Outdoor perception of robots based on SLAM technology and binocular vision positioning technology

To address the issues of low efficiency and poor reliability in traditional outdoor perception systems for robots, this study integrates stereo vision localization with map-building technology to propose a feature-fused stereo vision perception system. Its core innovation lies in constructing a dynamic weight fusion model based on the coupling of ORB point features and EDLines line features. By integrating a hierarchical bag-of-words tree structure with a sliding window marginalization mechanism, it effectively addresses misalignment and cumulative error issues under dynamic interference, significantly enhancing the system’s robustness and real-time performance. The experimental results show that in the simulation environment, the system achieves an average detection accuracy of 91.08% for obstacles and reconstructed object details with a minimum size of 2.47 cm. Meanwhile, the system operates with high efficiency, with an average response time of 49.68 ms and a computational resource consumption of 122 million floating point operations per second (1.22 GFLOPS). The average obstacle detection distance of the proposed system is 12.75 m. In actual model performance experiments, the average detection accuracy and minimum image reconstruction scale are 84.39% and 3.53 cm, respectively. The average positioning error and absolute deviation are 0.05 m and 0.02 m, respectively. The average frame processing rate and absolute deviation range are 23.52 Hz and 4.83 Hz, respectively. The research results show that the proposed system can improve the stability, real-time performance, and robustness of the robot outdoor perception. The proposed system can promote the application of intelligent robots in more outdoor scene tasks.