Abandoned Object Detection Based on Blob Analysis

Abstract

Abandoned object detection is vital & necessary in many video surveillance context. In this paper a
new framework to efficiently detect abandoned object in surveillance videos based on background
subtraction & foreground analysis is presented. The blob analysis technique is applied on the foreground object
pixels represented by binary images. Blob Analysis calculate statistics for labeled regions in a binary image. The
statistical quantities such as the area, centroid, bounding box, label matrix, and blob count are returned by the
block. On the binary images of the foreground pixels the morphological operations are done. From the blob
statistics, the blobs are tracked to find for how many frame it remains stationary. If the blob stays stationary for
predefined number of frames, it is declared as an abandoned object & the alarm is raised. Very small abandoned
objects in the low quality surveillance videos can be detected. Proposed method can even detect the abandoned
objects in the presence of varying illuminations and dynamic background.