摘要
文章提出了一种基于视频图像的行人检测算法。通过动态阈值目标分割算法获得运动目标,同时将分割结果块化处理,以减少算法时间复杂度。使用最小外接矩形获得目标的高宽比,结合基于质心的目标跟踪方法获得目标物体运动速度,进而对行人事件进行检测。实验结果表明,该方法能准确地检测交通场景中的行人事件,有效地解决抖动、光线及积水等因素对行人目标检测的影响,具有良好的实时性和准确性。
An algorithms for pedestrian detection based on video image is proposed in this paper. Target segmentation by dynamic threshold is used to detect moving target, and then get the block segmentation result to reduce time complexity of the algorithm. Pedestrian detection is judged by the depth-width ratio got by the minimum bounding rectangle and the speed of the target got by the tracking method based on the centre of gravity. Experimental results show that this method, with reasonably real-time performance and accuracy, can detect the pedestrian appearing in traffic scene, and effectively solve the influence of shaking, illumination change and water.
出处
《交通节能与环保》
2015年第4期94-97,共4页
Transport Energy Conservation & Environmental Protection
关键词
自适应阈值
目标跟踪
行人检测
dynamic threshold
target tracking
pedestrian detection
