Enhanced dynamic background updating MOD algorithm for harsh environmental conditions

This paper presents an algorithm which is capable of detecting moving objects in challenging environmental conditions. This work is mainly based on the recently published Dynamic Background Updating (DBU) for lightweight moving object detection. In the DBU algorithm, a combined temporal difference and background subtraction technique that exploits both frame-based and pixel-based background updates is used to detect some moving objects in the application scenario with a static camera. Even though its complexity can be significantly reduced, the main drawback of DBU is its lower performance compared to the GMM-based algorithms at harsh environmental conditions which have a non-uniform illumination distribution or extremely high/low light. By presenting a pre-pixel processing of adaptive histogram equalization and a post-pixel processing of bounding box interpolation, this paper proposes an Enhanced DBU (EDBU) algorithm to accurately detect moving objects even in harsh environmental conditions. Experimental results show that the proposed algorithm does not only increase the detection rate of the previous DBU algorithm but also exceeds the performance of GMM-based algorithms by at least 15%.