期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2020
卷号：98
期号：4
页码：714-724
出版社：Journal of Theoretical and Applied
摘要：The key contribution in this research work is introducing a technique that can be used to avoid unknown objects to a mobile robot. Most vision-based obstacle avoidance systems tend to have high computational complexity in order to compensate for accuracy. The proposed method tries to overcome that using pattern recognition of two laser pointers in the view of a single camera. This vision combination along with the program that runs on a base station is designed as a module that detects objects around the robot. Such obstacles are detected by calculating the intensity of the red laser points found on each frame being captured by the camera in real-time. Distance and angle to the objects is measured using Lagrange interpolation formula applied separately to each laser projection in the framed image. A map is created to show the robot’s actual distances versus estimated ones as the robot keeps track of objects which are in the camera’s view. The algorithm successfully manages to avoid obstacles as shown in the experiments. The effectiveness of the proposed system is evaluated by deploying the robot and performing a simple navigation task. The results show that the concept algorithm along with the hardware module is able to utilize the monocular vision with classification error of 3.63%.
