A bionic polarization semantic slam method based on neural radiation field

Abstract

The application provides a bionic polarization semantic SLAM method based on a neural radiation field, and the system comprises three parts: a bionic polarization pose estimation module based on factor graph optimization, a dense mapping module based on a neural radiation field, and a semantic information positioning module based on bionic polarization infrared. The system first collects multi-sensor data, establishes state equations and measurement equations, and estimates the pose using factor graph optimization. Secondly, using the estimated pose and lidar data, the neural implicit representation (NeRF-LOAM) model of lidar odometry and map construction is used to generate voxel embedding maps and grid reconstruction for dense mapping. Finally, the polarization infrared fusion image is processed by Yolov8 and SegNeXt, and the semantic positioning is realized in combination with the lidar features. The system significantly enhances the anti-interference and environmental adaptability, shortens the mapping training time, and can be applied to multiple fields such as autonomous driving, intelligent transportation, and smart logistics.

Images