Publication: Autonomous navigation using stereo vision
| datacite.subject.fos | oecd::Engineering and technology::Electrical engineering, Electronic engineering, Information engineering | |
| dc.contributor.author | Lok, Sook Wan | |
| dc.date.accessioned | 2024-07-04T01:35:27Z | |
| dc.date.available | 2024-07-04T01:35:27Z | |
| dc.date.issued | 2010-04-01 | |
| dc.description.abstract | Obstacle detection is the main issue for safe navigation and it can be achieved by applying stereo vision. In stereo vision, two cameras placed horizontally from one another on the mobile robot platform. These cameras are used to obtain differing views on a scene, in a manner similar to human binocular vision. The first step is camera calibration by using Matlab Camera Calibration Toolbox. The camera calibration process estimates camera’s intrinsic parameters and calculates extrinsic parameters of the camera. The result of the calibration process is used for image rectification to transform the corresponding epipolar lines in all images so that they become collinear with each other. Rectification is able to eliminate errors caused by lens distortion and camera displacement. This greatly simplified the implementation of the stereo matching algorithms. Stereo matching algorithm is used to find the corresponding points in stereo image and produce disparity map, which is inversely proportional to object distance. In this project, using a predefined window size, two different stereo matching algorithms, Sum of Absolute Differences (SAD) and Sum of Squared Differences (SSD) are implemented and their performance is compared. A pair of points is considered to match if the SAD or SSD value is small enough and minimum among all values computed within the possible disparity range. The disparity map generated is analyzed and processed to find the distance between the objects. The performance of SAD and SSD are quite similar. However, the processing time of SSD is longer than that of SAD because it requires an extra step in the squared calculation. As a conclusion, by using stereo matching algorithms to extract 3D depth information, the mobile robot is able to distinguish obstacles from the free space ahead. It is able to move forward by avoiding obstacles. | |
| dc.identifier.uri | https://erepo.usm.my/handle/123456789/19593 | |
| dc.language.iso | en | |
| dc.title | Autonomous navigation using stereo vision | |
| dc.type | Resource Types::text::report | |
| dspace.entity.type | Publication | |
| oairecerif.author.affiliation | Universiti Sains Malaysia |