CAJ | 학술논문

提出一种改进的分水岭运动目标检测方法。进行背景差分，利用Otsu算法进行阈值处理和形态学去噪，得到初始二值化的运动目标区域；对前景和背景对象进行标记，在此基础上采用分水岭算法进行运动目标的精细分割，根据分割产生的彩色标记矩阵的颜色特征进行目标跟踪；利用运动目标的面积和质心运动轨迹等信息对运动目标的一些典型行为进行理解，对车辆等目标的轧越分道线、异常停车、逆行等行为进行检测。实验结果表明，该算法克服了传统分水岭算法易产生过分割的缺陷，在目标聚集或粘连时能实现多目标的稳定跟踪，可对车辆等运动目标的异常行为进行判别。
제출일충개진적분수령운동목표검측방법。진행배경차분，이용Otsu산법진행역치처리화형태학거조，득도초시이치화적운동목표구역；대전경화배경대상진행표기，재차기출상채용분수령산법진행운동목표적정세분할，근거분할산생적채색표기구진적안색특정진행목표근종；이용운동목표적면적화질심운동궤적등신식대운동목표적일사전형행위진행리해，대차량등목표적알월분도선、이상정차、역행등행위진행검측。실험결과표명，해산법극복료전통분수령산법역산생과분할적결함，재목표취집혹점련시능실현다목표적은정근종，가대차량등운동목표적이상행위진행판별。
A moving objects detection algorithm based on improved watershed method was proposed .Background subtraction was applied ,and initial denoised threshold image which contained moving objects regions was obtained using Otsu algorithm and mor‐phologic methods .Foreground and background were labeled respectively .Based on above results ,moving objects were segmen‐ted precisely using watershed algorithm .These objects were tracked according to the color features of marked matrix that created in watershed segmentation .Finally ,some typical behaviors of moving objects were recognized according to the objects’ areas and the trajectories of objects’ centroids .The abnormal behaviors such as illegal parking ,rolling the solid line ,as well as retrogra‐ding were detected .Experimental results show that the algorithm can overcome the defect of excessive segmentations ,and can realize robust tracking when objects are clustering or adhesive .The system also can effectively detect abnormal behaviors of moving vehicles .