基于区域划分与标准时间的手部异常行为检测

通过监控视频自动检测操作任务中手部异常行为，能够预防人因差错，提高人因可靠性。针对手部操作任务运动特征不明显、常用异常检测和手势识别方法不适用的问题，提出基于区域划分与标准时间的手部异常行为检测技术。使用基于椭圆模型的肤色检测方法，检测视频中手部形心位置；提出工作区域划分方法，根据手部形心在操作过程中所处区域的不同，将连续操作分割为单元任务，获得各段单元任务的起止时间和持续时长；以正常工作时间为标准，对超出标准时间范围的单元任务提出异常警告。实验表明:所提方法的单元任务分割正确率高于93%，异常行为检测率高于86%，能够有效检测手部异常行为，为人为差错的监测与预警提供技术支持。 

Abnormal hand behavior detection based on area division and standard time

Abnormal hand behavior detection during operation based on intelligent video surveillance systems can prevent human errors and improve human reliability. In order to solve the problems that the motion characteristics of hand operation are not obvious, and common abnormal detection and gesture recognition methods are not applicable, a detection technology of abnormal hand behavior based on area division and standard time is proposed. The skin color detection method based on ellipse model was used to detect the hand centroids in the video. The work area division method is proposed to define the unit task. The continuous operation was divided into unit tasks according to the work area of the hand centroid in each frame, and the start and end time and the duration of each unit task were obtained. Standard time was defined by normal working hours. And warnings were given to the unit tasks which exceed the standard time range. Experimental results show that the accuracy rate of unit task segmentation is higher than 93%, and the detection rate of abnormal behavior is higher than 86%. The proposed method can effectively detect abnormal hand behavior and provide technical support for human error monitoring and early warning.