Lidar Camera Calibration Using Image Pyramid Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current calibration methods between laser radar and camera systems are inflexible and cannot perform real-time online calibration, requiring a fixed site and offline processes, which limits their application in dynamic environments like driverless vehicles.
Innovation Solution
A method and apparatus that obtain calibration data from multiple locations, build image pyramids, and perform a top-to-bottom search to determine optimal calibration parameters, enabling real-time online calibration without relying on specific targets or fixed sites.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional calibration methods using checkerboards or QR codes are employed, then calibration accuracy can be maintained, but the system requires fixed sites and cannot perform real-time online calibration
Solution Approach 1:
The system uses naturally occurring scene features (edges, corners, lines in the environment) as calibration targets instead of requiring external calibration objects like checkerboards. The laser radar and camera calibrate themselves using the environment they are already observing, enabling real-time calibration without fixed sites or special targets.
Solution Approach 2:
The calibration system transitions from static offline calibration to dynamic online calibration. The calibration parameters are continuously updated in real-time as the movable platform moves through different positions, allowing the system to adapt to changing environments and maintain accuracy during operation.
2Reliability
If offline calibration at fixed sites is performed, then calibration quality can be ensured, but the process cannot be performed online in real-time and requires maintaining fixed calibration sites
Solution Approach 1:
The invention extracts the calibration function from the fixed site infrastructure and embeds it within the movable platform itself. By using the platform's own sensors (laser radar and camera) to observe and calibrate against environmental features, the system eliminates the need for external fixed calibration sites while maintaining calibration quality.
Solution Approach 2:
The laser radar and camera serve dual purposes: they perform both operational sensing (detecting objects and environment) and calibration functions simultaneously. This multi-functionality eliminates the need for separate calibration equipment and fixed sites, reducing system complexity while maintaining reliability.
3Measurement precision
If multiple calibration parameters are optimized, then calibration accuracy improves, but the computational complexity and time required for calibration increases
Solution Approach 1:
The system performs preliminary actions by continuously collecting and preprocessing calibration data (edge detection, feature extraction) during normal operation. When calibration is needed, the optimization process starts from pre-processed data rather than raw data, significantly reducing computation time while maintaining accuracy for multiple parameters.
Solution Approach 2:
The calibration process is made continuous rather than periodic. The system continuously optimizes calibration parameters using real-time data from the laser radar and camera, ensuring that calibration accuracy is maintained without interrupting operation. The continuous optimization uses incremental updates rather than full re-calibration, improving speed.
Data Source
AI summary
The present disclosure provides a method and apparatus for calibration between a laser radar and a camera, a device and a storage medium, wherein the method comprises: respectively obtaining N stations of calibration data at N locations, each station of calibration data respectively comprising: image data collected by a camera and point cloud data collected by a laser radar, N being a positive integer; building an image pyramid for each station of image; performing a search for each layer of pyramid in a top-to-bottom order according to a preset initial calibration parameter and point cloud data, and determining an optimal calibration parameter according to search results. The solution of the present disclosure may be applied to achieve real-time online calibration.