3D Bounding Box Corner Point Detection for Occluded Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional point cloud object detection methods face challenges in accurately determining 3D center point coordinates due to objects being covered, leading to reduced accuracy in 3D bounding box detection, which affects the overall object detection precision.
Innovation Solution
A method involving the detection of 3D object bounding boxes using 3D coordinates of bounding box corner points, utilizing a point cloud object detection model trained with a regression loss function to adjust loss weights based on visibility and orientation analysis, ensuring accurate detection even when parts of the object are obscured.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional CSA mode is used to obtain 3D bounding box, then the detection process is simple, but the accuracy deteriorates when objects are covered and 3D point clouds are missing
Solution Approach 1:
The patent segments the 3D bounding box detection into two distinct parts: 2D bounding box detection on image planes and depth value detection along the Z-axis. The detector first identifies 2D bounding boxes in multiple image planes, then determines depth values for these boxes. Finally, it combines the 2D coordinates and depth values to construct the complete 3D bounding box. This segmentation allows the system to handle occluded objects more effectively by relying on visible portions for 2D detection while independently determining depth information.
2Measurement precision
If 3D center point coordinates are used to represent 3D bounding box, then the representation is compact, but the accuracy deteriorates when objects are covered
Solution Approach 1:
The patent transitions from representing 3D bounding boxes using center point coordinates (CSA mode) to using 8 corner point coordinates. By specifying all 8 corners of the bounding box, the system captures complete spatial information about the object's extent and orientation without relying on potentially inaccurate center point estimates from occluded objects. This dimensional change from center-based to corner-based representation preserves geometric information even when point clouds are incomplete.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to the technical field of autonomous driving, and specifically provides a point cloud object detection method, a computer device, a storage medium, and a vehicle, to solve the problem of improving the accuracy of point cloud object detection. The method includes: obtaining a three-dimensional(3D) point cloud frame collected by a radar, performing object detection on the 3D point cloud frame to obtain a 3D object bounding box represented by 3D coordinates of bounding box corner points, and obtaining an object detection result based on the 3D object bounding box. Through the method, even if an object is covered, coordinates of uncovered end points of the object can be accurately obtained based on 3D coordinates of bounding box corner points in a 3D object bounding box, so that the accuracy of object detection can be effectively improved, and effective tracking corner points are provided for object tracking, thereby ensuring the accuracy and reliability of object tracking.