Point Cloud Correction via Stereo-LiDAR Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current point cloud generation technologies face challenges in achieving high resolution and accuracy, particularly due to the high cost of high-resolution LiDARs and the low accuracy of stereo camera-based point clouds, making it difficult to generate reliable sensor data for machine learning algorithms in vehicle driver assistance and autonomous vehicles.
Innovation Solution
A system and method that combines low-resolution LiDARs and stereo cameras to generate high-resolution and high-accuracy point clouds by determining errors and correction functions based on reference LiDAR point clouds, iteratively refining the point clouds until error thresholds are met, allowing for the use of less expensive sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution LiDARs are used to generate accurate point clouds, then measurement precision is improved, but device cost increases significantly
Solution Approach 1:
The patent replaces expensive high-resolution LiDARs with cheaper low-resolution LiDARs and stereo cameras. The system uses multiple inexpensive sensors to achieve the functionality of a single expensive sensor, thereby reducing device cost while maintaining measurement precision through computational methods.
Solution Approach 2:
The patent combines data from low-resolution LiDAR and stereo camera into a unified point cloud representation. By merging the 3D geometric information from LiDAR with the high-resolution visual information from the stereo camera, the system achieves both cost reduction and maintained measurement precision.
2Ease of manufacture
If low-resolution LiDARs are used to reduce cost, then device cost decreases, but measurement precision deteriorates due to sparsity
Solution Approach 1:
The patent introduces a correction function as an intermediary that maps points from the low-resolution LiDAR point cloud to the high-resolution stereo camera point cloud. This correction function acts as a mediator that transfers the high-resolution details from the stereo camera to enhance the low-resolution LiDAR data, thereby improving measurement precision without increasing device cost.
Solution Approach 2:
The patent creates a high-resolution point cloud by copying and transforming stereo camera points into the LiDAR coordinate system using a learned correction function. This copying process allows the low-resolution LiDAR to benefit from the high-resolution details captured by the stereo camera, improving point cloud resolution while maintaining cost effectiveness.
3Ease of manufacture
If stereo cameras are used instead of LiDARs, then device cost decreases, but measurement precision deteriorates due to low accuracy in spatial coordinates
Solution Approach 1:
The patent employs a feedback mechanism where the system iteratively refines the correction function by comparing the transformed stereo camera points with the reference LiDAR points. The error between these point clouds is calculated and used to update the correction function, progressively improving spatial coordinate accuracy while maintaining the use of cost-effective stereo cameras.
Solution Approach 2:
The patent replaces the direct geometric measurement capability of LiDAR with a computational approach using stereo vision and learned correction functions. Instead of relying on the mechanical precision of LiDAR ranging, the system uses image processing and neural network-based correction to achieve accurate spatial coordinates, thereby reducing device cost while improving measurement precision.
Data Source
AI summary
A system, device and method of generating high resolution and high accuracy point cloud. In one aspect, a computer vision system receives a camera point cloud from a camera system and a LiDAR point cloud from a LiDAR system. An error of the camera point cloud is determined using the LiDAR point cloud as a reference. A correction function is determined based on the determined error. A corrected point cloud is generated from the camera point cloud using the correction function. A training error of the corrected point cloud is determined using the first LiDAR point cloud as a reference. The correction function is updated based on the determined training error. When training is completed, the correction function can be used by the computer vision system to generate a generating high resolution and high accuracy point cloud from the camera point cloud provided by the camera system.