Moving obstacle detection using images

摘要

Systems and methods for identifying moving objects from received images are disclosed. Images are received from a stereo image capture device and from a moving image capture device. Distances from the stereo image capture device to points in a captured image are stored in a stereo distance map and distances from the moving image capture device are determined from pairs of images captured by the moving image capture device and stored in a moving distance map. Stereo disparities are determined from distances in the stereo distance map and motion disparities are determined from distances in the moving distance map. A scale is generated from the stereo disparities and the motion disparities. The scale is applied to the motion disparities and the scaled motion disparities and the stereo disparities are used to identify pixels in an image associated with a moving object.