Research On Road Detection Technology Of Vision-based Navigation For Mobile Robot

Mobile robot have many valuable attributes that can benefit human beings in thefields of modern life. With the development of image processing technology and theenhancement of the ability of computers,vision-based mobile robot navigation systembecomes an important topic recently in the realm of robot intelligentize. The roaddetection is one of the key technologies in vision-based navigation,which have theimportant research meaning and use value in improve the mobile robot vision-basednavigation skills and intelligent level.Because the mobile robot working condition and road environment are complexand changeable, bring a lot of difficulties to the road detection algorithm design, howto enhance robustness of the algorithm under the interference of externalenvironmental factors, how to reduce the detection error due to the mobile robot’s ownmovement, and improve the real-time performance are road detection algorithmcurrently in need of improvement. In this article, monocular vision system is used torealize vision-based navigation for mobile robot by identifying road leading lines torealize mobile robot autonomous movement. Some key technology such as cameracalibration, image preprocessing, image grayscale, road edge and line detection arestudied to achieve a structured road environment of mobile robot visual navigationsystem. The content of the paper include:First of all, a modified camera calibration algorithm is proposed with the multipleregression model based on spline transformation, to slove the contradictions betweenfinite samples and high precision.Secondly, by analysing and comparing some relevant image preprocessingalgorithms, choose a kind of image preprocessing algorithm based on multi-color space,in the HSI model by histogram equalization for road image enhancement, in RGBmodel through the rapid median filtering smooth the road image.Then, a road identifying method that combine color and edge information isproposed, first the feature extraction criterion for the color of road leading lines is putforward and based on the method graying the road images, and then optimize andimprove the traditional edge and lines identifying algorithms, proposed a edge detection algorithm based on an adaptive Canny operator and a road leading linesdetection algorithm based on a modified Hough transform, achieve the purpose ofaccurater and faster for detecting road leading lines.Finally, in the mobile robot navigation system platform, the visualizationdevelopment tools Visual C++is selected as digital image processing software, throughexperiments and results analysis are verified for the algorithms proposed and used inthe article, and successfully used in the vision-based navigation for mobile robot.The experiment and application result shows that this research can obtained thedesired effect.