Investigation of Moving Object Detection Using Diffusion Layered ABM

In this paper, we propose DLABM-MP and DLABM-MA, which are methods that can be applied to moving cameras by using the optical flow in the video to perform motion compensation, which is diffusion in the time direction, and the spatial direction diffusion of ViBe, in addition to the conventional method DLABM. The conventional method DLABM was proposed as a method that can quickly respond to background changes using the diffusion of ViBe. However, it is based on the assumption of a stationary camera and does not assume a moving camera. To tackle this issue, we propose methods to perform the same motion compensation and diffusion in separate processes. The proposed method DLABM-MP incorporates the idea of ViBe diffusion to enable quick response to background changes. In contrast, the proposed method DLABM-MA incorporates diffusion that considers the LABM background model to allow it to respond to moving cameras. We evaluated the performance of the proposed methods DLABM-MP and DLABM-MA using a publicly available image database for evaluation.