Moving Target Person Detection And Following For Mobile Robot Based On Multi-Sensor Information

Moving target detection and tracking is the detection of moving objects from the scene, and the tracking of moving objects with the process. It is an important part of intelligent robotics and human-computer interaction. Currently, most tracking systems are based on visual information. Because of lacking distance information by a single vision, the information from other sensors can be used as a supplement to achieve better tracking results.For the purpose of moving target detection and tracking in more complex indoor environments, this paper, which is based on monocular camera mobile robot platform using sonar ring as an auxiliary sensor, proposed a multi-sensor data fusion tracking method based on the weighted distance. Through the use of visual and sonar ring for moving target information extraction, visual servo tracking and sonar grid method is used to develop control strategies based on the sensor characteristics. Then, the application system for moving targets detection and Robot tracking control is build.Firstly, a robot moving target detection platform system in software and hardware was build. Combined the project plan, this paper analyzed the Cartesian coordinate system, the robot kinematics model, the camera projection model and the image Jacobian matrix. Through the use of interface functions, the bottom-level control library functions was called by the top-level target tracking system in order to control the robot platform.Secondly, background subtraction and three frame differencing method was used to extract moving target region, and gravity method was used to determine the center of the target Projection. On this basis, because of the robot tracking process lags behind the target current location, Kalman filter is used to predict and update the position of the moving target.Finally, this paper used multi-sensor data fusion in tracking moving objects indoor scenes. A background compensation method based on robot motion is proposed to solve the Image deviation problem caused by robot motion. Through the analysis of the robot visual servo model and sonar grid model, Image-based visual servo tracking strategy and Grid-based sonar target tracking strategy were established. Weighting strategy was specified based on the analysis of the sensors range and characteristics. Considering the operation performance of the robot, the resulting control policy is established by tracking decision fusion. Experiments show that this method can quickly and accurately detect moving targets while tracking moving targets stably. In the tracking process, the robot control curve smoothing and have no velocity mutation, which is able to complete the task successfully.