Generally, in order to recognize the behavior of pedestrians in CCTV images, it is necessary to detect objects from the images.
However, occlusion problems arise due to the loss of three-dimensional information caused by projecting a three-dimensional real world onto a two-dimensional image.
This occlusion problem needs to consider the occlusion problem in the object detection process when one object is covered by another object.
Therefore, in this project, in order to recognize the human behavior from the CCTV image, depth information is added to the existing RGB information to detect the moving object by solving the obstruction problem in the object detection process.
Video surveillance, CCTV, RGB-D, Object Detection, Segmentation, Depth Map
