摘要： 视频运动目标检测是数字视频处理、分析应用的一个重要领域，其目的是把作为一个整体的视频图像序列，通过一定的方法挖掘出具有意义的运动实体数据。该文对传统阈值法的缺陷进行分析，采用改进的二维阈值结合遗传的方法提高求解寻优的速度和效率，并通过帧差结合背景补偿的方式，提出一种适合于在复杂背景环境下实时检测运动车辆的新方法。实验结果表明，该方法有较强的环境适应能力，能够很好地检测出运动车辆。

关键词: 阈值化, 遗传算法, 帧差, 背景补偿

Abstract: In the digital video signal processing, the technology of video moving object detection is very important. Mining meaning moving objects from a video image sequence by a certain method is its intention. In this paper, according to the analysis of the deficiency of the traditional threshold method, an improved 2D gray-level histogram combined with genetic algorithm is applied to enhance the speed and efficiency. A new method based on adaptive background subtraction and frame difference for the real-time detection of moving cars in complex scenes is proposed. Experimental results show that the new method has better environmental adaptive ability and can detect moving cars well.

Key words: thresholding, genetic algorithm, frame difference, background compensation

中图分类号: 

• TP18