Real-Time human Activity Recognition in Outdoor Environment

• Main Authors: Yueh-Feng Lin, 林岳鋒
• Other Authors: Fang-Hsuan Cheng
• Format: Others
• Language: en_US
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/23008046037987009699

碩士 === 中華大學 === 資訊工程學系碩士班 === 90 === Video surveillance becomes more and more important recently. Segmentation of moving objects is an important part of video surveillance. To subtract two successive images is a very common method for segmentation. Here, we present a moving objects segmentation method based on luminance and hue information. First, we learn the background models of luminance and hue. Second, it subtracts current pixels from background model of luminance. Then, it subtracts current pixels from background model of hue again. Finally, we can get the segmented objects. Recognition of Human activity is a very hard work, because the human activities are too complex. We cannot recognize all type of human activities, so we only recognize six basic types of human activities. The six basic types of human activities are squat, stand up, move on, far away, bend, and walk. After we segment the moving objects, we can locate the head point and foot point. Then, we can recognize the activities by their relation.