For the problems of complex structure, high power consumption, large size and poor anti-interference in the traditional DSP+FPGA design, a real-time multi-target tracking system based on a System on Programmble Chip(SOPC) was established and a software optimization method based on NIOS II core and its data link was proposed. The optimization technologies for algorithm structure, data link configuration, code optimization, custom instruction and C2H were researched. Then, these technologies were applied to the multi-target tracking SOPC system to optimize algorithm structure, reduce the size of code，remove the correlation of code and implement link-layer hardware conversion to enhance the performance of software pipelining. Experimental results indicate that these optimizing technologies can improve the efficiency of software by 9-18 times. The computing time of multi-target tracking is 1.3 ms/frame and the navigation operation time is 470 μs/cycle for the image sequence of 256 pixel×256 pixel in 100 frame/s. The system can realize real-time processing with low hardware costs, simple structure and stable performance.