Research On Traffic Indicatory Signs Recognition Of The Vision-based Intelligent Micro-vehicles

In the micro-traffic environment,the intelligent driving automatic driving simulation method based on visual-auditory information can solve the problem that the smart car experiment covers a large area and the vehicle modification cost is high under the real traffic environment,and it also can solve the problem that software simulation experiment difficult to accurately simulate the real traffic environment.This thesis studies the identification of traffic signs in the micro-traffic environment.Firstly,the design mode of hierarchical platform processing is adopted to separate the micro-car control and data processing,which improves the accuracy and real-time of the control.The upper computer collects and analyzes the traffic sign information and transmits it to the lower computer.The lower computer controls the steering and travel of the vehicle body according to the control signal.In the control system research,the incremental PID and positional PID algorithms are used to control the speed and steering respectively.Secondly,in the research of indicatory sign detection and recognition,a method based on color and shape features is adopted.In the color segmentation stage,Histogram equalization is used to improve the image contrast,and then three color segmentation algorithms are compared and analyzed to select the best HSI threshold segmentation algorithm.In the shape segmentation stage,the obtained binary image is preprocessed(including filtering,morphological processing and filling)firstly,and then the processed image contour is detected,and the noise is initially removed by limiting the aspect ratio and the pixel area.The non-interest region is removed again by the invalid pixel area constrained algorithm,and finally use the circularity algorithm to obtain the final region of interest.This thesis proposes an improved MBLBP+HOG+PCA feature extraction algorithm,and further verifies its superiority in recognition rate,false recognition rate,training time and detection time.Then the indicatory signs are identified by using the One-Against-One SVM multi-classifier,finalize the accurate information of the indicatory signs.Finally,the indicatory sign recognition experiments of the intelligent micro vehicles are completed on the micro-traffic environment.The experimental results show that the intelligent micro vehicles can accurately identify the category of the indicatory sign and perform the corresponding action in the micro-traffic environment.