Single NIR Camera Depth Mapping With Dot Pattern Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional depth estimation systems requiring multiple cameras for calibration are prone to calibration errors, leading to inaccurate depth calculations, and increase complexity and resource usage.
Innovation Solution
A system utilizing a single near-infrared (NIR) camera and dot illuminator projects uniform NIR dots on a scene, generating a dot image, which is processed by a machine-learned model to estimate depth without calibration, leveraging foreshortening effects and local disparities in the dot pattern.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple cameras are used for depth estimation, then measurement precision is improved, but device complexity increases and calibration errors occur
Solution Approach 1:
The patent combines the depth estimation function that previously required multiple cameras into a single camera system. The single camera captures images of a structured light pattern (dot grid) projected by an illuminator, and a processing system performs depth calculations using algorithms that analyze the distortion and displacement of the dot pattern in the captured image, eliminating the need for multiple cameras while maintaining depth estimation capability
Solution Approach 2:
The patent replaces the mechanical/optical system of multiple cameras with a single camera combined with computational processing. Instead of using stereo vision from multiple camera viewpoints, the system uses a single camera to capture a structured light pattern and employs image processing algorithms to derive depth information from the pattern distortion, substituting hardware complexity with computational methods
2Measurement precision
If multiple cameras are used for depth estimation, then measurement precision is improved, but calibration reliability deteriorates due to calibration errors
Solution Approach 1:
The patent extracts and eliminates the calibration requirement from the depth estimation system. By using a single camera to capture a known structured light pattern (dot grid) with defined geometric relationships, the system removes the need for inter-camera calibration. The depth calculation relies on the known geometry of the projected pattern and its distortion in the image, rather than on calibration between multiple camera coordinate systems
Solution Approach 2:
The system performs self-calibration through the use of a known structured light pattern. The dot grid illuminator projects a pattern with predetermined geometric relationships, and the single camera captures this pattern under various viewing conditions. The processing system uses the known geometry of the projected pattern as a reference to automatically determine depth without requiring external calibration procedures or multiple camera synchronizations
3Device complexity
If a single camera is used for depth estimation, then device complexity is reduced, but measurement precision may deteriorate
Solution Approach 1:
The patent introduces a structured light pattern (dot grid) as an intermediary between the single camera and the depth estimation process. This projected pattern serves as a known geometric reference that, when captured by the single camera, provides sufficient information for depth calculation. The pattern acts as a mediator that transforms the limited information from a single camera view into rich depth data through analysis of pattern distortion and displacement
Solution Approach 2:
The patent changes the parameter space by projecting a structured light pattern with specific geometric properties (dot grid with known spacing and arrangement). This transforms the problem from direct depth measurement to pattern analysis, where depth information is encoded in the distortion and displacement of the known pattern. The machine learning model is trained on these pattern variations to accurately predict depth from the captured images
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
This approach reduces complexity, resource usage, and eliminates the need for continuous camera calibration, while effectively preventing spoof attacks in authentication systems.
Implementation Method 1
projecting, using a dot illuminator, near-infrared (NIR) dots on the scene
Implementation Method 2
capturing, using a single NIR camera, the projected NIR dots on the scene
Data Source
AI summary
Provided are computing systems, methods, and platforms for using machine-learned models to generate a depth map. The operations can include projecting, using a dot illuminator, near-infrared (NIR) dots on the scene. The NIR dots can have a uniform pattern. Additionally, the operations can include capturing, using a single NIR camera, the projected NIR dots on the scene. Moreover, the operations can include generating a dot image based on the captured NIR dots on the scene. Furthermore, the operations can include processing the dot image with a machine-learned model to generate a depth map of the scene. Subsequently, the operations can further include evaluating the generated depth map of the scene and a ground truth depth map, and performing an action based on the evaluation.