Research On Lane Line Detection And Traffic Sign Recognition Of Outdoor Intelligent Vehicle Based On Machine Vision

With the rapid development of science and technology,intelligent vehicles play an important role in industrial production,and outdoor intelligent vehicle products are gradually upgraded to drive social development.This paper mainly focuses on the closed and semi-closed public space environment,and studies lane line detection and traffic sign recognition based on intelligent vehicles,so that it can complete the work in a specific environment.The main research work is as follows:(1)Based on the improved AHP and three-scale method,a multi-factor hierarchical structure model is established,and the optimal transfer matrix is introduced to optimize the calculation process of the whole method.Through quantitative evaluation research,the performance weights of outdoor smart vehicles are analyzed.It provides a research basis for the follow-up intelligent vehicle road information collection and analysis.(2)For the complex working environment of outdoor intelligent vehicles,this paper carries out road image preprocessing.First,the collected images are processed such as region of interest extraction,grayscale transformation,image filtering and image enhancement.A bilateral filtering method that can not only reduce noise,but also preserve image edge information.Based on the edge detection algorithm of Canny operator,an improved dual-threshold Canny operator is presented to detect the edge of lane lines.For straight lanes with disturbing signs,shadows,nighttime and low light intensity,the cumulative probability Hough transform algorithm and the piecewise Hough transform algorithm are used to detect the lane lines of the straight lane model and the curved lane model,respectively.Finally,the least squares method is used to fit the detected line segments.For straight lanes with interference signs,shadows,nighttime and low light intensity,experiments show that the cumulative probability Hough transform algorithm has higher accuracy and better real-time performance than the traditional Hough transform algorithm.On the one hand,the piecewise Hough transform algorithm can also well complete the curve lane detection.(3)Experimental research is mainly conducted on traffic sign detection algorithm based on color space model.Through literature and experiments,an improved threshold segmentation method based on HSV color space segmentation method is presented.The RGB color space of the image is transformed into HSV color space through color space conversion,and then the traffic sign detection is completed by adaptive threshold segmentation of H,S and V components.This method not only effectively realizes the segmentation of traffic signs,but also removes more interference noises.Finally,The improved threshold segmentation method based on the HSV color space is used to compare with the segmentation method based on the RGB color space model and the HSI color space model.The experimental results show that the method can effectively remove the background interference and greatly improve the detection speed and accuracy.(4)By studying the traditional template matching recognition algorithm,two improved template matching methods are proposed.Namely the center invariant moment template matching method and the correlation coefficient and HU invariant moment method.The central moment invariant template matching method calculates the central moment feature vector of the template image and the feature vector of the test image,and uses the similarity coefficient function to compare the similarity between the two feature vectors.By calculating the relatively simple template image feature vector,it overcomes the disadvantage that the traditional template matching method can only reduce the recognition accuracy in order to reduce the calculation amount.The traffic sign recognition principle of correlation coefficient and HU invariant moment is to combine the nonlinear combination of third-order normalized central moments to form seven invariants with translation invariance,rotation invariance and scaling invariance according to the principle of invariant moment.Aiming at the problem that the recognition rate is reduced due to the rotation or scaling of the image captured by the vehicle camera in this method,the invariant moment formula is not affected by the image rotation and the scale factor is reduced by using the ratio between the invariant moments.The effect of scaling changes.Based on this idea,10 invariant moments are expanded,and the scale factor is eliminated by the ratio between these 10 invariant moments to maintain the invariance of the invariant moments,thereby improving the recognition accuracy.