Research And Implementation Of Feasible Road Detection Based On Binocular Vision

In recent years,the research on intelligent mobile devices for outdoor scenes has become popular.Traffic safety has been a hotspot in the field of research,and undoubtedly road detection technology is one of the emphases.The main purpose of road detection is to determine the feasible area of the current environment.And then remind the mobile device to avoid obstacles along the specific edge,so as to help it reach the designated location.At present,the difficulty of road detection technology is to perceive the road area,road route and obstacle in the road.The feature used in traditional methods mainly include: color,texture,edge and road sign.However,due to the complex and varied nature of the road environment,it is difficult to meet the various needs of the actual project.Aiming at this problem,this thesis proposes a feasible road detection method based on binocular vision.We combine the disparity information with the traditional feature,and successfully improve the accuracy of the detection results.The main contents of this thesis include:Firstly,a road area detection algorithm based on image segmentation is proposed.The image is first segmented or evenly divided,and then the features are extracted based on the segment result.By using the binocular vision,the disparity information is added to the traditional features,which can improve the accuracy of the classifier detection.The post-processing by adjacency matrix or two histograms is necessary to further improve the detection accuracy.Secondly,a road lane and obstacle detection algorithm based on road area is proposed.The region of interest is extracted by using the road area,which can improve the efficiency of the straight line detection.The road lane is classified based on the color difference,the road edge and the lane distance.We can effectively detect the road edge and the zebra crossing.The distance information of the obstacle is obtained by the disparity value,and the relative motion of the obstacle is calculated based on the inter frame information of the video.Moreover,we get the feasible road in the image by combining the road results with the obstacles results.Thirdly,we build an embedded development platform which can detect road in real time under the mobile scene.Based on DM8168 development board we make a binocular vision embedded hardware platform.The algorithm is transplanted into the embedded system.By analysing and optimizing the problem,the function of the USB binocular camera input,DSP chip algorithm processing and real-time display processing results is realized.In this thesis,the road database is established.After testing and analysis,the experiments demonstrate that this algorithm can effectively detect the feasible road in the image.Besides the algorithm is efficient,which can meet the real-time processing requirements of the embedded operating platform.