Radar Image Matching/Reverse Positioning Based On Feature Points And Real-time Implementation

In the comprehensive processing and application of remote sensing images,image registration / matching technology is a pre-processing step,and its importance is self-evident.The matching of homologous / heterogeneous images is in image matching guidance,surface change detection,image fusion,etc.It has broad application prospects.As an all-weather and all-weather remote sensing device,radar has different image characteristics due to different observation angles,making image matching more difficult.This paper mainly studies the radar image matching algorithm and the platform inversion method based on the image matching results.An improved image preprocessing method for downsampling data fusion is proposed,which improves the success rate of the feature point matching algorithm in processing radar images when the resolution is different and the grid difference is large,and is based on the matching result of the image after In SAR terrain correction Combining In SAR observation geometry,a radar platform anti-location method is proposed.In view of the large amount of radar image data and the characteristics of computing resources and memory resources required by the feature point matching algorithm applied to remote sensing images in different observation scenarios,an adaptive memory management method based on the TMS320C6678 multi-core DSP platform is proposed to achieve feature point-based The hardware parallel acceleration of the image matching method improves the efficiency and timeliness of radar image matching.The main work of this article is as follows:1.Aiming at the shortcomings of traditional feature point matching method when dealing with radar images with different resolutions and large grid differences that are easily mismatched,an improved pre-processing method for downsampling data fusion is proposed.The grid / resolution of the reference image is down-sampled to approximately the same,and then the feature point matching algorithm is combined to effectively improve the matching success rate of the radar image pair,reduce the calculation amount of the algorithm,and in-depth analysis of each processing step of the feature point matching method.The impact of matching results,and the use of simulation data and airborne measured data on the influence of error factors were tested and verified,and the effectiveness of the improved method in this paper was also verified.2.A radar carrier inverse positioning method based on In SAR terrain correction image matching results is proposed.The digital elevation map obtained by In SAR can effectively correct the geometric distortion of SAR images introduced due to terrain fluctuations,and improve the matching probability and matching accuracy.Using the image matching results and In SAR observation geometry,the relative coordinate information of the region to be matched obtained by the observation is mapped to the absolute coordinate system of the reference image to complete the reverse positioning of the radar carrier,and the measurement errors of various parameters are analyzed for the radar based on the In SAR observation geometry The impact of platform anti-location methods.3.Based on the TMS320C6678 multi-core DSP platform,aiming at the feature that the hardware resource requirements of each module in the radar image matching algorithm based on feature points change dynamically when processing different images,an adaptive memory management method is proposed to avoid memory fragmentation and make full use of data Bit wide.Design a hardware parallel framework that can be reused by each module.Implement and optimize the improved feature point matching algorithm in this paper in parallel to achieve the purpose of balancing computing load,reducing the coupling between modules,and improving the real-time feature point matching algorithm.Based on simulation and The hardware operation results of the measured data verify the hardware acceleration effect of the improved method.