LiDAR Point Cloud Densification With Motion-Compensated Neighborhood Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
LiDAR sensors on moving vehicles produce sparse point clouds, leading to blurred or transparent objects during registration, and existing methods like stereo camera setups are costly, error-prone, and inaccurate.
Innovation Solution
A method to generate a densified LiDAR point cloud by combining subsequent LiDAR point clouds, correcting for movement, and applying neighborhood filtering to include only points within a predefined neighborhood of a reference point cloud, without requiring a stereo camera setup, and enhancing with additional points based on statistical distributions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If subsequent LiDAR point clouds are stacked onto each other using rigid body transformations to densify the point cloud, then the number of points increases, but moving objects become blurred due to registration at different time instances
Solution Approach 1:
The patent applies preliminary action by performing motion compensation using IMU data before stacking the point clouds. The LiDAR device's movement is corrected in advance based on inertial measurement unit measurements, allowing accurate temporal alignment of point clouds from different time instances without causing motion blur.
2Quantity of substance
If LiDAR point clouds from different time instances are registered onto a middle frame, then point density increases, but static objects may appear transparent when the vehicle moves behind them
Solution Approach 1:
The patent applies preliminary action by performing motion compensation using IMU data before stacking the point clouds. The LiDAR device's movement is corrected in advance based on inertial measurement unit measurements, allowing accurate temporal alignment of point clouds from different time instances without causing motion blur.
3Measurement precision
If a stereo camera setup is used to filter blurred and transparent objects, then object accuracy improves, but device complexity and calibration requirements increase
Solution Approach 1:
The patent extracts and uses only the essential motion compensation function from the complex stereo camera system. By utilizing the IMU's motion data to pre-compensate LiDAR point cloud positions, the solution achieves accurate object representation without requiring additional stereo cameras or their associated calibration procedures.
Solution Approach 2:
The patent applies self-service by using the LiDAR system's own integrated IMU sensor to perform motion compensation. The system leverages its existing inertial measurement capabilities to correct for vehicle movement, eliminating the need for external stereo camera systems and reducing overall device complexity.
4Measurement precision
If stereo matching depth estimation is compared with LiDAR measured depth, then filtering accuracy improves, but processing cost and parameter tuning complexity increase
Solution Approach 1:
The patent extracts and uses only the essential motion compensation function from the complex stereo camera system. By utilizing the IMU's motion data to pre-compensate LiDAR point cloud positions, the solution achieves accurate object representation without requiring additional stereo cameras or their associated calibration procedures.
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 significantly increases the number of points in the LiDAR point cloud, maintaining accuracy and eliminating blurring or transparency issues, while reducing costs and calibration needs, and allowing full 360-degree coverage.
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
Figure 1~3
Figure 4~6
AI summary
The invention relates to a method for generating a densified LiDAR point cloud by a device (1). A plurality of LiDAR point clouds is received, consisting of 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 (5) of a vehicle at subsequent measurement times. A densified LiDAR point cloud is generated by combining, in a common coordinate system, 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 (5), 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. The densified LiDAR point cloud is further enhanced by including a plurality of further points, selected based on a statistical distribution around the points of the densified LiDAR point cloud. Preferably, the step of correcting for a movement of the LiDAR device (5) between the measurement times is performed using sensor data obtained from an inertial measurement unit, IMU, (7) of the vehicle. The device (1) further comprises an interface (2), a computing device (3), and a storage device (8). By using a neighborhood filtering method, static objects and moving objects can be treated differently.