Publications

Abstract : Registration plate recognition plays a vital role in numerous numbers of applications in today’s world. Researches done so far could not completely meet the real time requirements of the system like different color, shape and size of plates, varied illumination conditions at image capture time, different positions and alignment of the plate. In the proposed system we introduce dynamic image processing techniques to meet the above mentioned short comings. We also introduce a new approach using genetic algorithm to find the location of the plate. Fluctuating illumination conditions are taken care-off by adaptive threshold method. Connected component labelling is used to identify the objects in blind folded regions. A matrix of invariant scale geometry is used for better system adaptability when applied to different plates. The convergence of the genetic algorithm is greatly improved by the introduction of a newly created mutation and crossover operators. We also modify genetic algorithm to overcome the drawbacks of connected component method by importing partial matching of the characters.