目标跟踪与识别技术是计算机视觉领域中的关键技术，该技术融合了图像识别、人工智能、图像处理等多门学科，在公共安全、视频监控、交通智能系统等领域具有广泛应用。本文在对运动目标跟踪与识别算法研究的基础上，搭建了实验平台，并通过实验验证提出算法的鲁棒性。 首先，在运动目标跟踪方面，本文重点研究了基于背景加权直方图的均值漂移目标跟踪算法，针对现有算法在出现近似物干扰以及运动目标与背景相似时跟踪准确度不高的问题，提出了将皮尔逊相关系数与巴氏系数结合，从而提高跟踪的准确度。在存在有背景扰动的实验中与Mean shift和CBWH算法进行比较，实验结果表明：本文提出的算法在背景扰动的情况下，跟踪准确度较高。 其次，在多源运动目标跟踪识别方面，着重研究了SIFT匹配算法，并且在运动目标识别的过程中，将帧间差分法、背景加权的均值漂移跟踪算法以及SIFT算法相结合。第一步，手动框选需要跟踪识别的目标，得到框选区域模板；第二步，对第二段视频运用桢差法，找到运动目标；第三步，用上述得到的运动目标模板与桢差法得到的运动目标进行SIFT匹配，找到运动目标；第四步，在第二段和第三段视频中找到运动目标。 最后，实验结果表明，本文提出的算法能够在多源环境中准确跟踪识别出感兴趣的运动目标。

Moving target tracking and recognition is a key technology in the field of computer vision, which combines image recognition, artificial intelligence, image processing and other subjects and used widely in the public safety, video surveillance, intelligent transport system. On the basis of moving target tracking and recognition algorithms, the paper put forwards the improved algorithms, and validated the effectiveness of the improved algorithm by experiments. First, in the terms of moving target tracking, the paper mainly studied the corrected background histogram based on Mean shift algorithm, and aiming at the problem of the existing algorithms in case of analogs in approximate interference and similar background with the target, the accuracy is not high. The paper combined the Person and Bhattachayya to improve the accuracy. Compared with the Mean shift algorithm and CBWH in the presence of the inference, the result showed that the improved algorithm has the high accuracy. Second, in terms of moving target recognition, the paper mainly studied the SIFT, and combined the interframe difference method, CBWH and SIFT in the moving target recognition in the field of multiple cameras. First of all, manually selected the target and get the target template; second, using the interframe difference method to get the moving target, then, using the SIFT between target template with the moving target to get the accurate target；finally，find the moving target in the second and third video. The result showed that the improved algorithm can find the target in the second and third video. At last, the result showed that the algorithms propose in the paper can track and recognize the target accurately.