摘要
针对驾驶员行车时双手离方向盘和接打电话等不安全操作行为的问题,提出了一种结合骨骼与肤色模型的不安全操作检测方法。该方法首先采用背景差分法检测出驾驶员区域,然后采用基于深度信息的骨骼定位方法来检测出驾驶员的手和头部区域,接着分析骨骼运动信息来定位驾驶员的手部动作,同时辅助肤色模型实现驾驶员手部的精确定位,进而识别出驾驶员接打电话或双手离开方向盘等不安全操作行为。实验结果证明骨骼与肤色模型结合的方法削弱了骨骼定位时关节节点重叠出现的抖动以及肤色检测时光照强度等外部条件的影响,具有很好的鲁棒性和实时性。
A detection method of drivers＇ unsafe operation was proposed, using the skeleton information combined with the skin color model to detect driver＇ s hands off the steering wheel or make a call when they are driving. Firstly, the background difference method was used to detect the pilot＇ s area. Then the skeleton positioning based on skeleton of depth information was used to detect the driver＇ s hands＇ and head＇ s location; through the analysis of the skeleton movement information combined with the skin color model, the location of driver＇ s hands could be accurately positioned. Finally unsafe operation behaviors including the driver＇ s hands off the steering wheel or having a call were recognized. The experimental results indicate that the proposed method using the skeleton information combined with the skin color model has good effects which weakens shake for joint nodes when the skeleton position overlapping and the influence of the external condition such as light intensity when skin color being detected. The proposed method has good robustness and real-time performance.
出处
《计算机应用》
CSCD
北大核心
2016年第A02期165-167,192,共4页
journal of Computer Applications
基金
北京市青年选拔人才培育计划项目(CIT&TCD201404009)
科技创新服务能力建设-协同创新中心-首都世界城市顺畅交通协同创新中心-参与单位(2011协同创新中心)(参与)(市级)(PXM2016_014212_000030)
关键词
不安全操作行为
深度信息
背景差分
骨骼定位
肤色检测
unsafe operation behavior
depth information
background difference
skeleton positioning
skin color detection
