People counting using object detection and grid size estimation

Abstract

The proposed system estimates the number of people in indoor or outdoor location. A background model of the scene is constructed to detect moving objects (people) in a video stream. Detected foreground objects are preprocessed to eliminate pixel noise and small artifacts by performing opening morphology operation in the foreground image. Counting persons in the occluded foreground object is estimated by counting the human blobs by estimating the grid size. The accurately normalized and calibrated reference grid model is necessary for the grid estimation. © 2011 Springer-Verlag.