Intelligent Vehicle Localization via Dynamic Object Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mapping and localization systems in autonomous driving struggle with accuracy in dynamic scenes due to movable objects, which affect the precision of vehicle localization.
Innovation Solution
A method and apparatus that utilize a combination of 4D millimeter wave radar, lidar, and RGB camera data to acquire and process point cloud data and images. This involves target detection, object classification into static and dynamic categories, and the determination of observation weights to improve localization accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If mapping and localization systems assume a static environment, then system complexity is reduced, but localization accuracy deteriorates in dynamic scenes
Solution Approach 1:
The patent segments the environment into static and dynamic objects by processing point cloud data to identify movable objects. This segmentation allows the system to handle dynamic scenes by separating static mapping from dynamic object tracking, resolving the contradiction between system simplicity and localization accuracy in dynamic environments
Solution Approach 2:
The patent introduces dynamic object detection and tracking mechanisms that adapt to moving objects in the environment. By making the localization system dynamic rather than static, it can accurately localize vehicles even when the environment contains movable objects, thus improving localization accuracy without requiring complete system redesign
2Ease of operation
If the system processes all point cloud data equally, then processing simplicity is maintained, but localization precision deteriorates due to interference from movable objects
Solution Approach 1:
The patent applies different processing qualities to different regions of the point cloud data. Static regions are processed with standard mapping algorithms while dynamic regions containing movable objects are processed with specialized detection and filtering algorithms. This local quality differentiation improves localization precision by reducing interference from movable objects while maintaining processing simplicity through region-based processing
3Measurement precision
If the system uses multiple sensors (4D millimeter wave radar, lidar, RGB camera), then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent merges data from multiple sensors (4D millimeter wave radar, lidar, and RGB camera) into a unified processing framework. By combining these sensors and their data streams into an integrated system that processes point cloud data together, it achieves high localization accuracy while managing device complexity through unified data processing rather than separate independent systems
Data Source
AI summary
A method and apparatus for localizing an intelligent vehicle in a dynamic scene are provided. The localizing method includes: processing an RGB image using a target detection model to determine a rectangular box of a movable object; converting first point cloud data and second point cloud data to a pixel coordinate system, and dividing the movable object into a static object and a dynamic object; converting the second point cloud data of the movable object in the pixel coordinate system to a map coordinate system to obtain a semantic point cloud map; converting the second point cloud data of a communication area, and of a rectangular box, in the pixel coordinate system to the map coordinate system to obtain a static point cloud map; and determining observation weights for objects in the second point cloud data, and thereby determining pose information of the intelligent vehicle at the current time.