Wavelet Based Algorithms for Tracking Moving Object By Mobile Robot
In this thesis, we present various algorithms to track moving objects by a mobile robot. Our algorithms based on hypothesis that the object areas are not defined for the robot. To this end, we have desigened a new wavelet based algorithm and a color texture based algorithm for real time object tracking in a video sequence. This purpose have been achived using a special search method for finding best match among feature vector of object and feature vector of the search window. Finally suitable feature for the object have been formed. This feature is dispersion of detail coefficients in the wavelet based algorithm. This featue alonge with the size invariant property, makes it possible to successfully track a object in the images.We use Euclidean distance measure for finding the distance of two feature vectorThe algorithm has been implemented in a PC running widows XP operating system with Pentium IV 3GHz full cash processor and 1GB of RAM memory. considering 240 320 pixels search window. The results have shown that our algorithms is often superior to traditional template methods for object tracking applications.