Knowledge-based Automatic Change Detection and Positioning System for Complex Heterogeneous Environments

We present a novel algorithm to achieve automatic detection and positioning of changes for monitoring systems in complex environments. The aim is to efficiently detect changes of unknown dimensions, shapes and velocity and to position them in a sequence of images. The practicality of the algorithm is simplified by the use of different decision rules in a multistage test for different purposes. These decision rules identify the changes and number of parts, as well as the position and its optimal pick-up points for each individual part. A lighting compensation method is embedded to maintain a constant lighting environment and therefore the error rate can be reduced. Experimental results on a variety of image sequences show that the proposed algorithm is effective and efficient, regardless of the irregularity and number of changes.