Polarization Shadow Detection via Sequence Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing algorithms struggle to accurately distinguish between shadows and material object edges, leading to false positives and negatives due to the assumption that shadow boundaries are soft and object edges are sharp, which is not always the case in real-world scenarios.
Innovation Solution
A method and system that determine illumination flux conditions by generating and analyzing a sequence of images taken in different polarization directions, using polarization sequence vectors to differentiate between shadowed and lit areas based on color information and polarization characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional brightness boundary analysis is used to detect edges, then the detection can be implemented by algorithms, but the results are incorrect due to false positives and false negatives
Solution Approach 1:
The patent transitions from analyzing a single brightness image to analyzing a sequence of images captured at multiple polarization angles (0°, 45°, 90°, 135°). This dimensional expansion from 2D brightness to 4D polarization space (x, y, angle, intensity) enables differentiation between shadow boundaries and material edges by examining polarization state variations across multiple angular dimensions.
Solution Approach 2:
The patent changes the measurement parameter from单一的 brightness intensity to polarization angle and polarization state. By capturing images at different polarization angles and analyzing the polarization sequence vectors, the system transforms the detection parameter space to distinguish between shadow regions (which maintain consistent polarization) and material edges (which exhibit polarization discontinuities).
2Ease of operation
If shadow boundaries are assumed to be soft and object edges to be sharp, then conventional algorithms can be executed, but significant possibilities for false positives and false negatives exist
Solution Approach 1:
Instead of assuming shadow boundaries are soft and object edges are sharp, the patent inverts the approach by using polarization analysis to directly characterize boundary types. The method computes polarization sequence vectors and analyzes their properties to determine whether a boundary is a shadow boundary or a material edge, regardless of its sharpness or softness appearance in the brightness image.
Solution Approach 2:
The patent introduces polarization sequence vectors as an intermediary representation between the raw polarization images and the final shadow/edge classification. These vectors capture the polarization state evolution across different angles and serve as a mediator that enables reliable discrimination between shadow and material boundaries through quantitative analysis of polarization characteristics.
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 allows for accurate identification of shadowed and lit regions, enabling correct recognition of object edges and improving image processing accuracy by understanding the interplay between material and illumination components in images.
Implementation Method 1
generating and storing a sequence of images of the scene, each one of the sequence of images comprising an array of pixels and corresponding to the scene photographed in a preselected polarization direction, different from the polarization direction of other ones of the sequence of images
Data Source
AI summary
In a first exemplary embodiment of the present invention, an automated, computerized method is provided for determining an illumination flux condition in a scene. The method comprises the steps of generating and storing a sequence of images of the scene, each one of the sequence of images comprising an array of pixels and corresponding to the scene photographed in a preselected polarization direction, different from the polarization direction of other ones of the sequence of images, determining a polarization sequence vector for at least one pixel in the array, as a function of color information for the pixel in the array, among the sequence of images; and utilizing the polarization sequence vector to determine one of a shadowed and lit illumination condition for the at least one pixel.