A Statistical Method for Generic Foreground Detection

Abstract

Traditional approaches such as Gaussian mixture model (GMM), Otsu's and moment preserving (MP) methods are developed for segmentation of opaque objects. For semi-opaque objects like flame and smoke the result is cluttered, due to inappropriate threshold, especially if one dominates the other. Besides, rapidly changing environments like foggy and rainy scenes increase the difficulty in foreground detection. We propose a statistical method for the detection of both ＂opaque objects＂ and ＂semi-opaque objects＂ that works in all weather conditions. We use difference of histogram and ANOVA for candidate foreground detection and Student's t-test for object segmentation. Experiments are conducted for both ＂opaque＂ and ＂semi-opaque objects＂ under both clear and severe weather conditions. The results show that for ＂opaque objects＂, the recall of the proposed is 0.941, while for ＂semi-opaque objects＂, the recall is 0.895. In the scenes where both types of objects exist, the recall remain at 0.901. In severe weather conditions, the recalls are 0.93 and 0.88 for ＂opaque＂ and ＂semi-opaque objects＂, respectively.