Research On Recognition Method Of Vehicle Logo Recognition Based On 2DLPP And 2DLDA

With the rapid increase in the number of vehicles,vehicle identification systems are increasingly important for vehicle information monitoring,vehicle violation detection,and vehicle service on the road.For certain decks,obstructing the license plate behavior,a separate license plate recognition system may cause misjudgment of vehicle information,and combining the vehicle logo recognition system may reduce the misjudgment rate.The traditional vehicle identification method will result in inaccurate positioning of the license plate when the vehicle body texture is similar to the license plate texture,and the traditional image dimension reduction method can not make good use of the spatial structure information embedded in the pixels and the neighborhood pixels,resulting in a low recognition rate of the vehicle logo.In this paper,these two problems are optimized in the two stages of the vehicle-logo location system and vehicle-logo feature extraction.For the traditional vehicle-logo identification method,if the vehicle body texture is similar to the license plate texture,it will lead to the problem of inaccurate positioning of the license plate.Based on the license plate location,this paper uses color space and texture detection combined to achieve accurate positioning of the license plate;Vehicle-logo positioning method using image Haar features and Adaboost algorithm.Experiments show that this method improves the positioning accuracy of the vehicle-logo compared with the traditional method.For the traditional image 1 dimension reduction method can not make good use of the spatial structure information embedded in the pixels and neighborhood pixels,this thesis uses 2 dimensions features to fuse each other to extract image features.The 2 dimensions feature uses the original image matrix to construct the image covariance,and extracts the feature vector of the image,which preserves the spatial structure information of the original image.Based on the combination of color space and texture detection to locate the license plate,and based on the vehicle-logo location positioning based on the combination of Haar features and Adaboost algorithm,this thesis conducts experiments on commonly used 1 dimension data dimensionality reduction methods and 2 dimensions feature fusion data dimensionality reduction methods.The experimental results show that the combination of 2DLPP and 2DLDA features has a better recognition effect.