Abnormal Object Detection Using Feedforward Model and Sequential Filters

Detection of stationary foreground objects: A survey

“…2, there are no strategies focused only in the detection of SFOs of type C. However, many strategies to detect multiple types of SFO have been proposed. Some algorithms try to detect types A and B [56,65,83], whereas other approaches are focused on the detection of the 3 types of SFOs [29,93,101].…”

“…Many of these specific mechanisms [8,31,40,62,83,84,101] rely on the analysis of the edges of both the current image and a BG image (point estimate of the BG that is typically obtained from the BG model). The intuition is that, in many cases, if the BG is covered with a A C C E P T E D M A N U S C R I P T SFO the number of edges in the image will grow.…”

“…The median filtering proposed in [119], which is able to classify among FG, shadows, and subtracted objects, has been adopted by some authors [30,83] to detect SFOs. In the case of [30], the original method was modified to selectively exclude stopped vehicles from the BG updating.…”

“…However, there are other works that consider more classes. For example, the classifier in [83], which is based on area, saliency and motion information, is able to classify among MOs, ROs, shadows cast by MOs or shadows cast by ROs.…”

“…For example, in [83] the ROs are detected thanks to a post-processing stage based on the analysis of the edges of the detected SFOs (see section 7).…”