| Space-borne topography surveying technology is widely used in earth observation and space exploration,among which stereo photography and laser altimeter are two commonly used technologies.Stereo photography has high spatial resolution,but it is difficult to improve elevation accuracy without ground control points.Laser altimeter has high accuracy in elevation,but the spatial resolution of cross track is insufficient.Topography surveying based on the combined adjustment of laser altimeter ground control points and stereo photography is widely studied.However,the traditional laser altimeter has large pulse power consumption and sparse footprints,and it is difficult to provide accurate laser control points for combined adjustment in mountainous areas.The new generation of spaceborne photon counting laser altimeter represented by ICESat-2(Ice,Cloud and land Elevation Satellite 2)shows the advantages of micropulse,high repetition frequency and multi-line push.In addition to providing traditional control points for stereo images,it can also provide terrain profile constraints in mountainous areas.However,photon counting laser altimeter still has problems,such as low signal-tonoise ratio(SNR),and its fusion method with stereo image has not been fully studied.This dissertation mainly studies the key technologies of topography surveying based on the fusion of photon counting laser altimeter data and stereo images.The aim is to develop the data processing technology of spaceborne photon counting laser altimeter and improve the elevation accuracy of stereo photography through active and passive data fusion.The research contents of this dissertation are as follows:(1)Research on signal photon extraction method of photon counting laser altimeter in multiple paradigms.ICESat-2 data show that,under the influence of diverse global topography,weather and illumination,low SNR signal photon extraction and adaptive signal photon extraction still face many challenges in complex scenes such as mountainous areas.In this dissertation,signal extraction in paradigms of traditional manual algorithm design and machine learning are studied.Under the design paradigm of traditional manual algorithm design,the problem of signal photon extraction under extreme conditions such as strong background noise,weak echo signal and complex topography was studied,and a multi-scale constrained signal extraction method was proposed.The experimental results show that the F value of the multi-scale constraint method is improved by 11%-20% and 2%-3% in weak light channel and strong light channel,respectively.The process of data analysis and algorithm coding in manual algorithm design is cumbersome,while machine learning can autonomously learn the distribution characteristics of data and generate the required processing methods,but there are few studies on the extraction of photon counting signals.In this dissertation,a partition convolutional point cloud segmentation network is designed for the complex structure of the traditional point cloud segmentation network,which does not combine the characteristics of photon counting measurement.The experimental results show that the prediction efficiency is 5-8 times that of the traditional point cloud segmentation network,and the F value is 1%-40% higher than that of the traditional manual design algorithm.(2)Pointing angle calibration based on small-range terrain iterative matching.Pointing calibration is an important step to eliminate the systematic errors of laser measurement and runs through the whole mission cycle.The existing pointing calibration methods have problems such as complicated process or large amount of calculation,this dissertation,based on the characteristics of high repetition frequency of photon counting laser altimeter and the popularity of unmanned aerial vehicle measuring equipment,a small range terrain iterative matching pointing angle calibration method is proposed,which calibrates the pointing error uses local high precision airborne laser data.The experimental results show that with a 2.5 km trajectory,the pointing calibration error is about 0.1 arc second,and the calculation time is only 1/6 of the traditional pyramid search method.(3)Research on the combined adjustment of photon counting laser control point and stereo image.The combined adjustment is the main fusion method of laser data and image data,but the combined adjustment of photon counting laser control point and stereo image is still in the development stage.In this dissertation,the key problems in the application of photon counting laser data to stereo image combined adjustment are redesigned and optimized,and experiments are carried out based on World View-3 stereo image and ICESat-2 data.The results show that the elevation error of 3d reconstruction results decreases from 4 m to 1 m after combined adjustment.(4)Research on the matching fusion method of photon counting track and stereo image in non-flat terrain.Aiming at the difficulty of accurate extraction of laser control points in non-flat areas such as mountainous,matching fusion methods to improve the elevation accuracy of stereo image by using laser trajectory constraint is studied.Based on the traditional minimum elevation difference matching method,a maximum likelihood elevation difference matching method is further proposed,and experiments are carried out based on gaofen-7 stereo image and ICEsat-2 data.The results show that the standard deviation of elevation matching is about 0.1 m,and the standard deviation of horizontal matching is 1 m ? 2 m.In addition,the maximum likelihood elevation difference matching method can directly use the original measurement results of photon counting with noise,which reduces the data processing steps.The research results of this dissertation are helpful to the data processing technology of the photon counting laser altimeter,and the active and passive fusion method can provide a reference for the system scheme design and data processing of the high scale topography mapping missions in the future. |
PDF Full Text Request