LiDAR Point Cloud Densification via Motion Correction and Neighborhood Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current LiDAR systems face challenges in generating high-quality densified point clouds due to blurring and transparency issues with moving and static objects, particularly when registering LiDAR point clouds from different time instances, which are exacerbated by the need for additional stereo camera setups and their associated costs and inaccuracies.
Innovation Solution
A computer-implemented method that combines LiDAR point clouds by correcting for the movement of the LiDAR device between measurement times and applying neighborhood filtering to include only points within a predefined neighborhood of the reference LiDAR point cloud, enhancing the densified point cloud with additional points based on statistical distributions, thereby improving the accuracy and density of the LiDAR data without requiring a stereo camera setup.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If LiDAR point clouds from different time instances are registered by stacking them onto each other using rigid body transformations, then the number of points is increased (densification), but moving objects become blurred due to registration at different time instances
Solution Approach 1:
The system performs preliminary classification of points as static or moving before densification. By identifying static points in advance using temporal consistency checks across multiple frames, the system can safely accumulate these points over time without causing blurring, while excluding moving objects from accumulation to preserve their accuracy.
Solution Approach 2:
The patent applies different processing rules to different regions of the point cloud based on object motion characteristics. Static regions allow point accumulation and densification, while moving regions use different criteria (such as latest frame dominance or motion-aware fusion) to prevent blurring, thus applying local quality differentiation to resolve the contradiction.
2Measurement precision
If stereo camera setup is installed to filter blurred dynamic and transparent objects, then filtering accuracy is improved, but device complexity and cost increase due to additional hardware and calibration requirements
Solution Approach 1:
The LiDAR system performs self-service by using its own temporal data to classify and filter points. The system analyzes temporal consistency of reflected intensity values across multiple frames to automatically identify static versus moving objects, eliminating the need for external stereo camera systems while maintaining filtering accuracy through self-contained temporal analysis.
Solution Approach 2:
The patent replaces the mechanical/optical stereo camera system with a computational approach using temporal analysis of LiDAR data. Instead of using additional optical hardware for depth estimation and filtering, the system substitutes this with software-based temporal consistency checks and machine learning classifiers that process LiDAR intensity data over time.
3Speed
If traditional convolutional networks are applied to sparse LiDAR data, then processing speed is maintained, but performance deteriorates due to poor handling of sparse data even when location of missing data is provided
Solution Approach 1:
The patent transforms the sparse LiDAR point cloud data into a denser representation by accumulating and interpolating points in 3D space. This parameter change from sparse to dense point cloud structure makes the data more suitable for convolutional network processing, improving both accuracy and utilization of spatial features while maintaining processing efficiency through optimized dense data structures.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach results in a significantly increased number of points in the densified LiDAR point cloud, with improved accuracy and reduced costs, as it effectively treats static and moving objects differently, maintaining useful LiDAR points and providing accurate depth values without the need for camera calibration or additional hardware.
Implementation Method 1
Emission of the laser pulse triggers a circuit for measuring the time-of-flight of the laser beam. Based on the measured time-of-flight of the laser beam, a respective distance to an object in the surrounding of the LiDAR system can be determined.
Data Source
AI summary
A computer-implemented method for generating a densified LiDAR point cloud includes receiving a plurality of LiDAR point clouds, including a reference LiDAR point cloud and remaining LiDAR point clouds, wherein the plurality of LiDAR point clouds is obtained based on measurements by a LiDAR device of a vehicle at subsequent measurement times. The densified LiDAR point cloud is generated by combining the reference LiDAR point cloud and the remaining LiDAR point clouds, wherein each remaining LiDAR point cloud is transformed by correcting for a movement of the LiDAR device in the time span between the measurement time of the remaining LiDAR point cloud and the measurement time of the reference LiDAR point cloud, and wherein only those points of the remaining LiDAR point clouds are combined into the densified LiDAR point cloud which are located in a predefined neighborhood around a point of the reference LiDAR point cloud.