一种直方图均衡化的五边形拟合老年人异常行为检测算法的研究

老年人异常行为检测是计算机智能视频监控的一项重要功能，其快速有效地检测老年人的异常行为信息，使老年人及时得到救治，本文提出一种直方图均衡化的老年人异常行为检测算法，首先根据定义的行为准则区分异常行为，然后对图像进行预处理-直方图均衡化修正直方图、前景图像和背景图像做差分、对图像进行二值化，采用形态学的算子进行膨胀连通断裂区域，腐蚀、轮廓提取。本文采用头、左手、左脚、右脚、右手五个突出顶点进行五边形拟合，边界跟踪去除影响的噪声。采用面积有效比、周长比、中心比、宽高比进行判断区分老年人的行为。经试验验证该算法提高了正确率，降低了误判率，具有很好的鲁棒性。

Research on Abnormal Behavior Detection about Older People Based on Pentagon Fitting of Histogram Proportion Equalization

Abnormal behavior detection about older people is a serious problem in intelligent surveillance, the abnormal behavior is detected fastly and effectively, the older people are rescued in time. In this paper, an algorithm which re-search on abnormal behavior detection about older people based on histogram proportion equalization is presented,which abnormal behavior is distinguished according to defined behavior norm;image preprocess is processed including with his-togram proportion equalization;the difference of foreground and background of the frame image;the image binarization, fracture area being dilated and connected by morphology operator, being eroded and contour extraction. In this paper, utilizing head, left hand, left foot, right foot, right hand five important vertex fit to pentagon, noise is deleted by boundary tracking . The older people’s behavior is judged by using area ratio, perimeter ratio, barycenter ratio, width and height rati.In this paper, using fewer vertex, using more judge rule, the experiment verify that the algorithm im-prove true ratio,decline false ratio,improve robustness in behavior detection.