Researches On Automatic Location Of Car License Plate

The automatic recognition of car license plate is an important step in Intelligent Transportation System. It can be applied to paying fee without car stop, finding stolen cars, managing the cars in parking lots, etc. The location of license plate is the most pivotal technology in the recognition system. It plays an important role in the performance of the whole system.This paper focuses on the automatic license plate locating in the recognition system. Based on a thorough analysis of the present algorithms, several automatic license plate locating algorithms are proposed. Here are the main contributions of this paper:1. An automatic car license plate locating algorithm based on the texture feature of characters is proposed.This algorithm is based on the texture feature of the characters on a license plate. Firstly, the long horizontal lines with small change in gray level in a car image are removed. Secondly, small areas are got rid of, then an analysis of the geometric features of the connected areas and the number of inner pixels and pixels on its edge is made, and with the help of the geometric features and the projected features of car license plate, the license plate area is found out.2. An automatic car license plate locating algorithm using ELMAN neural network is presented.Using the different gradients of the plate area and the non-plate areas in a car image, a neural network is trained. Then the image is classified with the trained network. Finally, with the help of the projection features and the geometric features of car license plate, the license plate is located.The license plate samples, which are taken under various conditions, such as illumination, weather, license plate type, etc., are used to prove the efficiency of the two proposed algorithms. The results indicate that the rate of correct location of the two algorithms reach 99%. These experimental results show that the two algorithms are effective and practical.