Research on Visual Sensor-based SLAM in Dynamic Scenes
In dynamic scenes,simultaneous localization and map construction(SLAM)are easily affected by various factors,which reduces the accuracy and poor robustness.Based on this,this paper first applies feature extraction,and then based on the semantic segmentation network,proposes a pixel-level image semantic segmentation method to segment a priori dynamic target objects.A method is designed to detect the dynamic points of multi-view geometry as a way to dynamically detect the motion state of the research object and combine static features for position estimation.Finally,combining keyframes and depth maps to generate point cloud maps and octree maps containing semantic information,the results show that the algorithm effectively improves the performance of the system in dynamic scenes.
visual sensorlocalization and map buildingdynamic scene
万方数据
