Http://blog.sina.com.cn/s/blog_77805ef60100pgip.html
UAV (UAV) Remote sensing is an important method of aerial remote sensing, and is increasingly becoming an important means of obtaining spatial data, it has the advantages of long endurance, real-time image transmission, detection in high-risk areas, low costs, and flexibility. It is a powerful supplement to satellite remote sensing and man-machine aerial remote sensing. In recent years, drones have achieved rapid development in military and civil fields, attracting the attention of many scientific research institutions. UAV remote sensing has become a trend of development.
However, due to the complexity of the drone Flight Environment and flight instabilityPosBecause the information is not accurate enough and the data volume is large, it is an urgent problem to deal with drone data.
ExploitationLPSThe software is used to process drone data and obtain the mosaic result of positive shot correction. It provides a new solution for UAV Data Processing and meets the needs of UAV in surveying and mapping and disaster emergency quick processing.
LPSLeica photogrammetry suite is a series of digital photography and Remote Sensing processing software released by Leica. LPS provides high-precision and efficient production tools for image processing and photography. It can process a variety of aerospace (the most commonly used satellite images include QuickBird/ALOS/GeoEye/SPOT5, etc.) and aviation (scanning aerial Films,Sensor Image targeting and empty three encryption, processing various digital image formats, black/Digital images such as white, color, multi-spectral, and hyperspectral. The Application of LPS also includes vector data collection, digital ground modeling, normal-shot image mosaic, and Remote Sensing processing. It is the first software series integrating remote sensing and photographic measurement on a single platform. LPS Product Structure 1 is shown.
Figure 1 LPSProduct Structure Diagram
The LPS modules used in UAV data processing include:: LPS coreAndAte(OrEate).
Data Description
The test data used in this article is composed (Canon EOS 5d Mark II) Shot473Drone data. In addition, the focal length (f0 = 24mm), pixel resolution (CCD size = 6.4um), aerial height, and POS data are also used.
LPSUAV Data Processing Process
LPSThe procedure for processing drone data is as follows:
Figure 2 procedure of using LPs to process drone data
(1)Create a project and import data
Create a project:LPSTo process data, you must first create a test area file, that isBlockFile. Select the corresponding ry correction model. The model we selected here is a digital camera (Digital Camera). Define the attributes of the measurement area, including the projection coordinate system and flight height.
Data Import: after the project file is created, import data: images, camera parameters (like the coordinate of the main point, focal length,CCDDimensions) andPosData.
(2)Automatically generate a vertex with the same name
According to the initialPosData to automatically generate vertices with the same name. You can modify the policy to generate a vertex of the same name as needed, so that the generated vertex of the same name is evenly distributed throughout the test area.
(3) Null 3 computing
Based on the generated large number of vertices with the same name and the initial internal and external Azimuth Information, perform null three operations.RMSEAnd preview the coordinates and precision of each point with the same name. A detailed precision report is provided. PreviewRMSERemove the points of the same name with a large error, and re-calculate null three until the requirements are met. Accept the null result and update the internal and external orientation information.
(4) Generate DTM
LPSExtractDTMIt is a fully automatic processing process with detailed contentDTMExtract reports. You can select to generate a singleDTMOr directly generate an embeddedDTM. Its accuracy is suitable for making normal images. To obtain high-precisionDTM, You still need some manual editing (for exampleTeModule) and good quality control.
(5)Image normal shooting correction
Efficiency has always been one of the most important issues for remote sensing image processing users,ERDASParallel Computing Technology and distributed processing are implemented through multiple processing or multiple computers, so that users can make full use of hardware investment in the data processing process to improve the efficiency of data production.LPSSupports parallel computing and distributed processing of batch processing during image re-sampling.
(6)Mosaic of positive shot correction results
Endotoxin 2010The mosaic makes full use of operating system resources, improving the mosaic speed10And supports input and output of Massive images. The stitching line is automatically generated and editable.AlgorithmAvailable.
Result
An operator uses a normal workstation to process 473 pieces of drone data. After 26 hours of processing, the automatic calculation time is 22 hours, and the accuracy of null three can be achieved.3Pixels. The results of LPS processing data are shown in:
Figure 3LPSProcessing drone data mosaic results
LPSSuggestions for processing drone data:
(1) If some images are blurred. This may affect the matching of image connection points in the future, so the data with poor image quality is removed.
(2) drone data provided includes corner data of some routes. The inertial data corresponding to this data is not accurate enough. Therefore, corner data will be removed in actual operations, to ensure the accuracy of the calculation of the entire test area.
(3) Remove drone dataPosData with sudden changes in information (especially angle information) ensures the accuracy of connection point matching and the final result.
(4) ensure that the photo is POS the data exactly corresponds to each other. Due to POS input problems, POS the information does not match the photo.
(5)To improve the accuracy of the results, we recommend that you use a secondary optimization strategy. First, according to the originalPosInformation, automatically generate the connection point, perform null 3, accept the internal and external orientation information of the optimization, and then automatically generate the connection point again based on the results of the first optimization, perform null 3 calculation.
Conclusion
Using LPs to process drone data and using a rigorous Beam Method for area network adjustment, we can use less manual intervention to ensure a certain degree of accuracy, it can process the entire test area as quickly as possible. Supports parallel computing and distributed processing of batch processing in the Process of image re-sampling, which improves the efficiency of data production and does not require high hardware. Generally, the workstation can process the data, this satisfies the needs of UAV data in Surveying and Mapping/Mosaic/DTMSuch as surveying and mapping products) and the need for rapid disaster emergency response, providing a new solution for the UAV data processing market.