Dissertation > Excellent graduate degree dissertation topics show

Research on Road Understanding Technology of Monocular Vision-based Navigation for Mobile Robot

Author: LiZuo
Tutor: ZhangMingJun
School: Harbin Engineering University
Course: Mechanical and Electronic Engineering
Keywords: Visual navigation Road Understanding Image Segmentation Color Feature Extraction Texture feature extraction
CLC: TP242
Type: Master's thesis
Year: 2009
Downloads: 211
Quote: 1
Read: Download Dissertation

Abstract

The mobile robot visual navigation technology is an important foundation and key technologies of the mobile robot intelligent research, is being more and more attention of research scholars at home and abroad, and has become the forefront of artificial intelligence and robotics research topics and research focus. Visual navigation of roads understanding of the technology is a basic module, its ability to understand a direct impact on the performance of the mobile robot autonomous navigation system of mobile robot visual navigation technology. The visual navigation road understanding of technology to improve mobile robot visual navigation capability and intelligent level has important significance and value in use. Understanding of technology as a research object mobile robot visual navigation road mobile robot visual navigation boundary extraction, the main line of the the road image area color features extracted texture feature extraction, feature fusion and the road image segmentation method specific issues related to research job. The road image color feature extraction studies, quantitative method for histogram classification based solely on the color segmentation area may not be complete, as well as the classification threshold selection and poor self-adaptive problem, this paper proposes a color characteristics based on the HSI color space extraction method. The method uses in line with the characteristics of human vision and the various components of independent HSI color model to represent color images, to take full advantage of HSI color model described the color stability, suppress the divergence of the color information in the classification, combined with histogram-based multi-threshold classification The quantized adaptive, to extract the color characteristics of the image. Verified through experiments on the effectiveness of the color feature extraction method proposed in this paper. Road image texture features extracted texture features undermine the traditional wavelet transform translational invariance problem, this paper proposes a framework transform based on DWF discrete wavelet texture feature extraction methods. Image texture description, the method proposed in this paper pyramid wavelet transform method can be very good in the decomposition process to retain the translational invariance of the image texture features, and for the final image clustering segmentation one-to-one correspondence with the original image pixel value of texture features; gray level co-occurrence matrix decomposition level selection, based on the low-frequency approximation sub-image texture ingredients are filtered degree of judgment in the process of decomposition of the wavelet frame transform to provide an intuitive and practical way for the decomposition level selection; DWF on the basis of the image multilayer filtering the improved method of the Laws texture energy measurement, decomposition calculation sub-picture of the wavelet coefficients, and each layer decomposition level , vertical and diagonal high-frequency coefficient combined, at the same time to reduce the feature dimension, but also makes the characteristic parameters have a rotation-invariant nature of the rotation of 90 °, 180 ° and 270 °, thereby obtaining the texture feature vector of the pixel of the image. Texture feature extraction and segmentation experimental results verify the effectiveness of the proposed method. On the basis of color feature extraction and texture feature extraction studies, understand the algorithm based on the fusion of the road color and texture features road. The algorithm uses the visual image processing technology, and will get the the road colors eigenvalue texture features value according to the combination of weight, eventually able to characterize the regional road image characteristics, based on the eigenvalues ??clustering technique to complete the road image in road regional division of the region and non-road mobile robot motion path boundary information is given, and edge detection method. The road color texture image segmentation experiments and the mobile robot real-time the road image segmentation experimental results verify proposed in this paper to understand the algorithm shows good results in guided mobile robots independently explore environmental. In the concrete realization of the process, the preparation of a number of important image processing functions, with tools such as MATLAB and Visual C achieve the road image segmentation and road area detection algorithm processing flow, in accordance with the road scene understanding. Mobile robot visual navigation control experiments were carried out in Pioneer31 robot navigation and control platform, experiments verify the feasibility of the proposed algorithm in autonomous navigation of mobile robots movement.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Automation technology and equipment > Robotics > Robot
© 2012 www.DissertationTopic.Net Mobile