ABGS Segmenter: pixel wise adaptive background subtraction and intensity ratio based shadow removal approach for moving object detection

Background subtraction approaches are used to detect moving objects with a high recognition rate and less computation time. These methods face two challenges: selecting the appropriate threshold value and removing shadow pixels for correct foreground detection. In this paper, we solve these challenges by proposing a new background subtraction method called ABGS Segmenter, which is based on a two-level adaptive thresholding approach where a reference frame is created using mean-based thresholding to generate the initial value of the threshold and accelerates the process of foreground segmentation for remaining frames by adaptively updating the threshold value at the pixel level. ABGS Segmenter is also capable of removing shadow pixels by fusing the chromaticity-based YCbCr color space model with the intensity ratio method for improving the percentage of correct pixels' classification measure. Comprehensive experiments are evaluated on three benchmark datasets (Highway, PETS 2006, and SBU) and observed that the proposed work achieves better results than existing methods.

Automated summary

In this paper, a new background subtraction method called ABGS Segmenter is proposed to detect moving objects with high recognition rate and less computation time. The method solves the challenges of selecting the appropriate threshold value and removing shadow pixels by using a two-level adaptive thresholding approach and fusing color space models.