针对智能视频监控中运动目标的检测，提出了一种基于字典学习的背景建模方法．结合时空域信息对视频中的每个位置进行字典学习来描述场景信息．利用背景频繁出现的特性，将字典中的词划分成描述背景的词和描述前景的词．用字典表达对应位置的结构，并根据字典中贡献最大词的属性对当前位置进行背景判断．根据判断的结果对字典进行实时更新．在公共的视频数据库上与传统的背景建模方法相比较，所提方法可以较好地检测出前景目标．-Intelligent video surveillance to detect moving targets against dictionary-based learning presents a background modeling method. Spatial information for each position in the video representation of the scene dictionary learning information combined. Frequent use of background characteristics, will be divided into descr iptive words in a dictionary of words and word that describes the foreground background. Structure corresponds to the location of expression dictionary, and the background of the judgment based on the current position of the largest contribution to the dictionary word attributes. According to the results of a dictionary to determine real-time updates. In the public video database compared with conventional background modeling methods, the proposed method can better detect the foreground object