Point Cloud Completion Using HSV Brightness Unification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for adding color information to 3D maps, such as those using LiDAR data, fail to accurately handle brightness information, leading to distorted color restoration in missing regions of objects due to light source effects.
Innovation Solution
A point cloud complementing device and method that uses a feature extractor and a point cloud complementary model to correct color information by unifying brightness across adjacent points, assuming they are on the same plane, thereby maintaining consistent brightness and improving color information restoration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If color information is added to point cloud using existing methods, then color information can be complemented for missing portions, but brightness information is not handled correctly causing distorted color restoration
Solution Approach 1:
The patent transforms the color space from RGB to HSV (Hue, Saturation, Value) to separately handle brightness (Value channel) and color information (Hue and Saturation channels). This parameter transformation allows independent optimization of brightness consistency and color restoration, resolving the contradiction between complementing color information and maintaining brightness consistency.
Solution Approach 2:
The patent segments the color information into separate channels: brightness (Value) and color (Hue and Saturation). By processing these channels independently through separate loss functions during training, the model can accurately restore color information while maintaining consistent brightness, avoiding the distortion that occurs when treating color as a unified parameter.
2Device complexity
If point cloud is complemented without considering brightness, then complementation process is simplified, but color information becomes distorted due to light source effects
Solution Approach 1:
By changing to HSV color space, the brightness (Value) component is separated from color information, allowing the model to focus on restoring color (Hue and Saturation) while using brightness from adjacent points as a constraint. This maintains reliability of color information accuracy without excessively complicating the complementation process.
Solution Approach 2:
The patent employs a loss function that provides feedback during training, comparing the complemented point cloud's brightness against the brightness of adjacent observed points. This feedback mechanism ensures that complemented points maintain consistent brightness with their neighbors, improving color information accuracy while keeping the process manageable through automated learning.
3Productivity
If brightness information is not unified across adjacent points, then processing is faster, but restored color information appears distorted in three-dimensional map
Solution Approach 1:
The HSV color space transformation allows the model to process brightness (Value) and color (Hue, Saturation) independently. This parameter separation enables efficient processing where brightness consistency can be enforced through simple Value channel comparisons while color restoration occurs in parallel through Hue and Saturation channel processing, maintaining both speed and quality.
Data Source
AI summary
A point cloud complementing device 100 receives a colored three-dimensional point cloud of each point including a missing region and the number of points to be complemented as inputs. A CPU 11 of the point cloud complementing device 100 extracts a feature vector of the three-dimensional point cloud using a feature extractor learned in advance. The CPU 11 uses a point cloud complementing model learned in advance in consideration of an error between color information and brightness information, has the feature vector and the number of points to be complemented as inputs, and outputs a point cloud obtained by complementing the input three-dimensional point cloud up to the number of points to be complemented by performing correction on point at which the brightness does not change among adjacent points of a predicted point cloud such that the points have close brightness information assuming that the points are on the same plane.