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

Mobile robot is an important branch of robotics and will be widely applied in both military field and civil area. The studying relates to theory and technique of multi-subjects, incarnates the up-to-the-minute fruits of the information science and artificial capacity technique. The vision-based navigation technique of mobile robot is the important base and key technique of the mobile robot intelligent research, and it is becoming an increasing recognition of scholars home and abroad, has become the forward problem and studying hotspot of the artificial capacity and robot researching. Road understanding technique of vision navigation is the basic module to study the vision navigation of mobile robot, and the understanding ability touch the stability of the mobile robot self- navigation system straightway. This paper takes the vision-based navigation road understanding technique of mobile robot as the research object, and the vision-based navigation boundary extracting as the masterstroke to carry through correlative research of color features extracting, texture features extracting, features amalgamation and road image segmentation method.In the research on extracting color features of the road image, aim at the area segmented by the histogram sort quantification method based on the single color features maybe not intact, and the self-adaptability of selecting sort threshold value is not good, a color features extracting method based the HSI color space is presented. This method adopts the HSI color model which accord with human vision character and the component is specialty to each other, so it makes the best of the description stability of HSI color model for colors, combining with adaptability of quantification based on multi-thresholding classification of histogram, and extracts color features of image. The experiments indicated the validity of the method of extracting color feathers presented in this paper. In the research on extracting texture feathers of road image, a texture feathers extracting method based on the discrete wavelet frame (DWF) transform is presented. When describe the image texture frame, comparing with the pyramidal wavelet transform, the method this paper presented can hold the translational invariance of the image texture feathers primely in the decompose process, and offers texture feathers value which opposite the former image pixels one by one for final image clustering segmentation; in the process of wavelet frame transform, we use the gray-level co-occurrence matrix to select decompose layers, and offers a intuitionistic and feasible approach for selecting the decompose layers based on the degree of the texture component of low frequency approximate Sub-image filtrated; On the base of the multilayer filtering of the DWF for image, a improved method of Laws texture energy measuring is presented, calculating the wavelet coefficient of the decomposed image, and combining level, upright and diagonal high frequency coefficient which is decomposed on each layer, so the feathers parameters are changeless when circumrotate in the directions of 90°,180°and 270°, thereby we gained the texture eigenvector of the image pixels. Texture feathers extracting and segmentation experiments validate the validity and accuracy of the method presented in this paper.On the base of the research of color and texture feather extraction, a road understanding arithmetic based on combining the road color and texture feathers is presented. This arithmetic makes use of vision image disposing technique to combine the gained road color and texture feathers according to stated power value, and finally acquires the eigenvalue which can express the feathers of each area in road image, then use clustering technique to measure off the road areas and non-road areas in road image, then with the edge detection method to provide the navigation path boundary information for the mobile robot. Chromatic texture image segmentation experiments and real-time road image segmentation experiments of the mobile robot validated the good effect of the road understanding arithmetic in the aspect of guiding mobile robot to explore environment autonomously. According to the road scene presented in this paper to under the arithmetic disposing flow, and in the process of concretely realizing, by dint of MATLAB and Visual C++ to compile some important image disposing functions, realize the segmentation of the road image and the availed detection of the road area. Under the navigation and control platform of Pioneer31 robot, the visual navigation and control experiments are completed, which validated the feasibility of this arithmetic in the mobile robot autonomous navigation movement.