Research On AdaBoost Optimization And Its Applications In License Plate Location

License Plate Recognition(LPR) is a technology that by using computer vision,image processing and pattern recognition techniques to extract the information of the license plate character from the image,in order to determine the identity of the vehicle. A complete license plate recognition system includes three parts,like license plate location,characters segmentation and characters recognition.Existing license plate location method usually uses the image pixels to locate,this not only takes a long time,but the location effect largely depends on the quality of the image.For the problem of the low-contrast license plate is difficult to locate,we start from both the feature extraction and classifier training,the Haar-like features,AdaBoost algorithm are introduced to the license plate location technology,and the algorithm has been improved.Experiments show that the proposed algorithm not only improves the license plate location accuracy but also shorten the location time.Haar features have been largely used for pattern recognition areas, such as face image analysis,and the license plate images match with the description of the Haar features. So this article introduces Haar features from the ares of the face image analysis to the license plate image analysis,and the structural properties of Haar features are reserved and redefine the eigenvalue of Haar feature,in order to construct a Haar-like features to adapt to the needs of the license plate image analysis.For traditional AdaBoost algorithm exists over-fitting in training classifier,so this this article has improved the algorithm, proposes an adaptive filtering algorithm based on Hoeffding inequality,which can divide training samples into reliable samples and temporarily unreliable samples.And proposes a dynamic weight update rule,so as to achieve the effective control of the structure of the classifier.The experiments show that the improved algorithm not only has good generalization performance,but also has good robustness.