实际应用中对目标跟踪的实时性要求越来越高。针对这个问题，设计并实现一种基于FPGA的Mean Shift跟踪系统。针对FPGA硬件平台的浮点运算复杂度高的特点，对核函数和权重计算进行优化，使用定点运算替代浮点运算。在处理同样分辨率的视频/图像数据时，与通用CPU E7400相比，该系统可使得性能有很大的提升。采用此方法大大提高了Mean Shift跟踪算法的计算速度，满足实时性的要求。%Thereal-timerequirementfortargettrackingisgettingincreasinglyhigherinpracticalapplication.Aimingatthisissue,we design and realise an FPGA-based Mean Shift tracking system.Considering the characteristic of the FPGA hard platform in high complexity of floating-point operations,we optimise the kernel function and weight computation and uses fixed-point operations instead of floating-point operations.Compared with CPU E7400,the performance of this system can be greatly improved when processing the video or image data with same resolution.The use of this method greatly improves the computational speed of Mean Shift tracking algorithm and meets the real-time requirements.
展开▼
