无迹粒子滤波在基于色彩的视频目标跟踪中的应用

Color Video Target Tracking Based on Unscented Particle Filter

在视频目标跟踪中把视频象素点的色彩值(RGB值)作为跟踪系统的观测量,用无迹粒子滤波算法估计活动目标在每一帧视频画面中的位置。在粒子滤波的重要性采样步骤中以色彩相似度作为更新粒子权值的标准。采用一种判断是否需要进行重采样的方法。在MATLAB环境下仿真时采取调用预先编译得到的DLL文件的方式减少仿真时间。实验结果表明,取300个粒子时,能以每秒24帧的速率跟踪随机运动的视频目标。

Unscented particle filter is used to estimate the position of a moving target in color video target tracking.The tracking system reads the colour value（RGB） of a pixel in the video image,and uses the color likelihood as a standard to update the weights of the particles which are generated in the particle filter algorithm.The calculating time was reduced because there was no calculation related with the observation function.A method of checking the necessity of resampling was used.In Matlab,the method of calling DLL files was used which were compiled in advance to reduce the simulation time.The simulation result shows that,using 300 particles,the unscented particle filter can track moving target in a rate of 24 frame per second.