Design And Implementation Of Visual Aid System For Service Robot

With the rapid development of artificial intelligence technology,the application of service robots has become more and more popular.The research on perfecting various functions of indoor service robots to better serve humans has received more attention from scholars.The autonomous movement of service robots and human-machine interaction in indoor scenes have been two hot topics in this research field.The visual aid system for service robot proposed in this paper mainly implements two functions: visual positioning,face detection and tracking.In the indoor structured environment,the laser odometer has poor applicability and cannot be used.The visual positioning system studied in this paper can make up for this deficiency by upwardly-obtained visual information,laying the foundation for the follow-up research on robot autonomous movement.The face detection and tracking system can accurately capture and track people's faces,thereby accurately distinguishing people from obstacles.This is the basis for the next man-machine interaction and other operations.In terms of visual positioning system,this article mainly uses the RGB-D data obtained from the calibrated Kinect sensor,and no longer extracts the feature points of the image,but based on the gray invariance theory,that is,the pixel gray levels of the same spatial point in the motion process remains unchanged.The gray information of the semi-dense pixels in the image is directly used.The robot's pose transformation matrix is obtained by minimizing the pixel gray difference between successive frames,and a semi-dense point cloud map is constructed.The experimental hardware platform is based on the Pioneer 3-DX wheeled mobile robot,equipped with experimental equipment such as the ORBBEC Kinect and GIGABYTE IPC.It performs closed-loop experiments in three different indoor experimental scenarios to verify the effectiveness of the positioning system.For face detection and tracking systems,this paper mainly extracts the Haar features of the image.After calculating the Haar feature values using the integral graph method,the strong classifiers are trained by the AdaBoost algorithm,and finally the strong classifiers are cascaded into the final face detector.Based on face detection,a normalized cross-correlation matching algorithm is added to implement face tracking: the face detector is enabled near the area saved in the previous frame,and the template matching algorithm is invoked if no face is detected.The saved template performs face feature matching within the area and finally the system can achieve real-time tracking of the face.And through a professional video set to test the system,the good robustness of the face detection tracking system is verified.