碩士 / 國立雲林科技大學 / 電子工程系 / 107 / In recent years, many robotic obstacle avoidance methods use high-priced laser detection technology, or a dual-lens camera as a sensor for robot control. In order to reduce costs, this thesis uses a single vision obstacle avoidance method, combined with three obstacle avoidance algorithms for obstacle detection. The techniques include top view gradient detection, continuous image detection and improved integral image segmentation algorithm. First, the gradient detection of the top view is to convert the original image into a top view using perspective conversion. The X, Y gradient is used to select the obstacle according to the direction of the obstacle using the Sobel operator. The continuous image detection is calculated using two consecutive images. In order to avoid errors when moving objects, the average of the surrounding 3x5 pixels is calculated. The improved integral image segmentation algorithm is modified to eliminate the ground marking lines and to classify the plane objects on the load. Since the single vision lacks in-depth information, it is impossible to estimate the direct estimation distance. In order not to increase the cost of robot, the original image is converted into a bird's-eye view image using a perspective transformation to detect the distance according to the pixel distance. To solve the universal single vision unmeasurable distance, the use of multiple algorithms for high correlation improves the accuracy of obstacle detection. The DBSCAN clustering method is employed for the selection of obstacles. The robot system had been implemented by ARM system, the obstacle can be detected on the real-time camera. The frame rate is about 20 fps.
In order to enable the robot to walk on outside, the Google Maps API is used. Javascript architecture calculates the direction angle and distance through the latitude and longitude, and combines the electronic compass. The single visual obstacle avoidance mode is according to the direction by Google Maps to control the motion direction for the robot with route planning on the outdoor.
| Identifer | oai:union.ndltd.org:TW/106YUNT0393059 |
| Date | January 2019 |
| Creators | Li, Xiang-Xuan, 李向璿 |
| Contributors | Shr-Chang, Shia, 夏世昌 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 113 |
Page generated in 0.0145 seconds