Lidar-Camera Alignment via Inverse Distance Transformation Edge Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately aligning sensor data from different sources, such as lidar and cameras, for precise navigation and data fusion, which requires effective calibration of extrinsic parameters.
Innovation Solution
A method and system that involves computing edge maps from lidar and camera data, aligning points using an inverse distance transformation (IDT) edge map, and determining extrinsic parameters through a two-stage search method with quality estimation, allowing for iterative refinement until a confidence threshold is met.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data from lidar and camera are combined for autonomous vehicle navigation, then the accuracy of environment sensing and navigation is improved, but the difficulty of aligning sensor data and calibrating extrinsic parameters increases
Solution Approach 1:
The patent introduces an intermediary calibration process that uses identified features (corners, edges, contours) from both lidar and camera data as mediator objects. These features serve as common reference points that facilitate the alignment between the two sensor coordinate systems, making the calibration process more manageable and accurate.
Solution Approach 2:
The patent implements an iterative feedback mechanism where extrinsic parameters are continuously refined based on the alignment quality between lidar and camera features. The system computes alignment errors, uses them to update extrinsic parameters, and repeats the process until convergence, ensuring high precision in sensor fusion.
2Ease of manufacture
If extrinsic parameters are calibrated using traditional methods, then the alignment process is simpler, but the precision of sensor data association deteriorates
Solution Approach 1:
The patent performs preliminary actions by pre-identifying and extracting key features (corners, edges, contours) from sensor data before the actual alignment process. This preprocessing step creates a structured set of reference points that simplify subsequent calibration operations while maintaining high precision through feature-based matching.
Solution Approach 2:
The patent employs dynamic refinement of extrinsic parameters through iterative optimization. Rather than using static calibration values, the system continuously adjusts parameters based on real-time feature alignment quality, allowing the calibration precision to adapt and improve throughout the operation.
3Measurement precision
If iterative refinement of extrinsic parameters is performed to improve alignment precision, then the accuracy of sensor fusion is improved, but the computation time and processing complexity increase
Solution Approach 1:
The patent segments the calibration process into distinct stages: feature identification, initial parameter estimation, iterative refinement, and convergence validation. This segmentation allows the system to apply different computational strategies at each stage, optimizing the balance between precision and processing time.
Solution Approach 2:
The patent implements a stopping criterion that performs iterative refinement only until a predefined precision threshold is met or a maximum iteration count is reached. This partial action approach avoids unnecessary computations beyond the required precision level, reducing processing time while maintaining sufficient alignment accuracy for autonomous navigation.
Data Source
AI summary
Systems and method are provided for controlling a vehicle. In one embodiment, a method includes: receiving, by a controller onboard the vehicle, lidar data from the lidar device; receiving, by the controller, image data from the camera device; computing, by the controller, an edge map based on the lidar data; computing, by the controller, an inverse distance transformation (IDT) edge map based on the image data; aligning, by the controller, points of the IDT edge map with points of the lidar edge map to determine extrinsic parameters; storing, by the controller, extrinsic parameters as calibrations in a data storage device; and controlling, by the controller, the vehicle based on the stored calibrations.