sinking basins. Qualitative analysis pass filtering. While determining the order of the filter, consider that although the higher the order, the faster the was performed on the acquired subsidence basin. Compared with the subsidence difference between the stopband attenuation, the system phase delay will also increase: two phases of ground observations, the two subsidence ������� values are basically controlled within 150 mm, and the −b ± b − 4ac median error is 102 mm. Although there is a certain gap in ⟦ cos t⟧ ⇌ tan(ab)cos t. (15) the monitoring of the 10 mm terrain subsidence with slight 2a ′ θ−2θ | | changes, the mining parameters of this type, the aerial survey After comparing the theory of tilt and vertical photog- monitoring method has high reliability. raphy, and based on the high-precision aerial photography In order to facilitate reading and identification, it is plan of vertical photography, that is, postdifferential tech- necessary to renumber the images acquired in all directions nology and reasonable image control point layout, tilt and modify the image names according to certain rules. photography uses its multiview data acquisition and mul- Since the original image is being shot, there are effects such tiview joint adjustment to compare the postdifferential. as uneven illumination and changes in the sun’s altitude In the application of technology and image control point angle, and the oblique image has forward and back light program, the elevation accuracy of tilt photography data conditions. A uniform color and uniform light processing results was compared with that of vertical photography. It can be performed during preprocessing to ensure the natural was found that the elevation accuracy of tilt photography color transition of the image: was improved to a certain extent compared with that of ′ ′ ∯ wt − 2θ wt + 2θ dwdt vertical photography, which provided high-precision pro- (16) gram support for subsidence monitoring of tilt photography. ′ ′ − ∯wt − θ wt + θ dwdt � 0. Under the condition of the same degree of overlap, the basis can only be increased by increasing the heading CCD After superimposing the predicted subsidence basin with width or shortening the focal length. However, the shorter the aerial photography subsidence basin, we can intuitively the photographic baseline B, the more automatic the see that the areas with larger subsidence values in the aerial matching points, the easier the image matching, and the photography subsidence basin tend to be on the west side of higher the relative orientation accuracy. SPP extracts fixed- the aerial photography subsidence basin and are basically the dimensional features from the input object image. First, the same as the subsidence center of the predicted subsidence maximum pooling divides the image into 4 feature maps, basin. .e operations such as cropping and scaling, which each of which extracts features of the same dimension and are easy to lose information on the original image, are re- then pools the image into 16 feature maps with consistent moved and replaced by a spatial pyramid pooling layer on dimensions. the convolutional features. After receiving the survey task, it is necessary to first .e coincidence reflects the accuracy of aerial photog- analyze the survey task and objectives in detail to determine raphy of the sinking trend of the subsidence basin; in ad- the flight platform and photogrammetric sensor to be used, dition, it can be seen that the subsidence range obtained by and then choose to carry out drone aerial photogrammetry drone aerial photography is more comprehensive and broad work based on the geological and meteorological conditions to reflect the area affected by the subsidence of the mining of the survey area in Table 3. .e best season and time are area, breaking the traditional monitoring methods that can finally used as the basis for field reconnaissance and com- only be from one line. Observational bottlenecks reflect the prehensive analysis based on the previous geological work comprehensiveness of using aerial photography to monitor degree and other data. subsidence trends: Optimizing the flight plan can effectively reduce the ����������� � ��������������� � 2 2 number of invalid flights and time. Based on a compre- lim Δθ lim Δ2θ + cos a + cos b � 0. (17) hensive understanding of the location and area of the θ⟶∞ θ⟶∞ target area, full consideration should be given to flight efficiency, and the minimum number of routes to the .erefore, this study adopts the second-order Butter- target should be used under the condition of ensuring full worth low-pass filter. .e original data of the accelerometer coverage of the target area. and the data changed by FFTare simulated in MATLAB, and Network load 8 Journal of Robotics 0.5 0.8 -0.5 0.6 0.4 -1 0.2 0.8 0.6 -0.2 0.4 0.2 -0.4 -0.6 -0.2 -0.4 -0.8 -0.6 -0.8 -1 -1 Figure 3: Distribution of data sent by landslide inspection management. Table 3: Photogrammetric control algorithm of flight platform. Steps number Algorithm description Code explanation 1 Autonomous navigation path U (s) Int Index(char∗ S, char ∗ T) 2 Vertical and oblique images (n!/r!(n − r)!) B�(int ∗ )malloc(sizeof(A)); 3 .e two phases of field t − 1 Int i � 1,j � 1; 4 Its accuracy results x − 1 While(i<�s[0]&&j<�T[0]) 5 During the application process lim (n!/r!n!) //Merge(A,0,3,7); x⟶∞ 6 Based on the Smart3D zszt If(S[i] � � T[j]){ 7 Contextcapture platform f(x, t) ++i; ++j; }else 8 .e point cloud of the survey area zf(x, t) If(j � � 0||S[i] � � T[j]){ 9 UAV can directly carry out stable {mergesort(A,0,7); 10 Drone tilt photogrammetry U (s, t, x) I � i − j + 2; j � 1; 11 Mages obtained by technology i, j, k ∈ R Printf(“%d ”,A[i]); } 12 For the overall modeling of the target F(a, b) If(j> T[0])return i-T[0]; 2 2 13 UAV can directly carry out stable w − s Else int next[], int pos √������� 14 .e host computer software −b + b − 4ac Void get_nextone(char T[], int next[]){ 15 Geometric center of the flight platform cos a While(i<�T[0]) 16 .e tilt angles of the pictures obtained cos b Int kmpone(char S[], char T[]) 17 Object from multiple perspectives R(sin x, cos s) While(i < S[0]&&j < T[0]) Table 4: Analysis of autonomous control of safety early warning. Parameter name Channel 1 subsidence value Channel 2 subsidence value Parameter requirements Control method 19.31 47.37 External connection Autonomous communication 19.62 47.74 Wireless communication Warning bandwidth 19.65 47.77 500∼1500 MHz Control subsidence supply 19.25 47.30 Reserved for external interface Control subsidence life 18.27 46.11 Continuous work conditions Warning range 16.56 44.03 −40∼−20 dbm Analysis of channels 14.14 41.11 10 then, the accelerometer data after the second-order But- values, image point coordinates (x, y) as observations, and terworth filter with a cut-off frequency of 30HZ are simu- image internal and external azimuth elements, various lated, and FFT simulation is performed. .e processing distortion parameters, and spatial coordinates of pending results are as follows. We regard GCP coordinates as true points as unknowns. Node point Landslide inspection management sends data Journal of Robotics 9 Aerial triangulation calculation is a critical step in UAV 5. Application and Analysis of UAV Surveying image processing. First, we use photogrammetry software to and Mapping Trajectory Scheduling and correct the distortion of the image acquired by the UAV, Autonomous Control Model for then perform air triple encryption, and use the results of air Landslide Monitoring triple encryption to perform high-precision matching editing and acquisition. .e DSM filter edit the DSM to get 5.1. Preprocessing of Landslide Monitoring Data. For the the DEM that filters the buildings, etc., and then, use the dangerous situation of landslide monitoring data, in addi- DEM to digitally differentiate and correct the image to tion to human operation errors, other situations can be obtain multiple digital orthoimages in units of frames, avoided as much as possible. .e DJI Phantom 4 Pro drone namely, the DOM, and finally stitch them into the final used in this article is equipped with front and rear obstacle image data. avoidance functions, which can avoid hitting obstacles. In the case of weak GPS information, the drone will auto- matically hover and wait for a signal or return home au- 5.2. Realization of UAV Surveying and Mapping Trajectory tomatically, which can avoid flight accidents. Simulation. In order to realize the autonomous navigation On the basis of theoretical research, this article has done and monitoring of UAVs, it is necessary to realize the systematic software design and development of Beidou-based technology of remote monitoring of UAVs by the host high-precision autonomous navigation and monitoring tech- computer and remote binding tasks for UAVs online, so it nology for UAVs. On the basis of self-made monitoring upper needs to have the function of wireless communication. At computer, self-made remote navigation communication sys- the same time, a WIFI chip is added. If there is a WIFI signal tem, and anonymous flight control system in Figure 4 UAV in the surrounding environment, the host computer soft- comprehensive test platform was built, and the software design ware of the monitoring system and the UAV can directly of UAV high-precision positioning, UAV autonomous navi- carry out stable and efficient two-way communication in the gation inertial navigation algorithm, and UAV autonomous local area network: navigation path planning algorithm was completed. .e bar is the b-box regression. As of now, the two network modes 1 and n! n! lim − lim 2 do not share parameters, but are trained separately. x⟶∞ x⟶∞ r!(n − r)! r!n! (19) After the drone completes the aerial photography of the overall inspection, due to the drone’s own image trans- � 0, fori, j, k ∈ R(sin x, cos s). mission function, the aerial image can be checked at the Oblique photogrammetry carries cameras evenly distrib- ground control station, and the landslide can be judged uted at different angles around the geometric center of the according to the method of visual interpretation. .e image flight platform. .ere are two-lens oblique photography and can be judged that the target landslide is within the image five-lens oblique photography. .e tilt angles of the pictures range. .en, check the quality of the image to see whether it obtained by oblique photography are all greater than 3 , and the is blurred or deformed. If such problems occur, it is generally pictures obtained are called oblique pictures. Since oblique caused by camera settings, lens contamination, high wind photography captures the target object from multiple per- speed, etc. After the problem is resolved, we retake aerial spectives, it is particularly advantageous for the interpretation photography or only re-shoot a part of the image: of ground objects and targets, and has a higher quality as- U (s, t, x) zf(x, t, s) zf(x, t) surance for the overall modeling of the target area: Cz + + � 1. (18) zszt zszt zs for{a,b,c ∈ C} √√√√√√√√√√√√√√√√ √√√√√√√√√√√√√√√√√ √ ������� ������� 2 2 a,c,t −b + b − 4ac −b − b − 4ac (20) F(a, b)∗ G(cos a, cos b) � ∗ · · · ∗ . 2a 2a s.t. Aerial triangulation is the air three encryption. .is is the simplifies the data set. .e computational complexity and most important link in the aerial photogrammetric image model complexity of 3, 4 improve the computational speed processing industry, and it is also a difficult point in the and the accuracy of the final result. entire processing process. Its accuracy results directly affect .e dense point cloud is generated using the result of air three encryption, and the digital surface model data of the the later digital elevation model DEM, digital ground model DSM, and digital positive. On the basis of datasets 1 and 2, it entire survey area are obtained. .e digital surface model is has been improved to form a new algorithm body model, filtered to remove some surface objects such as vegetation which introduces a network structure instead of the selective and then manually edited to obtain the digital elevation data search algorithm, and integrates the candidate frame gen- that meets the requirements. Finally, it can be used to make eration into the deep neural network, which greatly 3D terrain of landslide. 10 Journal of Robotics 0 10 20 30 40 50 60 70 80 90 100 Data number Figure 4: UAV autonomous navigation path planning. .e database is used to store the basic information of the powerful expansion functions, rich interfaces, including landslide in Figure 5, such as the name of the landslide, UART, SPI, and 12C, and flexible application capabilities, lithology, topography, and rainfall, and it can also store the including TCP sampan DP/FTP. During hardware design, UAV landslide images and the results of the image post- the module in Figure 6 is embedded in the navigation processing. .e rationality of database design is related to communication board and connected to the UART pin of system calculation speed and accuracy. .e research content STM32. of this article involves the import and export of a large .e point cloud generated by the tilted image of the amount of data. .ere are many types of stored data. drone contains a lot of noise and nonground points. .is After the air triple encryption is completed, the DTM is article processes the point cloud based on the 3D point cloud extracted. .e software calculates the radiation adjustment processing software and forms a high-precision point cloud on a single image during processing to compensate for the after processing vegetation and high-voltage towers. visual effect and then adjusts the matching degree of ad- .rough this module, data can be sent to the WIFI network jacent images to balance the color of the image in the so as to realize the control and management of the Internet measurement area. .e orthophoto map generated by the of .ings. software can be exported to TIFF format and can be pro- .e output candidate region of the first step of dataset cessed in ArcGIS software. mode 1 is used as the input of the detection network. A simple configuration before use can realize the networking function of the UART device. .e SIM7100 is connected to 5.3. Case Application and Analysis. During the application the physical pin of the serial port 3 of the STM32, and the process of the example, the point cloud of the survey area in USR-C322 is connected to the physical pin of the serial port this article is automatically processed based on the Smart3D 2 of the STM32: ContextCapture platform, and the two phases of field ver- R(sin x, cos s) −⟨(zf(x, t, s)/zszt) x> t tical and oblique images obtained by drone tilt photo- � 0, for grammetry technology are combined with camera f(s, t) × exp s f(s, t) × sin x y< t parameters and field image control points. (22) Using aerial triangulation technology to calculate the high-precision external orientation elements of each image, .e landslide inspection management system needs to generate dense image points with the same name through record and query the basic information of the landslide. In dense matching, and then, perform matching after adjust- order to facilitate data entry, management, analysis, and ment, until the matching points meet the accuracy re- query functions, corresponding data tables must be estab- quirements, and then, high-precision point clouds can be lished to realize the above functions. A table is the most basic generated: data storage unit of a database. It is a two-dimensional structure composed of rows and columns. ⎧ ⎪ u (t) � cos(wt) + exp(wt), Columns are called fields and are used to define the u (t − 1) � sin(wt) + exp(wt), (21) structure of the table. Rows are called records and are used to store a piece of data in the table. We create corresponding fields u (t + 1) � tan(wt) + exp(wt). in the table according to the characteristics of the data and set the appropriate data type for each field. In the landslide in- .e 4G module uses the SIM7100 chip, which is an LTE spection management system, the types of data stored are platform based on Qualcomm’s MHD9215 multiplexer. It different, and the fields in each table are also different. has multiple frequency bands, including TDD-LTE//GSM/ According to Figure 7, the analysis of the results of the GNSS SMT. Its processor is provided by Cortex A5 (550HZ) three sets of data under different image control point layout and three QD SP6 (up to 500 MHz) built up. SIM7100 has UAV autonomous navigation path planning (%) Journal of Robotics 11 1.5 0.5 100 200 300 400 500 600 700 800 Group 3 Group 4 1.5 0.5 100 200 300 400 500 600 700 800 Training times Group 1 Group 2 Figure 5: Dispatching distribution of 3D terrain trajectory of landslide. 500 550 600 650 700 750 800 Mode 1 -10 -20 500 550 600 650 700 750 800 Point interval Mode 2 Figure 6: Data distribution of landslide inspection parameters. schemes shows that the difference between the basic ori- source I can also be improved by more than 2 times entation point and the checkpoint of the data source I after compared to data source II. using the differential technology is generally smaller than .e texture mapping based on the reconstructed TIN that of the data without using the differential technology. In network mainly includes two parts: texture optimization and terms of in-plane error, data source I is at least 2 times higher texture extraction. Due to the large overlap of the oblique than data source II, and in elevation error accuracy, data images, most of the triangles reconstructed on the surface Trajectory scheduling value Chi-square test of local Landslide inspection Landslide inspection data 12 Journal of Robotics -35.00 -30.00 -25.00 -20.00 -15.00 -10.00 -5.000 5.000 2 4 6 810 Test point Figure 7: Layout of drone navigation image control points. correspond to the corresponding textures of multiple im- internal and external orientation elements of the image ages. .e texture selection is based on the generated triangle using photogrammetric computer vision methods: mesh combined with the distortion parameters and the � � � � � � � 2 � � � � � − Kwt/2 � � � � � � cos1 − K wt/2U 1 cos t −t � � � � � Ke tan t � � � � � � � � � � � � � ∗ � . (23) � � � � � � � � � � � − Kwt/2 � 2 � � � � � � t sin t −tan t −Ke � 1 cos1 − K wt/2U � and solving the ambiguity of the whole week, and derives the .e monitoring data of the wooden piles showed that during the comparison time period, it shows a certain theoretical formula for the calculation of virtual observa- amount of deformation. .e cumulative deformation of tions and the establishment of the double-difference model. PM1 was 50 cm, and the cumulative deformation of PM2 First of all, related software development is carried out on was 4.4 cm. It can be seen that the trailing edge of the the network landslide monitoring technology, and the landslide is relatively broken and divided into pieces by network landslide monitoring is applied to the positioning of cracks. Specifically, we output a candidate frame, intercept the UAV. Secondly, it studied in detail the UAV surveying the original image through the candidate frame, pass the and mapping scheduling method based on complementary intercepted image through conVtool several times, and then filtering, including Butterworth low-pass filtering technol- output two branches through ROI-pooling, one for target ogy and accelerometer gyroscope complementary filtering classification and the other for target classification. technology. On the basis of surveying and mapping scheduling, the vertical third-order complementary auton- .is is because the data source I adopts the post- difference technology, and the acquired image POS posi- omous control algorithm with observer delay correction and tioning information is processed by the difference software the horizontal dual-observation autonomous control algo- and the accurate latitude, longitude, and ellipsoid height at rithm are studied, and simulations have verified the prac- the time of camera exposure are calculated with higher ticability. After that, according to the geometric attribute accuracy, so the accuracy of the data source I will be higher information of the ground fissures in the UAV image, the than that of data source II; in addition, with the increase in MF template and FDOG template that meet the vertical the number of control points and the changes in positions, profile curve are constructed, and the UAV image is filtered the error changes in the empty three planes have not and calculated separately to achieve the purpose of en- hancing the ground fissure signal, and linear stretching is changed significantly. .is is because the nine points of the image control point scheme 1 are already available in this adopted. .e method eliminates the difference in the value range obtained by the filtering operation and performs the survey area. .e plane accuracy is controlled to a higher level, but the error in the elevation is significantly improved, difference operation between the two results to highlight the and the minimum is increased to more than 2 times. ground fissure signal in the image and weaken the edge signal of the ground feature. Finally, on the basis of au- tonomous control, the path planning algorithm is studied 6. Conclusion and the feasibility of the algorithm is verified by simulation, .is article studies the principle of landslide monitoring so that the UAV can avoid obstacles and complete the technology, and its double-difference positioning model landslide monitoring task according to the optimal path introduces the core technology of UAV repair and detection under certain conditions. UAV navigation image control point layout value Journal of Robotics 13 [13] A. W. A. Hammad, B. B. F. da Costa, C. A. P. Soares, and Data Availability A. N. Haddad, “.e use of unmanned aerial vehicles for dynamic site layout planning in large-scale construction .e data used to support the findings of this study are projects,” Buildings, vol. 11, no. 12, p. 602, 2021. available from the corresponding author upon request. [14] M. M. Eltabey, A. A. Mawgoud, and A. Abu-Talleb, “.e autonomy evolution in unmanned aerial vehicle: theory, Conflicts of Interest challenges and techniques,” Advances in Intelligent Systems and Computing, pp. 527–536, 2020. .e authors declare that there are no conflicts of interest in [15] A. Khan, S. Gupta, and S. K. Gupta, “Multi-hazard disaster this article. studies: monitoring, detection, recovery, and management, based on emerging technologies and optimal techniques,” International Journal of Disaster Risk Reduction, vol. 47, Acknowledgments Article ID 101642, 2020. [16] R. Zanol, F. Chiariotti, and A. Zanella, “Drone mapping .is study was supported by “.e Special Fund Construction through multi-agent reinforcement learning,” IEEE Wireless Project of Superior Subject in Shaanxi Province colleges and Communications and Networking Conference (WCNC), vol. 4, universities (Grant no. 0602).” pp. 6-7, 2019. [17] H. Valente de Figueiredo, O. Saotome, E. H. Shiguemori, P. Silva Filho, and V. V. Estrela, “Image database of low- References altitude UAV flights with flight condition-logged for pho- togrammetry, remote sensing, and computer vision,” Imaging [1] S. Aggarwal and N. Kumar, “Path planning techniques for and Sensing for Unmanned Aircraft Systems: Volume 2: De- unmanned aerial vehicles: a review, solutions, and chal- ployment and Applications, vol. 2, pp. 91–111, 2020. lenges,” Computer Communications, vol. 149, pp. 270–299, [18] K. Julge, A. Ellmann, and R. Ko¨ok, ¨ “Unmanned aerial vehicle surveying for monitoring road construction earthworks,” Ce [2] L. Patan´e, “Bio-inspired robotic solutions for landslide Baltic Journal of Road and Bridge Engineering, vol. 14, no. 1, monitoring,” Energies, vol. 12, no. 7, p. 1256, 2019. pp. 14–17, 2019. [3] M. Erdelj, M. Krol, ´ and E. Natalizio, “Wireless sensor net- [19] H. Zhang, E. Aldana-Jague, F. Clapuyt, F. Wilken, works and multi-UAV systems for natural disaster manage- V. Vanacker, and K. Van Oost, “Evaluating the potential of ment,” Computer Networks, vol. 124, pp. 72–86, 2017. post-processing kinematic (PPK) georeferencing for UAV- [4] D. Giordan, M. S. Adams, I. Aicardi et al., “.e use of un- based structure- from-motion (SfM) photogrammetry and manned aerial vehicles (UAVs) for engineering geology ap- surface change detection,” Earth Surface Dynamics, vol. 7, plications,” Bulletin of Engineering Geology and the no. 3, pp. 807–827, 2019. Environment, vol. 79, no. 7, pp. 3437–3481, 2020. [20] R. Basir, S. Qaisar, M. Ali, N. Ahmad Chughtai, M. Ali Imran, [5] P. Arena, F. D. Pietro, A. L. Noce, S. Taffara, and L. Patane, and A. Hashmi, “Performance analysis of UAV-enabled di- “Assessment of navigation capabilities of mini cheetah robot saster recovery networks,” Autonomous Airborne Wireless for monitoring of landslide terrains,” Research and Technology Networks, pp. 157–194, 2021. for Society and Industry (RTSI), vol. 9, pp. 540–545, 2021. [21] Y. Sahraoui, C. A. Kerrache, M. Amadeo et al., “A cooperative [6] D. Mishra and E. Natalizio, “A survey on cellular-connected crowdsensing system based on flying and ground vehicles to UAVs: design challenges, enabling 5G/B5G innovations, and control respiratory viral disease outbreaks,” Ad Hoc Networks, experimental advancements,” Computer Networks, vol. 182, vol. 124, Article ID 102699, 2022. Article ID 107451, 2020. [22] D. Hermle, M. Keuschnig, I. Hartmeyer, R. Delleske, and [7] C. Kyrkou, S. Timotheou, P. Kolios, T. .eocharides, and M. Krautblatter, “Timely prediction potential of landslide C. Panayiotou, “Drones: augmenting our quality of life,” IEEE early warning systems with multispectral remote sensing: a Potentials, vol. 38, no. 1, pp. 30–36, 2019. conceptual approach tested in the Sattelkar, Austria,” Natural [8] H. Hildmann and E. Kovacs, “Review: using unmanned aerial Hazards and Earth System Sciences, vol. 21, no. 9, pp. 2753– vehicles (UAVs) as mobile sensing platforms (MSPs) for 2772, 2021. disaster response, civil security and public safety,” Drones, ´ ´ ´ [23] A. To¨rok, ¨ A. Barsi, G. Bog ¨ oly, ¨ T. Lovas, A. Somogyi, and vol. 3, no. 3, p. 59, 2019. P. Gor ¨ og, ¨ “Slope stability and rockfall assessment of volcanic [9] H. Wang, H. Zhao, J. Zhang, D. Ma, J. Li, and J. Wei, “Survey tuffs using RPAS with 2-D FEM slope modelling,” Natural on unmanned aerial vehicle networks: a cyber physical system Hazards and Earth System Sciences, vol. 18, no. 2, pp. 583–597, perspective,” IEEE Communications Surveys & Tutorials, vol. 22, no. 2, pp. 1027–1070, 2019. [24] I. Jeelani and M. Gheisari, “Safety challenges of UAV inte- [10] J. Besada, L. Bergesio, I. Campaña et al., “Drone mission gration in construction: conceptual analysis and future re- definition and implementation for automated infrastructure search roadmap,” Safety Science, vol. 144, Article ID 105473, inspection using airborne sensors,” Sensors, vol. 18, no. 4, p. 1170, 2018. [25] S. Ekici, “Design of unmanned helicopter equipped with [11] M. Labbadi and M. Cherkaoui, “Robust adaptive backstepping turboshaft engine for agriculture spraying mission based on fast terminal sliding mode controller for uncertain quadrotor thermodynamic analysis,” Journal of the Institute of Science UAV,” Aerospace Science and Technology, vol. 93, Article ID and Technology, vol. 10, no. 1, pp. 532–546, 2020. 105306, 2019. [26] L. Tan, J. Guo, S. Mohanarajah, and K. Zhou, “Can we detect [12] K. O. Said, M. Onifade, J. M. Githiria et al., “On the appli- trends in natural disaster management with artificial intelli- cation of drones: a progress report in mining operations,” gence? a review of modeling practices,” Natural Hazards, International Journal of Mining, Reclamation and Environ- vol. 107, no. 3, pp. 2389–2417, 2021. ment, vol. 35, no. 4, pp. 235–267, 2021.
Journal
Journal of Robotics – Hindawi Publishing Corporation
Published: Mar 24, 2022