Continuous Surface And Depth Estimation for Low-Power AR Vision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques for determining depth and surface normals in augmented reality (AR) devices are computationally expensive and require special hardware like depth sensors, consuming additional power, which is not suitable for compact AR devices.
Innovation Solution
A continuous surface and depth estimation system using stereo vision within a predetermined window, limiting its application to a sub-portion of captured images, and employing methods like RANSAC and ray casting to estimate depth and surface normals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If depth sensors and special hardware are used to determine depth and surface normals, then measurement precision is improved, but use of energy and device complexity increase
Solution Approach 1:
The patent replaces specialized depth sensing hardware with a computational approach using standard camera images. By applying continuous surface and depth estimation algorithms to regular 2D images, the system derives depth information without requiring additional sensors, thereby reducing power consumption and device complexity while maintaining depth measurement capability
Solution Approach 2:
The patent creates a virtual depth map by computing depth values from 2D image data rather than directly capturing it with depth sensors. This computational copy of depth information allows the system to function with standard imaging hardware, avoiding the need for power-intensive specialized sensors
2Measurement precision
If comprehensive depth and surface normal determination is performed across entire images, then measurement precision is improved, but productivity decreases due to computational expense
Solution Approach 1:
The patent divides the image processing task into discrete pixel-level operations where depth and surface normal are estimated independently for each pixel or small regions. This segmentation allows for optimized computation where only necessary calculations are performed, improving processing speed while maintaining estimation accuracy across the entire image
Solution Approach 2:
The patent implements an iterative estimation process that can terminate early when sufficient precision is achieved, or process only necessary regions of interest. This partial action approach prevents unnecessary computational overhead while ensuring adequate precision for the application requirements
3Measurement precision
If continuous surface and depth estimation is performed on full images, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent replaces complex specialized depth sensing hardware with computational algorithms running on standard processing units. By using mathematical models to estimate depth from 2D images, the system achieves depth mapping capability without requiring additional sensors or specialized hardware components
Solution Approach 2:
The patent makes standard camera hardware perform multiple functions by using it both for 2D image capture and for generating depth information through computational estimation. This multi-functionality eliminates the need for separate depth sensing hardware, reducing device complexity while maintaining depth measurement capability
Data Source
AI summary
Disclosed are systems, methods, and non-transitory computer-readable media for continuous surface and depth estimation. A continuous surface and depth estimation system determines the depth and surface normal of physical objects by using stereo vision limited within a predetermined window.