Research Of Road Recognition Algorithm Based On Texture Feature And Vanishing Point

The vision-based intelligent navigation system which is a challenging subject inthe field of computer vision can be widely used in national defense and economy. Asone of the key technologies for intelligent navigation system, the road detectionalgorithms are the key problem and many scholars from home and abroad havestudied it.Because of high robustness in many complex road environments, the roaddetection algorithm based on vanishing point estimation and texture feature, hasbecome the focus among various research institutions. However, there are still someproblems about the high complexity and low accuracy of vanishing point location.Although some scholars have improved accuracy by introduction of a new votingmechanism, the computation of the method would be higher. For this kind ofalgorithms, there are problems about low accuracy on vanishing point location andhigh complexity in searching road borders. In order to address these problems, a roaddetection algorithm which fuse multiple information and based on vanishing pointestimation, would be proposed. The major work of this paper is as follows.In order to address the problem about high complexity and low accuracy inprocess of vanishing point estimation, we have analyzed the existing method based onlocal soft-voting mechanism and summed up formation of the problems. According tothe principle of the vanishing point formation, a multi-dimensional estimation methodbased on edge detection has been provided. At first, the image would be preprocessedwith canny edge detection operator. Then, a definition about the main road directionwhich generated by statistical analysis, has been used to filter the effective votingpoints. Based on the above work, a principal direction of multi-dimensional votingscheme has been proposed. For global perspective, the proposed algorithm hasdescribed the detailed process about extraction of the effective pixels and theperformance about computational complexity and accuracy have been improvedsignificantly.Because of the wide range in road boundary searching, the computationalcomplexity is too high. In order to address the problem, the analysis on the meaningof the voting vector from initial vanishing point would be given. Based on theinformation in distribution of vote numbers, a multi-information fused algorithm can be presented. The method have made the high-score direction as the first searchingorientation and reduced the number of searching directions. Therefore, the searchingscope can be narrowed effectively. In addition, the way fused texture, color and otherfeatures have provided a reliable guarantee for extracting road boundaries quickly andaccurately.In order to verify the above algorithm, quantitative analysis and subjectivejudgment have been used on vanishing point estimation and road border searching.Compared with other typical algorithms, the experimental results show that thepresented method in this paper can get high performance on real-time, accuracy androbustness and also has a certain theoretical value and practical value.