CAJ | 학술논문

为了提高监控视频中人体异常行为识别的实时性和准确率，提出了基于运动特征的人体异常行为识别方法。利用分块更新的背景差法从图像中提取出完整的人体轮廓，通过区域关联结合颜色直方图实现人体目标跟踪，解决了非线性运动时漏跟和误跟的问题。通过人体运动轨迹、运动姿态及运动时间3个参数，对人的5种异常行为进行分析判断。实验结果表明，所提算法不仅能实时地对人体进行检测和跟踪，还能快速、准确地识别出异常行为，具有简单实用的特点。
위료제고감공시빈중인체이상행위식별적실시성화준학솔，제출료기우운동특정적인체이상행위식별방법。이용분괴경신적배경차법종도상중제취출완정적인체륜곽，통과구역관련결합안색직방도실현인체목표근종，해결료비선성운동시루근화오근적문제。통과인체운동궤적、운동자태급운동시간3개삼수，대인적5충이상행위진행분석판단。실험결과표명，소제산법불부능실시지대인체진행검측화근종，환능쾌속、준학지식별출이상행위，구유간단실용적특점。
A characteristics-based human abnormal behavior detection method is proposed to improve accuracy and real-time per-formance of human abnormal behavior detection in surveillance video.Chunked update-based background subtraction technique is used to extract complete human silhouette from the image.The regional association is combined with color histogram to track hu-man target to solve the miss tracking and target-missing problems in nonlinear movement.Three parameters of human movement trajectory,moving posture and time of movement areselected to identify five human abnormal behaviors.In the experiments,the human body is detected and tracked in real time,and the abnormal behaviors could be identified quickly and accurately using pro-posed algorithm.