自适应双阈值的运动目标检测算法

An Adaptive Double Thresholds Algorithm of Detecting Moving Objects

针对噪声、不同的天气状况和光照强度等环境变化对运动目标检测的影响,提出了一种自适应双阈值运动掩膜算法.为提高复杂环境条件下运动目标检测的识别率,该算法首先利用多帧平均法初始化背景,采用函数链接型神经网络算法动态更新高低两个阈值,自动适应光照变化.根据运动掩膜算法判定前景和背景区域动态更新背景后,采用自适应双阈值背景差法分割得到前景目标区域,并结合数学形态学方法,消除阴影,准确识别出前景目标.实验结果验证了该算法对运动目标检测的高准确性和良好的鲁棒性.

Due to the environmental change of the noise, different weather conditions and illumination, which influence the results of moving object detection, this paper proposed an adaptive double thresholds motion ob- ject mask algorithm. To improve the rate of motive vehicle recognition, this novel method first used multiple- frame average algorithm to initialize the background, and adopted functional chain neural network method to update the two of high and low thresholds dynamically, which can adjust to changeable illumination automatic- ly. According to the motion mask algorithm, the region of the foreground and background was identified and the current background was updated. Then the region of the foreground object could be attained by dynamic double thresholds background difference method. Combined with the mathematical morphology method, much shadow was deleted and the foreground object was recognized correctly. The experimental results demonstrated that this detecting algorithm was more accurate and robust.