Research On Vehicle Logo Recognition Method And Software Implementation

As one of the emerging technologies in intelligent transportation system, vehicle logo recognition has attracted much attention and interest of scholars in related research areas recently because of its practicle value in deck car detection and vehicle tracking. We analyzed and improved the existing vehicle logo recognition and detection method in this paper.Besides we also designed the vehicle logo recognition system,which has a good algorithm performance and user-friendly interface,and implemented it by programming.The main contents of this paper are as follows:(1)Research of vehicle logo detection algorithm.To eliminate the impact of the vehicle leaning to the logo detection module, this paper firstly proposed a stable and sophisticated symmetry detection and tilt correction method based on SIFT descriptors.After that we modified the Adaboost lisence plate detection method by combining its characteristics and the detection performance is somewhat improved.At last we put forward two vehicle logo fine detection method based on edge detection and Gabor filter respectively and compared the strengths and weaknesses of the two methods through experiment.(2) Research of vehicle logo recognition algorithm.First of all, we introduced a new vehicle logo feature extracting algorithm. The proposed new texture descriptor which was consisted of the PHOG descriptor and MB-LTP descriptor was improved to be more robust and invariant to slight locating error. Meanwhile, this paper analyzed the basic principles of the support vector machine and put forward a vehicle logo classifier based on SVM.We compared it with other common classifiers in the experiment to verify its advantages.(3)System design and implementation.Based on the research of the vehicle logo detection and recognition algorithm,a vehicle logo recognition system is designed from the perspective of system function and software architecture.Then the major implementation details of the core modules of the system,including the vehicle logo detection module and the vehicle logo recognition module, are presented in this paper.Finally functional test and performance test were conducted to demonstrate the performance of the system.