Hybrid Algorithm For Fast Detection And Recognition Of Signage And Obstacle Avoidance For Robot Navigation

Abstract

Robot navigation means the robot's ability to determine its own position in its frame of reference (map) and then to plan a path towards some goal location by avoiding dangerous situations such as collisions and unsafe conditions. The use of sign recognition and obstacle avoidance for the navigation of the robot has been increasing due its importance in many sectors such as for autonomous car, for military surveillance robot, for indoor surveillance robot and for assisting visual impaired people with robot. Therefore, ability to observe sign and understand the meaning by the robot is essential part for navigating the robot safe and to reach the desire end without any difficulty. There has been very little work directed toward assuring safe condition for navigation of the robot using computer vision and also current robot navigate through the unknown environment by using sensing mechanisms. So introducing the mechanism that the robot uses vision techniques will improve the performance of the robot to navigate through the environment better. Therefore, integrating fast vision algorithm that can enable the robot to understand sign and avoid obstacle make the robot to navigate safe within unknown indoor environment. This study enables the robot to detect and recognize sign with high speed and enable the robot to avoid any obstacle. This study integrates fast recognition algorithm with obstacle avoidance mechanism to make the robot to navigate by understanding the environment it exists and to plan paths accordingly by avoiding obstacles. To give solution for the problem first current existing v ision algorithms is studied and the best in computation time and reliable in detection have been chosen then simulation for executing the proposed algorithm is selected by their computability time, flexibility and by their feature factor which are best are selected. Because of robot requires fast and accurate response various vision algorithm have been studied and due to the problem of large computation and slow speed, a fast matching algorithm based on the combination of ORB feature point and SURF descriptor is proposed. Experiments show that compared to the classic SIFT and SURF matching algorithm, the method is very good to achieve the goal of fast matching, in addition to the speed also improves the accuracy of the matching. The navigation is done by identifying the indoor sign and making the decision to either to re arrange its path again or to stop its motion based on the sign it observes. A proof of concept is presented using TurtleBot3, a robot kit, which runs on robot software development framework Robot Operating System (ROS).