The tracking effect of the traditional tracking algorithms is poor due to the low video resolution,blurred frame image or heavy noise. To solve the above problems,a target tracking algorithm combining hybrid feature matching with Viterbi data association is proposed. The histogram back⁃projection technology is used to effectively segment the target edge in local bi⁃threshold image to overcome the problem of low contrast. The neighborhood feature,regional feature,movement direction feature and histogram feature are taken as the target characterization features to establish the cost function of hybrid feature. The Viterbi data association is used to calculate the sum of cost function to obtain the most similar target. The experimental results show that,in the condition of blurred frame image or heavy noise,the proposed algorithm has stable and effective target tracking,the accuracy rate of single target tracking is 0.89,the accuracy rate of multi⁃target tracking is 0.975,the recall rate is 0.920,and the algorithm is superior to other similar tracking algorithms.%传统跟踪算法在视频分辨率低、帧图像模糊或噪声较多时跟踪效果较差。针对此情况，提出一种混合特征匹配结合Viterbi数据关联的目标跟踪算法。首先，采用直方图反向投影技术对双局部阈值图像中的目标边缘进行有效分割，克服了低对比度问题；然后，将邻域特征、区域特征、运动方向特征和直方图特征作为目标表征特征，建立混合特征代价函数；最后，采用Viterbi 数据关联计算代价总和，求得最相似目标。实验结果表明，在帧图像模糊或噪声较多的情况下，目标跟踪稳定且有效，单目标跟踪准确率为0.89，多目标跟踪精度达0.975，召回率达0.920，优于其他几种同类跟踪算法。
展开▼
