A Research on Vehicle Identification based on Deep Learning

In order to confirm the domestic license plate damage, fake brand car and clone car, the vehicle specific information mainly rely on artificial experience for recognition, as well as the vehicle manufacturers to produce more models not effectively identify a problem on issues such as the introduction of models, deep learning theory, puts forward a specific vehicle sub deep learning does not depend on the license plate class identification method based on. Compared with the traditional manual method, this method has the characteristics of automatic learning and can be used in the structure of the convolution neural network can be directly input image and so on. Based on Mxnet depth learning architecture, using InecptionAB-Net model for feature extraction, using Softmax classifier for vehicle classification. The results of experiment showed that the InecptionAB-Net model has a high recognition of vehicle sub class accuracy, fast training speed, has good robustness to illumination changes in the larger image, on different occasions, weather conditions of the vehicle image effectively vehicle recognition.