Discussion on Key Issues of UAV Large-scale Mapping

Yonghua ZHAO

Abstract The main constituent parts of unmanned aerial vehicle aerial photogrammetry systems are discussed. The key issues including the division of regional networks， the layout of regional networks， the correction of lens distortion， the optimization of external orientation elements， the aerial triangulation， the image matching and fusion， and the production of digital elevation models and digital orthoimages， tilt real 3D models， and digital line drawings， were analyzed. The advantages of UAV aerial photogrammetry were compared. This study provides reference for measuring large-scale topographic maps by UAV photogrammetry systems.

Key words UAV; Large scale; GPS; Data transmission system

Received： April 7， 2021  Accepted： June 3， 2021

Yonghua ZHAO （1969- ）， male， P. R. China， senior engineer， devoted to research about engineering survey and cadastral survey.

*Corresponding author.

With the rapid development of Chinas economy， the construction of digital cities has gradually been put on the agenda， so the demand for spatial geographic information is increasing， especially for the rapid acquisition and real-time update of large-scale topographic maps. The main methods of obtaining topographic maps include theodolite surveying， total station surveying， RTK surveying and plane table surveying. With the innovation and development in the field of surveying and mapping， new technologies such as satellite remote sensing， three-dimensional laser scanning and unmanned aerial vehicles have gradually been applied to the production of topographic maps.

With the rapid development of UAV （unmanned aerial vehicle） hardware equipment and its post-processing software， especially the successful development of high-stability measurement UAVs and the increasingly perfect performance of the equipment they carry， there are more and more applications of UAV aerial survey and remote sensing technology. The area of UAV surveying and mapping is small， the amount of data is relatively small， and the post-processing is relatively simple. Compared with manned aircrafts， UAVs are also less affected by weather conditions and airspace approvals， and there is no approval link for airport use. The flight speed and flight range can be set according to the needs， with extremely high mobility. UAVs can take off with the assistance of catapults， and can roll down or parachute down as needed， without a dedicated take-off and landing site. They can operate under extremely harsh environmental conditions. Meanwhile， UAVs are inexpensive and relatively low-cost， and they have significant advantages when applied to large-scale， small-scale， and tight-construction-period surveying and mapping[1-2].

In this study， we discussed the key issues of measuring large-scale topographic maps using unmanned aerial vehicles， aiming to provide reference for the field and office links of measuring large-scale topographic maps by UAV aerial survey systems.

UAV Aerial Photogrammetry System Composition

UAV aerial photogrammetry system mainly includes six parts： flight platform， flight control system， ground station system， data transmission system， aerial photography system， and launch and recovery system[3-4].

Fight platform

At present， the most commonly used UAV flight platforms in China mainly include fixed-wing UAVs， rotor-wing UAVs and composite-wing UAVs. Comparing the three， the advantages of fixed-wing UAVs are good high-flying performance， fast flying speed， and high operating efficiency， but they have high requirements for take-off and landing sites， and small-scale operations are more difficult. Rotor-wing UAVs have strong flexibility， almost no site restrictions for take-off and landing， but their flying speed is slow and unstable， and their anti-turbulence ability is poor. Composite-wing UAVs combine the advantages of fixed-wing and rotor-wing UAVs， but the technology is not mature yet， and the equipment is more expensive and the operation cost is high.

Flight control system

The flight control system uses GPS equipment for positioning， uses IMU system to obtain platform attitude parameters such as gyroscopes and accelerations， uses GPS and IMU combined algorithms to calculate the UAVs flight altitude， flight rate， roll and pitch and other data， and receives and processes the information sent by the ground control system in time. Onboard computers use the central processing unit as the core to calculate the offset law according to the selected route， control the aircraft to fly according to the pre-designed route and expose the flight at fixed points.

Ground station system

The ground station system includes mission planning and monitoring system， communication system and data processing system. It uses the route design software of the ground monitoring system to design the flight route before aerial photography. During the flight， it guides the UVA to fly autonomously along the designed route. Meanwhile， the ground monitoring system can display aircrafts flight speed， flight altitude， flight attitude， weather and other parameters in real time during the flight， and record and store the flight parameters and data. The ground station personnel can monitor the UAVs flight status in real time based on the flight parameter data displayed in real time.

Data transmission system

The data transmission system mainly includes ground data transmission and air data transmission. Both ground data transmission and air data transmission include transmission antennas， data interfaces， digital transmission stations， etc.， mainly for mutual transmission of command data from the flight control system and ground monitoring center.

Aerial photography system

The aerial photography system includes high-resolution cameras， multi-spectral cameras， and hyper-spectral cameras.

Launch and recovery system

The launch part is to increase the speed of the UAV to the take-off speed within a certain period of time; the recovery part is to ensure the landing safety of the UAV; and there are two ways to launch and recover： the sliding method and the catapult method. The sliding method is generally carried out on the runway， and the requirements are relatively high. The catapult method is an emerging linear propulsion technology， which is mostly used for short-stroke launching large loads.

Key Issues of Large-scale UAV Mapping

Layout and measurement of field photograph control points

The reliability and accuracy of the layout of the photo control points directly affects the accuracy of the air triple encryption， and affects whether the final positioning can be achieved. The photograph control point layout plan needs to pay attention to the division of the regional networks and the distribution of points in regional networks.

① Regional network division： The division of the regional networks needs to consider factors such as the specific topographic features of the survey area， the map scale， the opposite resolution， and the division of planned UAV aerial photography areas. In general， the shape of the divided regional networks should be mainly rectangular and square as far as possible. It should be noted that adjacent regional networks of route edges require at least 4 or more baseline repetitions between the control points on the route， and the adjacent regional networks of side edges require at least two or more baseline repetitions[5-6].

② Requirements for the distribution of regional network points： The photograph control points between the divided regional networks should be selected as far as possible in the photograph overlapping area， so as to facilitate the sharing of photograph control points between adjacent regional networks. Horizontal and vertical control points are set at the periphery， center and corners of regional networks， and elevation points are set at both ends of routes and center line areas[1].

External data preprocessing

Lens distortion correction

Because UAVs generally have a small load， they can only carry non-measurement cameras， which are affected by many factors during the processing and installation process， which will cause different degrees of error in the camera lens， further leading to lens distortion， which makes the final aerial images obtained with poor distortion. When the distortion is large， that is， the deviation between theoretical values and actual values in aerial survey images caused by lens distortion is large， then the collinear relationship between the projection center and the corresponding principal point of photograph is destroyed at this time， and the offset of the image point coordinates will eventually lead to a large error in the air triple encryption results and low image matching accuracy[1，4].

Lens distortion mainly includes three types： radial distortion， image principal point offset distortion and tangential distortion. The radial distortion is mainly caused by the radial deviation of principal points of images caused by the radial curvature error of the lens， and has the characteristic that a farther distance from image centers causes greater the distortion. Both the principal point offset distortion and the tangential distortion are caused by the assembly of lenses， and there are both radial and tangential deviations. Since the cameras carried by UAVs are fixed-focus lenses with a fixed focal length， the differences between theoretical values and actual values in aerial survey images caused by lens distortion are systematic error. The correction of lens distortion generally adopts the radial compound distortion and thin prism distortion model （brown model for short） to correct image coordinates.

Correction method of lens distortion： A correction field with high-precision specifications is first established， in which a certain number of high-precision mark points with known spatial coordinates are set， and digital cameras to be corrected are used to take pictures of the mark points arranged in the correction field. And the image point coordinates of the mark points on the photos are extracted. The image point coordinates are brought into collinear equations， and the object coordinates of the marker points are used to inversely calculate the ideal coordinates of the corresponding marker points on the images through perspective transformation. And the ideal coordinates are brought into relevant formulas to obtain distortion correction parameters. Air triple encryption can be performed on the basis of the correction of the distortion difference[1].

Optimization of external orientation elements

The external orientation elements of the image data obtained by UAV aerial survey are generally included in the POS data or satellite navigation and positioning auxiliary data. Before the data are processed， they needs to be optimized to reduce the trajectory system deviation of UAVs in flight and optimize the eccentric vector between the satellite navigation positioning and the camera projection center and the collimation axis eccentric angle between the inertial measurement unit and the camera coordinate axis. For this type of optimization processing， Chebyshevs quartic polynomial is used to optimize the external orientation elements.

Aerial triangulation

Aerial triangulation is also called space three encryption. It is a method of encrypting more control points indoors based on the spatial geometric relationship between aerial photographs and ground targets， combining with a small number of photo control points in the field. In aerial photography， the aerial photos of the same route and adjacent routes are designed to overlap to a certain extent， and there are no aerial photography loopholes. The photograph control points are selected according to the layout plan required by the specification to conduct field surveys. And a regional network model corresponding to the field is established according to the method of photogrammetry， and the plane coordinates and elevation coordinates of the encrypted points are calculated.

Aerial triangulation methods include air strip method， independent model method， and beam method. The aerial triangulation mode is divided into two types according to its degree of automation： fully automatic mode and semi-automatic mode. In the fully automatic mode， the all standard points are automatically matched through computers by the image automatic matching technology. This method has strict requirements on image quality. In addition， there are also requirements for surface objects， and terrain types in the measurement area. The semi-automatic mode requires manual measurement and selection of standard points on screens， then obtains the points with the same name through image matching， and then performs adjustment on these points with the same name and the standard points together. This mode is characterized by high efficiency， and there will be many connection points between routes and models， so the adjustment network is very stable， and the accuracy is also high， which is suitable for complex measurement areas[1，7].

Image matching and fusion

Since most UAVs fly at low altitudes with a relatively small field of view and a small coverage of a single photo， it is necessary to use image stitching technology to stitch out the overall orthophoto of the entire work area. The image stitching technology is to stitch two or two sets of images with common overlapping parts into a larger high-resolution image， which specifically includes the two tasks of image registration and image fusion.

Image matching is to find out the positional relationship between the overlapping areas of the images， and transform the images to a unified coordinate system through the corresponding mathematical model. There are many methods of image matching. According to the different image information used in the matching process， it can be roughly divided into four types： methods based on coordinate information， methods based on gray information， methods based on transform domain， and methods based on feature information matching. Among them， the method based on feature information matching is currently the most commonly used method for image matching in UAV aerial photogrammetry. Compared with other three methods， it has certain resistance to transformation in terms of distortion， noise and grayscale changes， and has the advantages of small amount of calculation and high efficiency.

The method of image matching based on feature information has four steps： ① extracting the feature information of images （the quality of feature information extraction directly affects the accuracy and speed of matching）， ② performing feature matching on the extracted image feature information to establish a corresponding relationship between adjacent images， ③ establishing a mathematical model to determine the overall transformation relationship of adjacent images according to the existing feature matching relationship， and ④ transforming the images into a unified coordinate system according to the mathematical model， and performing interpolation on the images[1].

After the image matching stage is completed， if the images are superimposed on each other directly and simply according to the geometric change model between the images， there will be obvious color difference and boundary lines at the positions of image stitching， and even partial misalignment， making the visual effect of the overall image after superimposition very poor， and the UAV images need to be fused to achieve seamless stitching.

The currently commonly used image fusion method is the best stitching line fusion algorithm.  The principle of the algorithm is based on the idea of graph cutting. An irregular stitching line is chosen in an image overlapping area to replace the original stitching line， so that the overlapping area is divided into two irregular areas. The UAV images can be bounded by the best stitching line when stitching， and randomly picking the content of an area on both sides can effectively solve the problem of low registration accuracy and realize seamless stitching[1].

Agricultural Biotechnology2021

Surveying and mapping products

Digital elevation model

Digital elevation model （DEM） is a data set of plane coordinates and elevations of regular grid points on a certain projection plane （such as Gaussian projection plane）. The main method for UAV aerial photogrammetry to obtain a DEM is to automatically match a large number of three-dimensional discrete points by high-precision digital image matching algorithm based on the aerial triangulation results， extract and automatically measure multiple points， and construct the DEM by interpolation after filtering out the unqualified points[1].

Digital orthograph model

When constructing a digital orthograph model （DOM）， digitized aerial photographs or remote sensing images are scanned and processed under the support of a digital elevation model to correct the projection difference corresponding to each pixel. Then， insertion is performed according to the images， and the generated image data are finally tailored according to the countrys basic scale map frame range.

Tilt real 3D model

The real 3D model is the main output product of the tilt photogrammetric 3D modeling technology. Real 3D models constructed based on the UAV tilt photogrammetry technology can truly display the multi-angle texture information of the terrain and features with higher accuracy. During UAV tilt photogrammetry， the software calculates a triangular irregular network based on the aerial triangulation results under rigorous algorithms to generate a white model of the 3D model， selects corresponding scenes from specific aerial photos for texture fitting according to the shape of the 3D model， and finally outputs the real 3D model with real texture[1，8].

Digital line drawing

Digital line drawing （DLG） is a vector dataset that expresses landforms and features in the form of points， lines， areas or map-specific graphic symbols. For traditional UAV surveying topographic maps， aerial triangulation is performed using high-resolution images obtained by UAVs， combining with a certain amount of ground control points， to obtain undistorted images， and stereomapping is performed combining with vector collection systems provided by some UAV image processing software. When the terrain and features in the stereomapping process are unclear on photos or when the eaves are corrected， supplementary survey should be carried out in the field using a total station and RTK， and the digital line drawing is finally output and evaluated for accuracy. The disadvantage of this method is that the office personnel need to wear stereo glasses and coordinate the use of the hand wheel caster during the stereo collection. The process is cumbersome and difficult to get started. In the method of collecting the line drawing on the basis of a tilt 3D model， the real 3D model produced is loaded into the Tsinghua Sunway EPS software based on the UAV tilt photography technology for naked-eye stereo collection and office editing and processing into a map. Compared with the traditional method， this method is quicker to get started， and has no need to wear stereo glasses and operate hand wheel castors. Since three-dimensional models can truly reflect the multi-angle attribute data of ground objects， they also reduce the eaves correction link in the field mapping work， and the collected line drawing is more accurate[9-10].

Conclusions

The UAV aerial photogrammetry system mainly includes six parts： flight platform， flight control system， ground station system， data transmission system， aerial photography system， launch and recovery system.

The key issues of large-scale UAV mapping include the division of regional networks， the layout of regional networks， the correction of lens distortion， the optimization of external orientation elements， the aerial triangulation， the image matching and fusion， and the production of digital elevation models and digital orthoimages， tilt real 3D models， and digital line drawings.

UAVs have the advantages of low price， efficient operation， low flying height， etc. They can quickly obtain clear and high-precision image data information， and can simultaneously greatly reduce the workload in field and reduce production costs， thus having very considerable economic efficiency. UAV photogrammetry technology is more and more widely used in large-scale surveying and mapping.

References

[1] YUAN XX. Large scale measurement technology and application research of UAV[D]. Huainan： Anhui University of Science and Technology， 2019. （in Chinese）

[2] MA TC. Research on the application of UAV photogrammetry system in large-scale mapping[D]. Chengdu： Chengdu University of Technology， 2018. （in Chinese）

[3] SUN YL. Research on large-scale mapping method for small unmanned aerial vehicle aerial survey system[D]. Chengdu： Chengdu Univerisity of Technology， 2017. （in Chinese）

[4] WANG JJ. Research on low-altitude aerial photographical systems based on unmanned aerial vehicle[D]. Harbin： Harbin Institute of Technology， 2016. （in Chinese）

[5] SUN L. Accuracy analysis of surveying and mapping large-scale topographic maps by UAV aerial photography system[D]. Kunming： Kunming University of Science and Technology， 2017. （in Chinese）

[6] SEN B， ZHOU XH， LIN XB， et al. Application of UAV aerial photography system in large-scale surveying and mapping of coastal zone[J]. Bulletin of Surveying and Mapping， 2017（9）： 156-158. （in Chinese）

[7] ZHOU XM， ZHAO LB， ZHANG XL， et al. Discussion on technique and method of low level UAV image processing[J]. Architectural Engineering Technology and Design， 2014， 35（2）： 182-184. （in Chinese）

[8] HU ZT， LIU YF. A Method for large-scale topographic map based on the combination of realistic 3D mapping oblique photography and stereo-pair vertical photography[J]. Urban Geotechnical Investigation & Surveying， 2018（2）： 82-85. （in Chinese）

[9] CUI J， LI J， WANG P， et al. Research on aerial photography of large scale topographic map and quality control of indoor work [J]. Journal of Shandong Jianzhu University， 2018， 33（1）： 7-13. （in Chinese）

[10] SONG WP. Research on integration of low altitude remote sensing system of UAV and related questions of image data processing[D]. Xian： Changan University， 2016. （in Chinese）