Improvement of Kalman Filtering in Mobile Robot Target Recognition and Intelligent Localization Research

Addressing the challenge of depth information loss faced by monocular mobile robots in the process of target recognition and position estimation, and aiming to enhance localization accuracy, especially when integrating lidar for distance measurement, we propose an innovative Kalman filtering algorithm. This algorithm integrates radar ranging information with azimuth calculation results. Initially, the YOLO network is utilized for efficient target recognition, followed by the acquisition of target azimuth angle information through standard computer vision calibration techniques. On this basis, lidar is employed for multiple directional ranging to obtain multiple observations. During the calculation of the target position, different confidence weights are assigned to each observation based on its distance and direction information. These weights are used to dynamically adjust the observation noise and system noise of the Kalman filter, thereby optimizing the Kalman gain and achieving adaptive target position estimation. The effectiveness of the proposed method is verified through simulation experiments and practical tests. The results demonstrate that compared to traditional methods that solely rely on radar ranging for supplementation, our method achieves more accurate localization and exhibits broad application potential.