The Research Of Vehicle Recognition Method Based On The Images From Vehicle Intelligent Monitoring And Recording System

As an important part of Intelligent Transportation System, vehicle type recognition is widely used in stolen car crime solving, traffic order regulating, and highway automatic charging. A great amount of vehicle exists, some of which share the similar structure or identical logo. Influence by complex background (such as light and weather) and some other reasons (such as no vehicle plate or stained), it is difficult to obtain the exact visual information of vehicle. Thus, vehicle type recognition has become the research focus and difficulty in computer vision area.Based on Anhui provincial science and technology key project "Research on intelligent vehicle automatic recognition system based on video images", the research of vehicle type recognition is carried out. First, vehicle face is located. Then base on feature extraction and recognition technique, through vehicle logo and face recognition, the vehicle type is identified. The work of this dissertation includes:1) The location of vehicle face: Based on the knowledge of traditional vehicle plate location methods, a multi-feature based vehicle plate location method is proposed. It locates the vehicle plate through color and texture features. Then based on edge detection and vehicle plate location, the vehicle face can be located.2) Vehicle type recognition base on vehicle logo and face:Using the common feature extraction methods to extract the vehicle face features, and then using classifier to recognize vehicle type, the accuracy of which can’t meet the requirements. Thus, a vehicle face feature extraction method through convolutional neural networks is proposed for vehicle type recognition, in which the feature can be learned through self-learning. Consider the complexity of vehicle type, the vehicle logo is recognized in order to assist the identification of vehicle brand, combined with face feature model, the vehicle type recognition is finally realized.3) The realization of vehicle type recognition system: Based on the techniques proposed in the above chapters, the vehicle type recognition system based on VIMRS images is developed. The detection results of the system is evaluated and analyzed based on the test datasets.