Camera-LiDAR Calibration Using 3D Hole Marker Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sensor fusion technologies face challenges in accurately calibrating heterogeneous sensors like cameras and LiDARs, particularly in determining the relative position and direction between them, which affects the accuracy of fused data, especially for autonomous driving systems that require reliable recognition of distant objects.
Innovation Solution
A method using a 3D marker board with holes to calibrate camera and LiDAR sensors by determining a translation vector based on the radii of the holes recognized in both image and point cloud data, allowing for accurate projection of point cloud data onto image data and evaluation of fusion data accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data fusion is performed between heterogeneous sensors (camera and LiDAR) without accurate calibration, then the system can process data from multiple sensors, but the accuracy of fusion data deteriorates due to ambiguous correspondence relationships
Solution Approach 1:
A marker board with holes serves as an intermediary object between the camera and LiDAR sensors. The marker board provides common reference features (holes) that both sensors can detect, enabling establishment of correspondence relationships. The holes appear as circular features in both image data and point cloud data, providing a reliable basis for calculating the translation vector and achieving accurate calibration between heterogeneous sensors.
Solution Approach 2:
The invention uses the radius of holes as a measurable parameter that can be extracted from both camera image data and LiDAR point cloud data. By comparing the radius measurements from both sensors and using optimization algorithms, the system determines the translation vector parameters that maximize the correspondence between sensor coordinate systems, thereby achieving accurate calibration.
2Measurement precision
If a 3D marker board with holes is used for calibration, then the accuracy of detecting distant objects is improved, but the device complexity increases due to the need for marker board recognition and processing
Solution Approach 1:
The calibration process is segmented into distinct steps: (1) detecting the marker board in the camera image, (2) identifying holes in the marker board, (3) extracting hole radius from image data, (4) detecting the marker board in LiDAR point cloud data, (5) identifying holes and extracting radius from point cloud data, and (6) calculating the translation vector. This segmentation makes the complex calibration process more manageable and systematic.
Solution Approach 2:
The marker board with holes serves as a self-contained calibration target that provides all necessary information for calibration. The holes themselves provide both positional and dimensional (radius) information that both sensors can independently measure, allowing the system to self-calibrate without requiring external reference measurements or complex calibration apparatus.
3Measurement precision
If the translation vector is determined based on hole radius from both image and point cloud data, then the calibration accuracy is improved, but the difficulty of detecting and measuring increases due to the need to process both data types
Solution Approach 1:
The invention focuses on detecting specific local features (holes) in the marker board rather than attempting to process the entire scene data. By concentrating on the hole features which have distinct characteristics (circular shape in images, spherical point distributions in point clouds), the system simplifies the detection process while maintaining high calibration accuracy through optimized radius measurement algorithms.
Data Source
AI summary
An apparatus and method for data fusion between heterogeneous sensors are disclosed. The method for data fusion between the heterogeneous sensors may include identifying image data and point cloud data for a search area by each of a camera sensor and a LiDAR sensor that are calibrated using a marker board having a hole; recognizing a translation vector determined through calibrating of the camera sensor and the LiDAR sensor; and projecting the point cloud data of the LiDAR sensor onto the image data of the camera sensor using the recognized translation vector to fuse the identified image data and point cloud data.