DESIGN AND IMPLEMENTATION OF IMAGE FEATURE PARALLEL COMPUTING SYSTEM BASED ON FGPA

The extraction of image features is the basis for the realization of image tracking algorithms. In the field of civil radar, an algorithm for pedestrian recognition by fusion of visual information and millimeter-wave radar information has emerged. The fusion of visual feature information with radar information is a development direction in the field of target detection in the future. This paper introduces an feature calculation system, which uses the high parallelism of FPGA to complete the calculation of HOG and LBP features and uses advanced fusion to quickly complete the fusion of the two features. At the same time, the system designed an improved double-pipeline structure to calculate the HAAR feature. The final feature calculation system is implemented on the Zedboard of Xilinx. And apply it to a specific tracking algorithm, the algorithm processing efficiency is greatly improved.