Traffic Sign Recognition And HMI Design In Intelligent Transportation System

ITS plays a more and more important role in our daily life. As an important part of ITS, the electronic police plays an irreplaceable role in maintaining social order and deterring illegal crime. As important parts of electronic police, TSR and human-computer interface have important research value. Efficient and reliable TSR system can improve the safety of driving. Human-computer interface is the medium of transmitting and exchanging information between man and computer. This thesis mainly studied TSR and human-computer interface of ITS, and the main content is summarized as follows:1. Traffic sign recognition based on the geometric moment invariants and the SVM classifier. This paper makes a study on a common TSR method under the complex traffic environment background, based on the geometric moment invariants and SVM. Upon this basis, the SVM algorithm is deeply probed, and the influence of employing different kernel and parameters before and after for optimum on the performance of the algorithm to recognize is further discussed. The thesis offers an effective TSR based on the geometric moment invariants. Aiming at fuzzy problems at the early stage of the image processing, such as image blur under greasy weather, the thesis improves the quality of image based on the dark grey priori algorithm---firstly employs the knowledge of the dark channel priori to estimate the atmosphere light and transitivity, and then uses guided filtering to revise the transitivity; At last, recover the image that owns a relatively- good quality according to the degradation model. The experimental results show that the TSR based on geometric moment invariants and the SVM has poor robustness to fog, but the algorithm proposed by the thesis is not affected by this and shares an obvious promotion in recognition performance.2. Traffic sign recognition based on the template matching. The template matching for TSR is adopted in the thesis, because the correct recognition rate of traffic signs based on the geometric moment invariants and the SVM classifier has much room for improvement. Considering the certain robustness possessed by the algorithm, such as affection-excluded by image scaling, rotation, and translation, this paper has adopted the SIFT algorithm and SURF algorithm,which extract the feature points from the traffic signs and generate corresponding descriptors. Then when the descriptors are matched, the recognized results are output according to the correct matched feature points. The experimental results show that this method is more effective. In view of the real scenario, the actual shooting angle may change and the object shot may be distorted, which makes it difficult to recognize, the thesis has adopted the ASIFT algorithm. In the end, the thesis will focus on the effectiveness of the above three algorithms from two aspects---the normalized size and the affine deformation.3. A HMI of electronic police is developed. Via the interaction, people can manage multiple electronic police system synthetically, configure the parameters of each electronic police, display and store the output of electronic police and so on. On the basis of needs of the project and a thorough investigation, this thesis makes an overall plan on human-computer interface of electronic police and realized functions, such as displaying the real-time traffic scene, displaying the results of the snapped vehicle and the recognized plate, setting the traffic reticule and detecting box, counting the traffic information, setting the parameter of smart camera and so on.