碩士 / 國立臺北科技大學 / 機電整合研究所 / 90 / Due to vast variability of the brightness in outdoor environment, the roads, trees, and lawns usually are not easy to recognize. In this study, we have developed a new navigating approach based on binocular stereovision that has successfully applied to an autonomous land vehicle (ALV) guidance and collision avoidance in outdoor environment. The study topics in the thesis include the information of road learning, stereo corresponding of the sensor-like points, collision avoidance and navigation.
In a known outdoor environment, a learning approach to record the information of road before navigating is useful. The information of road includes hue, saturation and intensity (HSI) color feature distribution and the range of their color gradients. This approach will not be influenced by brightness easily, because the approach does not apply the intensity property. According to the learning information, we can label the pixel in the image whose H and S values are similar to the learned information as road. Furthermore, to reduce the cycle time, the image buffer is segmented into blocks of size .
Stereo corresponding is a time-consuming task. If we perform corresponding on whole image, we can not navigate the ALV in real time. In the thesis, we propose an efficient and fast approach to sense five suitable points for corresponding, called sensor-like points approach. After sensing points, we suggest an improved approach to solve the stereo corresponding. The approach combines the advantages of feature-based matching and block matching to solve the optimal correspondence point. Using the corresponding point, we can reconstruct five 3-D points, and obtain five coordinates relative to ALV. Using those information we can navigate in the outdoor environment, and perform collision avoidance. The ALV system has been implemented, and it can be flexible to navigate in campus.
| Identifer | oai:union.ndltd.org:TW/090TIT00651037 |
| Date | January 2002 |
| Creators | 劉明豐 |
| Contributors | 駱榮欽 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | en_US |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 66 |
Page generated in 0.0114 seconds