Figure 3
From: Modeling and Visualization of Human Activities for Multicamera Networks
Use of geometry in multiview detection: (a) snapshot from each view, (b) object free background image, (c) background subtraction results, (d) synthetically generated top view of the ground plane. The bottom point (feet) of each blob is mapped to the ground plane using the image-plane to ground-plane homography. Each color represents a blob detected in a different camera view. Points of different colors very close together on the ground plane probably correspond to the same subject seen via different camera views.Video frames captured from 4 different viewsBackground images corresponding to each viewBackground subtraction results at each viewProjection of detected points onto synthetic top view of ground-plane