Moving Object Detection for Medical Gait Analysis in Complex Scene

摘要

中文摘要：针对医学步态分析中的复杂场景下运动目标检测问题，提出了基于贝叶斯决策规则的方法。该方法由变化检测、变化分类、前景目标提取和背景更新四部分组成。变化检测采用自适应阈值法来二值化变化点和非变化点，变化分类基于颜色共生特征向量，采用贝叶斯规则进行决策，前景对象的提取融合了时间差分法和减背景法。针对复杂场景中背景的“渐变”和“突变”情况，提出了不同的背景更新策略。实验表明，该方法在包含有摇动的树枝，或者灯的开关等复杂背景中能准确地提取运动目标，因此可用在医学步态分析的研究中。英文摘要：Abstract This paper proposes a novel method for moving object detection from a video in medical gait analysis whichcontains not only stationary background objects but also moving background objects. It consists of four parts: changedetection, change classification, foreground object abstraction, and background learning and maintenance. We use theBayes decision rule for classification of background and foreground changes based on a special feature vector--- colorco-occurrence feature. Foreground object abstraction fuse the classification results from both stationary and movingpixels. Learning strategies for the gradual and "once-off" background changes are proposed to adapt to various changesin background through the video. Extensive experiments on detecting foreground objects from a video containing waver-ing tree branches, or light open/close demonstrate that the proposed method is effective and can be used in medical gaitanalysis.