Active Stereo Depth Prediction Using Coarse Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing depth camera technologies face challenges in computing accurate depth maps, especially in textureless regions, occlusions, repetitive patterns, thin structures, and reflective surfaces, while also requiring significant processing power and memory.
Innovation Solution
The method involves using an active stereo depth camera system that projects an illumination pattern, captures depth images with an infrared stereo camera pair, and processes these images to estimate a depth map. This includes downsampling and matching stereo images to generate a coarse disparity map, upsampling it to full resolution, and refining the results to handle occlusions and improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional stereo triangulation is used to compute depth maps, then depth accuracy can be achieved, but processing power and memory requirements become excessively high
Solution Approach 1:
The patent divides the depth computation process into multiple stages: coarse depth map generation using downsampled images, followed by refinement stages that progressively improve accuracy. This segmentation allows the system to achieve high precision without requiring full computational resources at every stage, thus resolving the contradiction between depth map accuracy and processing power requirements.
Solution Approach 2:
The patent performs preliminary downsampling of stereo images before depth computation, and generates a coarse depth map first. This preliminary action reduces the data volume and computational complexity before the refinement stages, enabling accurate depth maps to be computed with reduced processing power and memory requirements.
2Measurement precision
If traditional stereo triangulation is used to compute depth maps, then depth information can be obtained, but memory requirements exceed computational budget
Solution Approach 1:
The patent segments the depth computation into coarse and refinement stages, where each stage processes data at appropriate resolution levels. This prevents the need to load and process full-resolution data throughout the entire computation pipeline, thereby reducing memory requirements while maintaining depth map accuracy.
Solution Approach 2:
The patent performs preliminary downsampling of input images and generates coarse depth maps before refinement. This preliminary action significantly reduces the memory footprint of intermediate data structures, allowing the system to compute accurate depth maps within available memory constraints.
3Productivity
If coarse matching is used to reduce computational demands, then processing efficiency improves, but depth accuracy may deteriorate
Solution Approach 1:
The patent uses coarse matching on downsampled images as a preliminary step to generate an initial depth map quickly. This preliminary coarse matching provides a good starting point that guides subsequent refinement stages, ensuring both processing efficiency and final depth accuracy are achieved.
Solution Approach 2:
The patent implements refinement stages that use the coarse depth map as input and iteratively improve accuracy by comparing with original high-resolution images. This feedback mechanism ensures that the initial coarse matching results are corrected and refined to achieve high depth map accuracy while maintaining processing efficiency.
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
The approach enables efficient and accurate estimation of depth maps, supporting location-based functionalities like augmented reality, virtual reality, and simultaneous localization and mapping, while reducing computational demands.
Implementation Method 1
captures depth images with an infrared stereo camera pair
Implementation Method 2
captures depth images with an infrared stereo camera pair
Data Source
AI summary
An electronic device estimates a depth map of an environment based on matching reduced-resolution stereo depth images captured by depth cameras to generate a coarse disparity (depth) map. The electronic device downsamples depth images captured by the depth cameras and matches sections of the reduced-resolution images to each other to generate a coarse depth map. The electronic device upsamples the coarse depth map to a higher resolution and refines the upsampled depth map to generate a high-resolution depth map to support location-based functionality.