Lidar Point Cloud Completion Using Time-Division Gray Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lidar systems face challenges in generating accurate 3D point clouds due to missing data or holes, especially in complex environments and adverse weather conditions, leading to difficulties in detecting key targets accurately.
Innovation Solution
A method and apparatus that utilize a single sensor to alternately collect 3D point clouds and 2D gray images in a time-division manner, eliminating the need for calibration and registration operations, and leveraging the 2D images to complement missing data in the 3D point clouds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If calibration and registration operations are performed to combine 2D image and 3D point cloud, then point cloud complement accuracy is improved, but algorithm complexity and processing time increase
Solution Approach 1:
The patent extracts the essential function of point cloud complement by using only the 2D gray image information, eliminating the need for complex calibration and registration operations between 2D images and 3D point clouds. The method directly uses the gray image to identify target contours and complement missing point cloud data, simplifying the overall algorithm while maintaining accuracy.
Solution Approach 2:
The patent segments the point cloud data into complete point clouds and incomplete point clouds based on whether they contain target information. By segmenting the processing flow into separate modules (obtaining gray image, extracting contour, complementing point cloud), the algorithm becomes more manageable and efficient without sacrificing complement accuracy.
2Measurement precision
If multiple sensors are used to collect 3D point cloud and 2D gray image simultaneously, then coordinate consistency is improved, but hardware cost and device complexity increase
Solution Approach 1:
The patent uses periodic action by alternately collecting 3D point cloud data and 2D gray image data in a time-division manner using a single sensor. The sensor switches between laser illumination mode for point cloud collection and ambient light mode for gray image collection, achieving coordinate consistency through temporal separation rather than simultaneous multi-sensor capture.
Solution Approach 2:
The patent makes the single sensor universal by enabling it to perform both 3D point cloud collection and 2D gray image capture functions. The same sensor collects different types of data at different times, eliminating the need for separate sensors for each function and reducing hardware complexity while maintaining data consistency.
3Measurement precision
If 2D gray image is used to complement missing point cloud data, then detection accuracy of key targets is improved, but processing time increases
Solution Approach 1:
The patent applies preliminary action by pre-processing the 2D gray image to extract target contours and characteristics before using them to complement the point cloud data. By preparing the gray image information in advance and storing it for later reference, the system reduces real-time processing requirements and speeds up the overall target detection process.
Solution Approach 2:
The patent uses the 2D gray image as an intermediary mediator between the laser illumination data and the target detection algorithm. The gray image serves as a reference that bridges the gap between available point cloud data and the need for complete target information, enabling efficient complement operations without requiring complex real-time processing.
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 enhances the quality of lidar point clouds by reducing hardware costs and improving detection accuracy of key targets without the need for additional processing steps.
Implementation Method 1
a point cloud of a first unit is obtained under laser illumination
Implementation Method 2
a gray image of the first unit is obtained under ambient light illumination
Implementation Method 3
the sensor is configured to separately collect optical signals when the laser is turned on or turned off, and convert the optical signals into electrical signals
Data Source
AI summary
This application provides a method includes: obtaining a to-be-processed point cloud and a reference gray image corresponding to a radar field of view, where the to-be-processed point cloud is synthesized by point clouds of a plurality of units, the reference gray image is synthesized by gray images of the plurality of units, a point cloud of a first unit is obtained under laser illumination, a gray image of the first unit is obtained under ambient light illumination, the point cloud of the first unit and the gray image of the first unit are alternately collected by a same sensor, and the first unit is any one of the plurality of units; obtaining a contour of a target based on the reference gray image; and complementing the to-be-processed point cloud based on the contour of the target to obtain a target point cloud.