Polarization Surface Normal Imaging With Event-Based Vision Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing surface normal estimation methods using polarization information face challenges with speed vs. accuracy tradeoffs, high error rates, and susceptibility to noise, particularly in dynamic scenes, and struggle with non-Lambertian surfaces.
Innovation Solution
An apparatus and method utilizing a rotatable linear polarizer and event-based vision sensor (EVS) to detect polarization information as events, combined with a shape estimation processor to compute surface normals, enabling high-speed and accurate estimation of both static and dynamic scenes, including non-Lambertian surfaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional frame-based SfP with fixed polarizer angles is used, then polarization information can be captured at full resolution, but image capture speed is limited by frame rate and cannot handle dynamic scenes
Solution Approach 1:
The patent applies the Dynamics principle by replacing static, fixed-angle polarizers with a dynamically rotating linear polarizer. The polarizer rotates at controlled speeds (e.g., 300-1000 RPM) to capture polarization information at multiple angles within a single frame exposure period. This dynamic approach allows the system to capture full-resolution polarization data at video frame rates (30-60 FPS), simultaneously achieving high spatial resolution and fast capture speed for dynamic scenes.
2Speed
If mosaicking with micro-optical polarizers is used, then all polarization angles can be captured simultaneously, but spatial resolution is reduced to one quarter of original
Solution Approach 1:
The patent avoids the resolution loss of mosaicking by using a dynamic rotating polarizer instead of static micro-polarizer arrays. The single lens captures full-resolution light from the scene, and the rotating polarizer modulates the polarization angle over time. The EVS captures events at full spatial resolution for each polarization angle during the rotation, eliminating the 4x resolution penalty of mosaicking while maintaining simultaneous multi-angle capture capability.
Solution Approach 2:
The patent transitions from spatial multiplexing (mosaicking different polarizer angles at different pixel locations) to temporal multiplexing (rotating the polarizer to present different angles at different times). This dimensional change from space to time allows the system to capture all polarization angles at each pixel location sequentially during the exposure period, achieving full spatial resolution with temporal resolution sufficient for dynamic scenes.
3Measurement precision
If conventional cameras are used for SfP, then full resolution images can be captured, but captured light intensity is reduced by 50 percent leading to noise susceptibility
Solution Approach 1:
The patent replaces conventional frame-based cameras with an Event-based Vision Sensor (EVS). The EVS uses event-driven sampling where each pixel independently generates an event when light intensity changes exceed a threshold, rather than capturing complete frames at fixed intervals. This substitution allows the system to capture polarization information at full resolution without the 50% intensity loss penalty, as events are generated asynchronously based on actual light changes, improving signal-to-noise ratio and reducing noise susceptibility.
4Speed
If high frame rates are used to capture dynamic scenes, then speed is improved, but accuracy and error rates increase due to reduced light integration time
Solution Approach 1:
The patent implements continuous polarization angle sampling during the exposure period through polarizer rotation. Instead of capturing discrete frames at high speed with limited integration time, the system continuously modulates the polarization angle and the EVS continuously generates events as long as light changes occur. This continuous useful action allows sufficient light integration even at high effective frame rates, maintaining surface normal estimation accuracy while capturing dynamic scenes at video frame rates.
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
Enables fast and precise surface normal estimation with reduced data input, allowing for high-speed scanning and reconstruction of 3D shapes without compromising spatial resolution, particularly effective for non-Lambertian surfaces.
Implementation Method 1
a rotatable linear polarizer configured to pass light from the scene at a first rotation angle of the linear polarizer and to subsequently pass light from the scene at a second rotation angle of the linear polarizer
Implementation Method 2
an event-based vision sensor, EVS, configured to detect a first set of events associated with the passed light of the first rotation angle of the linear polarizer
Data Source
AI summary
The present disclosure relates to an apparatus for polarization-based surface normal imaging, the apparatus comprising a linear polarizer configured to rotate and to subsequently pass light from a scene, an event-based vision sensor. EVS, configured to detect a set of events of the scene based on the rotation angle of the linear polarizer, and a shape estimation processor configured to compute surface normal information of the scene based on the set of events and the corresponding rotation angle of the polarizer.