摘要
针对现有异常行为识别方法在车内场景应用少,并且受车内空间狭小、异常行为复杂多变等影响导致识别有效性差等问题。在Alpha pose模型提取驾乘人员骨架关键点基础上,构建驾乘人员人体坐姿模型,采用关键点位置信息描述异常状态,最后利用概率学习模型将位置信息转换为概率对行为进行识别分类。经实验测试,该方法对车内前排人员异常行为的识别准确率能够达到90%以上,且具有一定的实用价值。
In view of the fact that the existing abnormal behavior identification methods is rarely used in the car scene,and due to the small space in the car,abnormal behaviors are complex and changeable,resulting in poor recognition effectiveness and other problems,the Alpha Pose model was used to extract the key points of the driver s skeleton,the sitting posture model of the driver s body was constructed,and the position information of the key points was used to describe the abnormal state.Finally,the probabilistic learning model is used to trans form location information into probability to identify and classify behaviors.The experimental results show that the recognition accuracy of the method can reach more than 90%,and it has a certain practical value.
作者
赵雄
陈平
潘晋孝
ZHAO Xiong;CHEN Ping;PAN Jinxiao(Shanxi Key Laboratory of Signal Capturing and Processing,North University of China,Taiyuan 030051,China)
出处
《机械与电子》
2021年第3期10-15,共6页
Machinery & Electronics
基金
国家自然科学基金资助项目(61801437,61871351,61971381)
山西省自然科学基金资助项目(201801D221206,201801D221207)
山西省研究生创新项目资助(2020BY098)。
关键词
Alpha
pose
坐姿骨架关键点
概率学习模型
车内异常行为识别
Alpha pose
key points of sitting skeleton
probabilistic learning model
abnormal behavior recognition in the car
