Research On The Application Of Pavement Traffic Signs Recognition Based On Machine Vision

This paper aims to study the traffic sign recognition, including direction indicative markings and lane markings, and provide the drivers the necessary traffic information to ensure their safety.Firstly, this paper analyzes the drivers’ perception characteristics, finds out the general time required as the driver addresses the traffic condition, and thus we come up with the curve of perception characteristics. Meanwhile, the experimental survey on traffic behaviors of drivers by different ages and genders is conducted, and the conclusion is that traffic sign recognition is critical to the driving auxiliary function.Traffic sign recognition system mainly consists of three parts: image acquisition, image detection and image recognition. In order to diminish the various disruption and noise during the image acquisition, this paper has already pre-processed the image, including graying, image filtering and enhancement. Among these, it adopts the median filtering to de- noising the images, and the effect is ideal.To meet the requirement of timeliness and accuracy, the image shall be simplified and the information shall maximally be reduced. Given that, this paper applies binary processing on the images, and the Minimum error method achieves the optimal effect via the experimental comparison. With the analysis of direction indicative markings’ shape feature, it extracts seven types of invariant moments such as direct movement, left turn and right turn as the basis of recognition. During the extraction, this paper respectively introduces and applies the Hu moments and Zernike moments to provide the reliable evidence to identify the traffic signs.As to lane marking recognition, Steerable Filter method is adopted in this paper. It is an edge filter that controls the direction of filtering, and the good detection effect can be obtained with the different light intensity. And it also boasts the good anti-interference, and applies geometric calculation to determine whether the vehicle derivates the lane. With the judgment of the drivers’ lane-changing intention, it can reduce the misjudgment over the lane departure alarm. This paper uses Viola-Jones (VJ) method to conduct the face-centric abscissa detection, and then track the face with Kanade-Lucas-Tomasi (KLT) method. It recognizes the driver’s intention of lane-changing by the movement of face abscissa.Finally, this paper conducts the experimental verification on direction indicative markings through the BP neural network recognition system. Such recognition system can meet the requirements of timeliness and accuracy. To identify the driver’s lane-changing intention by displacement of face abscissa, and combine the experiment of departure from the lane marking, it can design a departure early warning system with the recognition of driver’s lane-changing intention.