Research On Visual Tracking System Of Service Robot

With the development of robot intelligent technology and increasing number of elderly population,intelligent service robot which apply to old people's home is growing up.As robot receptionists and robots feeding applied to commercial places,People are demanding more and more intelligent levels of robots,hoping that robots can handle more dynamic,more random daily affairs and emergencies in the natural environment.Several factors,such as the dynamic and uncertainty of the nursing environment,a variety of indoor furniture,different places decoration of different objects,contacts and occlusions between objects,etc.,higher requirements for the service robot daily work intelligence.In this paper,the service robot vision tracking system for nursing home is used as the research object,and the video image processing technology is used to construct a set of functions such as specific object determination and feature extraction,specific object detection and specific object tracking system in the tracking process.The robot visual motion model is established to study the background motion problem caused by robot movement.A modified background compensation method based on ORB feature matching is proposed to extract and reduce the target area's computational complexity of subsequent image processing.In the detection and recognition phase of a particular object,a new weight calculation method is designed and applied to the calculation of the image face proportion,and the judgment is made on the aspect ratio calculation of the image by using the third-party dynamic face matching system to complete the region division and feature extraction of the tracking object.So as to determine the target object is full body or half body into the image.The second step is to propose a fast pedestrian area detection method based on the vertex features of the human head,and to carry out pedestrian detection with the half-body model or DPM model,and to extract the characteristics of the extraction area.Finally,the situation is compared with the detected pedestrian characteristics and target object similarity,so as to determine whether the detected pedestrian is not to track the target object.Using the improved particle filter technology to track the target object.Combined with the third party dynamic face matching software quickly to the particle filter algorithm an initial target,and then track the target.In the target tracking process,in the case of possible deviation or loss of target,the introduction of a specific pedestrian detection program to track the object to be modified to improve the correctness of tracking to prevent the tracking process of missing targets.Finally,summarized the full paper,analyzed the existing problems of the tracking system,and proposed some suggestions for the next research.