Multi-Source Depth Map Fusion to Reduce Point-Cloud Noise
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for fusing depth data from multiple sources, such as 3D laser scanners and photogrammetry, face challenges in efficiently combining data from different techniques due to noise, outliers, and computational complexity, leading to large and noisy point clouds without valuable information.
Innovation Solution
A novel method for fusing depth data in a 2D space using weighted averaging and correction vectors to merge depth maps from different imaging devices, reducing noise and combining redundant data effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If depth data from multiple sources (3D laser scanners and photogrammetry) are fused using existing methods, then more complete 3D coverage is achieved, but noise and outliers increase leading to large and noisy point clouds
Solution Approach 1:
The patent merges depth maps from multiple sources (3D laser scanners and photogrammetry devices) by converting images to depth maps in 2D space and combining them using weighted averaging. This approach integrates data from different sources to achieve more complete 3D coverage while managing noise through the merging process.
Solution Approach 2:
The patent converts the harmful effect of noise and outliers into a benefit by using weighted averaging where reliable pixels from either source can compensate for noisy pixels from the other source. The correction vectors adjust pixel positions to eliminate outliers while maintaining valuable information from both sources.
2Quantity of substance
If depth data from multiple sources are fused using existing methods, then more data points are obtained, but computational complexity increases
Solution Approach 1:
The patent changes the processing dimension from traditional 3D point cloud operations to 2D depth map operations. By converting images to depth maps and performing merging in 2D space, the computational complexity is reduced while still achieving complete 3D coverage through the correction vector approach.
3Reliability
If depth maps from different imaging devices are merged, then noise is reduced, but resolution may be lost
Solution Approach 1:
The patent applies local quality by treating pixels differently based on their reliability. High-resolution pixels from either source are identified and used to correct lower-resolution pixels, ensuring that the final merged depth map maintains the highest possible resolution locally while reducing noise through the combination of multiple sources.
Data Source
AI summary
A method includes capturing images of an object in a three dimensional (3D) space and converting a first image into a first depth map having first pixels and converting a second image into a second depth map having second pixels, at least one of the second pixels overlapping at least one of the first pixels. The method further includes identifying from the first depth map a first pixel that overlaps a second pixel from the second depth map, the second pixel representing a correct position of the first pixel and the second pixel in the 3D space. The method further includes determining a correction vector for a position of the first pixel, determining adjusted positions of the first pixels using the correction vector, determining an adjusted first depth map with the adjusted positions of first pixels, and merging the second depth map with the adjusted first depth map.