Research Of Traffic Signs Recognition Based On Driving Video And Its Android App Development

Intelligent driving is one of the hot topics in recent years.Before the realizations of a intelligent driving system,several key techniques,such as Internet of Vehicles,object detection and recognition,obstacle avoidance and route planning,must be solved.In this thesis,Machine learning and Convolutional Neural Network are used to study the traffic sign detection and recognition for intelligent driving.The main contents are given as follows:1.Histogram of Gradien and Support Vector Machine are used to realize the traffic signs regnition system and the corresponding Android APP is written.In this system,the extracted video is preprocessed firstly.The edge detection operator was used to detect each frame with Canny edge detection operator secondly.The contours that were not satisfied with the conditions of length width ratio and area ratio were used to cut the remaining outline,and then the image was encoded with the character of HOG,and then the image was coded after the segmentation.The first class SVM classifier is classified into two kinds,and the traffic sign area is identified.The remaining regions of interest are reused with the HOG feature description word to be coded.then,the SVM multiple classifier is used to identify the traffic signs.Finally,the effect of automatic identification of traffic signs is achieved.2.In order to improve the accuracy,Region Proposal Networks and Cascaded Convolutional Neural Network is used to implement the traffic signs and detection and regnition system,in which region proposal networks and the cascade Convolutional Neural Network are respectively adoped for the object detection and the object recognition.Experimental results show results show that compared with the method based on HOG and the cascade support vector machine,the deep learning system has higher accuracy.