运动目标的检测和跟踪技术作为视频图像处理中的一个关键技术，在人机交互、交通安全、视频监控、军事和公共安全管理等领域得到了广泛的应用。本文在对运动目标检测和跟踪算法研究的基础上，提出了一些算法的改进方法，并通过实验验证了改进方法的有效性和稳定性。本论文的主要工作如下： 首先，在运动目标检测方面，通过对已有的检测方法进行深入分析，并重点研究了基于码书模型的运动目标检测算法。码书模型是一种基于运动信息的检测方法，当目标的运动信息不足时，可能会出现误检或局部检测等问题。针对码书模型存在的不足，通过联合目标的空间整体信息，提出了一种基于码书模型的自适应背景更新算法，使其在处理缓慢移动目标和只有局部运动目标时减少误判。该方法通过对运动目标空间信息变化进行分析，寻找前景中潜在背景，然后联合像素时域统计信息，得到真正的背景模型。实验结果表明，该算法可以快速适应背景变化，提高码书模型处理能力，能明显减少对运动信息不足目标的误判，同时保证目标检测的完整性。 其次，在目标跟踪方面，本课题重点研究了基于卡尔曼滤波器和mean-shift的目标跟踪算法。针对现有基于多特征融合的跟踪算法在复杂环境下跟踪准确度不高，且大部分采用单一判定方式来实现多特征融合的问题，提出了一种多准则判定的自适应多特征融合方法。首先引入局部背景信息加强对目标的描述，然后在多特征融合过程中利用多种判定准则自适应计算特征权值。最后，在均值漂移框架下，结合卡尔曼滤波完成对目标的跟踪。在各种场景下的实验结果比较表明：本文融合算法比单种判定融合有更好的稳定性和鲁棒性，有效地提高了复杂环境下跟踪准确性。

Moving target detection and tracking has become one of the key technologies in video image processing, and has been widely used in the field of human-computer interaction, video surveillance, traffic safety and public safety management. On the basis of the researches on the moving target detection and tracking algorithms, this paper proposed some improved algorithms, and the effectiveness and the stability of the algorithms have been verified by experiments. The main works of this paper are as follows: First of all, in the moving target detection, through analyzing the existing detection methods, and further study of the target detection based on codebook model. Codebook model is a kind of detection method based on motion information. When the motion information of largest is insufficient, there may be error detection or local detection and so on. Aiming at the shortcomings of the codebook model, this paper proposed an algorithm of the adaptive background based on codebook model by the joint space overall information of the target, to reduce the error detection in the process of the slow moving target and local moving target only. The method through analyzing the changes of mining target spatial information, to find the background in the foreground, then combined with temporal statistical information to obtain the real background model. The experiments result show that, the algorithm can quickly adapt to the changes in the background, improve the processing ability of codebook model, reduce the error detection significantly when the motion information insufficient. Secondly, in target tracking, this paper focuses on the target tracking algorithm of Kalman Filter and Mean Shift. In the view of the precision is not high based on multi-environment tracking, and most of them use single way to implement the multiple feature fusion problem. A multiple feature fusion based multiple criterion decision adaptive was proposed. Firstly, introducing the local background information to s strengthen the description of the target and then using a variety of criteria adaptive to calculate feature weights in the process of multiple features fusion. In addition, the framework of mean shift combining with Kalman Filter was considered to realize target tracking. An extensive number of comparative experimental results show that the proposed algorithm has better stability and robustness than the single fusion rule and improves the tracking accuracy in complex environment significantly.