Research on Ortho-rectification of High-resolution Satellite Images Based on GPU
Aiming at the problem of low processing efficiency of ortho-rectification for high-resolution satellite images,an ortho-recti-fication method based on GPU for high-resolution satellite remote sensing images is proposed. Limited by the global memory capacity of GPU,DEM data and original image were processed by rectangular blocks in batches. The processing time of ortho-rectification was greatly shortened by optimizing the configuration and memory of GPU. In execution configuration optimization,the number of threads in the thread block were optimized. In memory optimization,RPC parameters were stored in constant memory,GPU was adopted the computing workloads mode,and the on-chip cache was divided into 32 KB shared memory and 64 KB L1 cache. Experimental results show that the proposed method is about 8.6 times compared with CPU multithreading on the time acceleration ratio,and it can greatly improve the efficiency of ortho-rectification of high-resolution satellite images and meet the requirement of rapid emergency response.
万方数据
维普
