Delta E colour feature based mutiobject tracking using merge and split approach and Kalman filtering

Automatic monitoring of the region using single stationary camera had become a vital part of the surveillance. The tracking of moving objects can undergo limitations such as illumination variation, camera noise, occlusion. Many methods exist to overcome this limitation. In the multiple moving object system detection is done by using Mixture of Gaussian based on background subtraction to get foreground mask. In tracking the occlusion is handled by merge and split technique, feature correspondence is done using Delta E based color feature of the multiple moving objects. Delta E gives single value for the color difference between the objects. Delta E can extract even minute difference in the scene which is not perceivable to human eye. The prediction for the next position of the object is done by using Kalman filter. Experiment on PETS 2009 dataset video show that the proposed system gives more accurate results than the existing system which uses color histogram based matching.