Research On Scene Adaptive Methods For Multi-Source Pedestrian Navigation

In the satellite denial environment,the real-time robust navigation of pedestrians is a challenging problem.Typical pedestrian autonomous navigation systems usually estimate the position and attitude relative to a known starting point based on vision or inertia.However,due to the limitations of vision and inertia,the common visual slam and INS systems have poor adaptability and robustness to pedestrians and the environment.To improve the adaptability of pedestrian autonomous navigation systems in different scenes or for different pedestrians,this paper integrates other sensors based on foot MIMU and camera,designs a pedestrian autonomous navigation framework based on multi-source information fusion,and studies the scene adaptability of strapdown inertial navigation system and visual slam respectively.The main research results are:1.Aiming at the insufficient adaptability of ZUPT-aided SINS to pedestrians and their motion state,the paper proposed the zero-velocity correction method of pedestrian navigation based on RNN and the optimization method of pedestrian inertial navigation parameters based on factor graph,so that the system can adaptively adjust the navigation parameters and effectively suppress the error divergence,so as to improve the positioning accuracy and robustness of the navigation.2.Aiming at the poor robustness of traditional visual slam system and poor real-time performance of dynamic SLAM,the paper implemented the construction of pedestrian visual navigation system based on ORB-SLAM 2.Combined with deep learning,the paper designed the target recognition and boundary segmentation algorithm for pedestrian autonomous navigation and the pedestrian visual navigation algorithm to suppress the influence of environmental dynamic information,so as to accurately remove the dynamic objects in the image frame and improve the dynamic environmental adaptability of the visual navigation system.3.combined with the motion characteristics of pedestrians,a pedestrian multi-source fusion navigation system combining optimization and filtering is designed based on inertial navigation system and visual SLAM algorithm.The simultaneous interpreting and positioning method can give full play to the advantages of different sensors and achieve accurate and reliable navigation of pedestrians in complex scenes.The adaptability of the algorithm in complex scenes is verified by lots of experiments,and the experimental results are compared with a typical visual slam system and inertial navigation system.The results show that the proposed algorithm has higher positioning accuracy and robustness in complex scenes.