This study proposes an approach to segment human object from a depth imagebased on histogram of depth values. The region of interest is first extractedbased on a predefined threshold for histogram regions. A region growing processis then employed to separate multiple human bodies with the same depthinterval. Our contribution is the identification of an adaptive growththreshold based on the detected histogram region. To demonstrate theeffectiveness of the proposed method, an application in driver distractiondetection was introduced. After successfully extracting the driver's positioninside the car, we came up with a simple solution to track the driver motion.With the analysis of the difference between initial and current frame, a changeof cluster position or depth value in the interested region, which cross thepreset threshold, is considered as a distracted activity. The experimentresults demonstrated the success of the algorithm in detecting typicaldistracted driving activities such as using phone for calling or texting,adjusting internal devices and drinking in real time.
展开▼