LiDAR Point Association via Color-Coded Image Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for associating LiDAR points with objects detected by computer vision algorithms in autonomous vehicles are computationally intensive, requiring numerous calculations to determine which LiDAR points correspond to detected objects, especially when dealing with large numbers of points and bounding boxes.
Innovation Solution
A method where LiDAR points are projected into an image with bounding boxes, and pixels outside or inside the boxes are colored differently, allowing each LiDAR point to be labeled as associated with an object by determining the color of the corresponding pixel, significantly reducing the number of computations required.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the system calculates whether each LiDAR point falls within each bounding box using coordinate comparison, then accurate object association is achieved, but the computational complexity increases significantly (20 million calculations for 2 million points and 10 bounding boxes)
Solution Approach 1:
The patent introduces an intermediary color-coded image that maps bounding box regions to specific colors. Instead of directly comparing LiDAR point coordinates with bounding box coordinates, the system projects the colored image onto the LiDAR point cloud and uses color matching to determine associations. This intermediary representation transforms a computationally intensive coordinate comparison problem into a simpler color lookup operation, reducing calculations from 20 million to 2 million while maintaining association accuracy
Solution Approach 2:
The patent changes the parameter space from coordinate values to color values. By representing bounding box spatial regions as color-coded pixel values in an image, the system transforms the problem from checking coordinate inequalities (x_min ≤ x ≤ x_max and y_min ≤ y ≤ y_max) to simply comparing color identifiers. This parameter transformation dramatically reduces computational complexity while preserving the spatial association information
2Reliability
If the system processes all LiDAR points with coordinate comparisons for each bounding box, then complete object association is achieved, but the processing time increases
Solution Approach 1:
The patent performs preliminary action by pre-processing the bounding box information into a color-coded image representation before processing LiDAR points. The colored image is created in advance, with each pixel's color encoding which bounding box it belongs to. When LiDAR points are processed, this pre-computed color map allows immediate association determination without repeated coordinate comparisons, significantly reducing processing time while ensuring complete association coverage
Solution Approach 2:
The patent creates a simplified copy of the bounding box spatial information in the form of a color-coded 2D image that mirrors the spatial relationships. This color image copy serves as a lookup table that preserves all necessary association information in a more efficient format, allowing the system to query associations by simply matching point project 3D coordinates to 2D image coordinates and reading the corresponding color value
3Productivity
If the system uses color-coded pixels to represent bounding boxes and determines LiDAR point associations by color matching, then computational efficiency is improved (reduced from 20 million to 2 million calculations), but the system complexity increases
Solution Approach 1:
The patent makes the color-coded image serve multiple functions simultaneously: it acts as a visual representation of bounding boxes, a spatial lookup table for association determination, and a computational optimization structure. This single multi-functional data structure replaces what would otherwise require separate coordinate comparison logic, bounding box storage, and association tracking, reducing overall system complexity despite the added color encoding layer
Solution Approach 2:
The patent substitutes the mechanical coordinate comparison system with an optical/color-based lookup system. Instead of using arithmetic operations to check if a point lies within bounding box boundaries, the system uses color value matching, which is computationally simpler. The complex coordinate geometry operations are replaced by straightforward color equality checks, reducing processing complexity while maintaining the essential spatial association functionality
4Measurement precision
If the system processes LiDAR points without optimization, then accurate associations are made, but the scalability with respect to the number of bounding boxes deteriorates
Solution Approach 1:
The patent transforms the problem from a multi-dimensional coordinate comparison task to a two-dimensional image lookup task. By projecting the 3D bounding box associations onto a 2D color-coded image plane, the system reduces the dimensional complexity of the search space. The number of bounding boxes no longer directly impacts the per-point processing complexity, as all bounding box information is encoded in the 2D color map that can be queried in constant time regardless of how many bounding boxes define the colors
Data Source
AI summary
In one embodiment, a system for associating LiDAR points with objects is provided. Objects in an image are identified and bounding boxes are placed on each object in the image. LiDAR points are received and are projected into the image based on a reference frame of the image. All pixels of the image that are not inside a bounding box are changed to a color such as black. All points that are inside a bounding box are changed to a color that is unique to the particular bounding box. To determine which LiDAR points are associated with the objects in the image, for each LiDAR point a color of the pixel having the same coordinates in the image as the LiDAR point is determined. If the determined color is black than the LiDAR point is not associated with any object. Else, the LiDAR point is located in the bounding box that uses the determined color, and the LiDAR point can be labeled as being located with the object corresponding to the bounding box.