改进Camshift算法的DSP硬件实现目标跟踪方法

针对Camshift算法只对前一帧预测而导致的目标像素脱靶现象,以及目标像素在帧间位移较大的问题,本文提出一种改进Camshift算法的目标跟踪方法.该算法将加权背景直方图和贪心算法融入Camshift算法,利用贪心算法对前两帧图像信息进行处理,预测出目标在当前帧图像中的位置,再根据目标颜色概率,用Camshift算法找到目标的真实位置,最后在TMS320DM642(数字媒体应用的定点DSP)上对该文算法进行硬件系统的实现,并使用EDMA(增强型直接内存访问)方式和Cache技术对系统进行优化.实验结果表明,与传统Camshift算法相比,该文方法在背景与目标相近的情况下跟踪效果更佳,具有很好的鲁棒性和稳定性,适用于复杂环境下的目标跟踪.在系统实现上,优化后的系统平均帧率提升在3帧/s以上,增强了算法的速度.

Target Tracking Method Based on DSP Hardware Improved Camshift Algorithm

In order to solve the problem that the target pixel is lost caused by the prediction of the previous frame of the Camshift algorithm and the target pixel is greatly shifted between frames, a target tracking method based on improved Camshift algorithm is proposed in this paper. The algorithm combines the weighted background histogram and the greedy algorithm into Camshift algorithm, and uses the greedy algorithm to process the image information of the first two frames to predict the position of the target in the current frame image. Then the Camshift algorithm is used to find out the target''s reality according to the target color probability Location. Finally, the hardware system is implemented on the TMS320DM642. Then using EDMA and Cache technology optimize the system. Compared with the traditional Camshift algorithm, Experimental results show that the proposed method has better tracking performance and better robustness and stability under the condition of similar background and target, which is suitable for the target tracking in complex environment. In the system implementation, the average system frame rate of the optimized system is increased by more than 3 frames/second, which increases the speed of the algorithm.