Robot patrolling using sensor data and image features
Sheng-Bin Hsu, Cheng‐Chang Lien, Chang-Hsing Lee, Chin‐Chuan Han, Shaopeng Chen, Yang-Lang Chang · 2014
Location identification is an essential module for robot patrolling. Robots follow the planed routes for security patrolling. It spends much computational time for identifying the robot location in the traditional image-based methods during the patrolling process. In this paper, a fast robot patrolling framework is proposed by integrating sensor data and image features. In the sensor-based module, a small amount of WiFi signals is collected for roughly identifying the robot location. Moreover, the travel direction is determined by the electronic compass. Using the position and orientation information, few gallery images are matched by visual features in the image-based module. SURF-based features are extracted for finely identifying the robot location and moving direction. A wide area has been patrolled using the proposed framework. In addition, pedestrian are also detected by the histogram of gradient (HOG) features during robot patrolling.