Research On Indoor Self-localization Technology Of Mobile Robot Based On Kinect Sensor

Autonomous localization and trajectory construction of mobile robots is one of the key technologies to realize robot intelligence.Because of the depth sensor can directly acquire the depth information,a new method to realize visual location based on RGB-D sensor has been a hot research topic in recent years.In the thesis,the Kinect sensor is used to research the indoor self-location technology of mobile robots.Aiming at the shortcomings of traditional vision location technology,an optimization scheme based on Kinect for indoor self-location technology of mobile robot is proposed to improve the robustness and accuracy of mobile robot self-location.The main work of the thesis is as follows:1.The main structure and depth measurement principle of the Kinect v2 sensor are described;A detail derivation is carried out on the mathematical modle of camera imaging,determines the required for camera calibration parameters;The iai?kinect tool was used to calibrate the Kinect sensor,The Kiect sensor was calibrated using the iai?kinect tool,obtains the camera intrinsics and distortion coefficients.2.The principle of visual location algorithm based on feature points method is studied,and a visual location optimization method based on improved ORB features is proposed.Aiming at the problem of large number of mismatches in the feature matching stage,combined with the feature matching method based on grid motion statistical optimization,the wrong matching point pairs are effectively eliminated,which provides more accurate input for solving motion estimation.The accurary of location estimation is increased by 23%.3.Aiming at the low robustness of the visual location algorithm based on feature points method in the indoor environment with lack of texture,the thesis analyzes the principle of visual localization based on direct method,and combines the improved ORB feature points method and semi-dense direct method to realize a visual location method combining feature point method and direct method.The method evaluates the texture features of the environment where the mobile robot is located by establishing an environmental feature evaluation function.When the texture features of the environment are rich,the improved ORB feature point method is selected for location estimation;when the environmental texture features are sparse,the semi-dense direct method is selected realize location estimation.4.Based on the improved related modules,the Kinect sensor-based mobile robot indoor self-location software system is designed and implemented,and the system is verified by the standard public dataset and the real environment test data collected by the mobile robot.The experimental results show that the self-localization system with improved ORB feature points method and semi-dense direct method has reliable accuracy and good robustness in indoor environment with rich texture features and sparse texture features.