Intelligent Behavior for Mobile Robot Based on Indoor Scenes Recognition by Means of Human-Computer Interaction

Abstract

An intelligent behavior research for mobile robot based on scenes recognition under human-computer interaction is proposed. To address the mobile robot reconnaissance indoor environment issue, intelligent behavior is autonomously planned based on scenes recognition. A method of images feature extraction based on phase congruency using LogGabor filter is accepted, after PCA transformation and SVM, the robot identifies the scene. In complex environment, robot firstly recognizes the scene to plan behavior autonomously, meanwhile returns detection data from sensors. Once the robot misidentifies the scene, the data from sensors would correct the error. If the detection data is incorrect, active intervention under human-computer interaction is executed to modify the behavior planning for mobile robot. The algorithm has been tested on the mobile robot Hunter-5. The experiment results indicate that the algorithm can reach up to a correct localization rate of 95.5%, and is robust to the various illumination and contrast.