LiDAR Object Tracking with Overlap Removal for Accurate Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inaccurate information processing by lidar sensors in autonomous vehicles leads to reduced reliability, necessitating improved object detection and tracking methods.
Innovation Solution
A method and device using lidar sensors to determine whether overlapping objects within detected boxes are removable based on shape information, allowing for accurate removal of irrelevant objects and enhancing system performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If lidar sensor detects all objects within bounding boxes, then detection coverage is improved, but false detection of removable overlapping objects increases
Solution Approach 1:
The patent extracts and removes removable overlapping objects from bounding boxes based on shape information analysis. By identifying objects that can be removed (such as objects completely contained within another object's bounding box or objects that do not contribute to the main detection target), the system eliminates false detections while preserving true targets, thereby improving both detection accuracy and information reliability
Solution Approach 2:
The patent performs preliminary shape information analysis and determination of removable objects before final object detection and tracking. By pre-processing the point cloud data to identify and mark removable overlapping objects, the system prepares the data in advance, preventing false detections from propagating through subsequent processing stages
2Measurement precision
If all detected objects are tracked, then tracking completeness is improved, but system processing efficiency deteriorates
Solution Approach 1:
The patent extracts and removes removable overlapping objects from the detection results before the tracking stage. By eliminating objects that are completely contained within other objects or do not represent independent tracking targets, the system reduces the number of objects requiring tracking, thereby improving processing efficiency without compromising tracking completeness for meaningful targets
Solution Approach 2:
The patent performs preliminary filtering and removal of removable objects before the tracking process begins. This pre-processing step prepares a refined set of objects that actually require tracking, preventing unnecessary processing overhead in subsequent tracking operations while ensuring all relevant objects are still captured
Data Source
AI summary
An object detection method using a lidar sensor of an embodiment includes determining whether a box of a target object is a box in which an overlapping object present therein can be deleted on the basis of shape information of the target object obtained by the lidar sensor, and generating a box track of the target object after removing the overlapping object according to a determination result.