Research On License Plate Recognition In Low Illumination Environment

As one of key technologies in intelligent transportation,license plate recognition technology has been widely used in vehicle intelligent parking lot management,highway toll management,and vehicle security,etc.The license plate's edge and color information is not obvious in low illumination environment and it is difficult to detect and recognition plate by using traditional methods.According to the problem,a new recognition method in low illumination environment is proposed in this thesis.The main work is as follows:In the plate location module,a license plate detection method based on improved maximally stable extremal regions is proposed in this thesis.Firstly,character candidate regions are extracted by the improved maximally stable extremal regions detection algorithm.Then character candidate regions can be found according to the geometric characteristics of the character.Next a text classifier is used for further filtering,and strong seeds are selected according to the classification probability,weak seeds can be found around strong seeds according to the seed growth method.If the total number of seeds is less than seven,a sliding window is added to find missing characters.Finally,the real license plate area can be found through the support vector machine classifier.In the character segmentation module,an improved vertical projection method is proposed in this thesis to realize character segmentation.Firstly,the improved maximum interclass variance method is used for binarization operation.Then the border and the rivet need to be removed,and the improved vertical projection method is used for roughly division operation.Finally,the precise division is completed based on the prior knowledge of the character.In the character recognition module,the self-adaptive evolutionary extreme learning machine is used to train the model and recognize characters.In this method,the self-adaptive differential evolution algorithm is used to optimize the input weight and hidden layer deviation of the extreme learning machine network.And the improved local binary pattern features are used to train the final classifier.The experimental results show that the algorithm proposed in this thesis can be used to detect and identify the license plate in low illumination environment.It is satisfied for the real-time requirement and can also be adapted to other complex environments.The algorithm also provides a new thinking for license plate recognition.