License Plate Recognition Algorithm Combining With Artifical Bee Colony

With the development of economy, science and technology, our living standards are improved and transportation is also developing. Since the car’s number is increasing, the requirement of transportation is higher and higher. The Intelligent Transport Systems came into being. License Plate Recognition (LPR) is a very important part of the system, consist by License plate location, License plate segmentation and License plate recognition.Artificial bee colony algorithm is a new method to find the optimal solution, which is proposed by Karaboga. The algorithm applied the bee colony model to the actual optimal problem. The algorithm is fast and easily implemented.In order to improve the robust of license plate recognition algorithm, this paper applied artificial bee colony algorithm to the license plate location, license plate segmentation and character recognition.First, at the edge detection part, it is difficult to determine the threshold. A bad threshold may lose edges or produce false edges. This paper uses artificial bee colony algorithm to search the optimal threshold automatically, and then we will obtain a good edge image. Then combined morphological processing,pixel change,and the license plate’s color texture feature to find the license plate’s location.Second, at the license plate segmentation part, the binarization of the license plate image is very important in vertical projection. While the binarization is mainly to find a suitable threshold, this paper will use artificial bee colony algorithm to search the optimal threshold automatically. With the optimal threshold,we can get a good license plate binarization image, and then use the vertical projection to divide the license plate character.Finally, the dissertation applied Back Propagation Network(BP network) to the part of recognition of the license plate character. While it is very important to find approprate weights and bias for BP network, the dissertation use artificial bee colony algorithm to search the initial weights and bias. At last we use the BP network to identify the license plate character.Simulation results show that the algorithm can identify license plate well.