Multicore DSP Based Infrared Target Tracking Technology Feature

With the rapid development of science and technology, as well as national attention on areas of national defense and security, with it computer vision are constantly developing and expanding, target tracking become one indispensable part. While the performance of digital signal processors also continue to increase, making previously complex algorithms and methods which difficult to handle can not meet the real-time has been resolved, especially the launch of TI’s high-performance DSP, image processing technology to enhance a highly.In this paper, the design of the target tracking system based on feature points, which contains the algorithm research, design, hardware platform, the algorithm implemented on a hardware platform to optimize the system in four parts. Most existing feature point tracking algorithms only consider two adjacent at a time, and forget the feature information of previous frames. In this paper, based on the original Kanade-Lucas-Tomasi (KLT) algorithm, a new EigenTracking method is presented that learns an eigenspace representation of training features online, and finds the target feature point with Gauss-Newton style search method, effectively avoiding the target dimension, Angle of great change and occlusion. Analyzing the performance of the algorithm by the actual image, and comparing with other algorithms. Meanwhile designing bases on the eight-core DSP (TMS320C6678) as the core processor, a combination of FPGA digital video image acquisition hardware platform, a series strategies of the algorithm optimization are proposed considering the TMS320C6678.Finally achieved on tracking the moving target by the real-time transmission of video images.