Multi-Modal Impurity Detection Using Polarization and Depth Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing technologies for impurity detection in industrial settings, such as in coal processing, face challenges in accurately distinguishing impurities from homogeneous backgrounds due to limitations in RGB and RGB-D image processing algorithms, which can lead to system failures and significant losses.
Innovation Solution
A method and system that utilize RGB-P images captured with a polarization camera, pre-processing to remove motion blur, and graph-based spectral clustering to identify impurities by generating a Degree of Polarization (DoP) map and constructing joint graphs from RGB, RGB-D, and RGB-P images, merging their spectra to enhance segmentation and classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If RGB and RGB-D image processing algorithms are used for impurity detection, then the detection system can be implemented, but the accuracy in distinguishing impurities from homogeneous backgrounds is insufficient
Solution Approach 1:
The patent introduces polarization dimension to the traditional RGB and depth imaging modalities. By capturing polarization information at multiple angles and computing Degree of Polarization (DoP) maps, the system adds a new dimensional characteristic that enables differentiation of impurities from homogeneous backgrounds based on their distinct polarization properties, thereby resolving the limitation of conventional imaging algorithms
Solution Approach 2:
The patent fuses multiple imaging modalities (RGB, depth, and polarization) into a composite detection system. The joint graph construction integrates features from all three modalities, creating a composite representation that leverages the complementary strengths of each modality to achieve superior impurity detection accuracy and reliability
2Measurement precision
If manual inspection is used for material inspection, then impurity detection can be performed, but the process is cumbersome and error-prone
Solution Approach 1:
The patent replaces manual mechanical inspection with an automated multi-modal imaging system. The system uses polarization cameras, depth sensors, and RGB cameras coupled with automated image processing algorithms to perform impurity detection, eliminating the need for manual inspection while achieving higher accuracy and efficiency simultaneously
3Productivity
If single-modal image processing is used, then the processing speed is maintained, but the ability to differentiate native material and impurities is limited
Solution Approach 1:
The patent merges multiple imaging modalities (RGB, depth, and polarization) into a unified detection framework. The joint graph construction combines features from all modalities, allowing the system to maintain processing speed through efficient feature fusion while achieving superior material differentiation accuracy through the complementary information provided by each modality
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 significantly improves impurity detection accuracy by effectively differentiating between native materials and impurities, achieving higher precision and recall rates compared to traditional methods, with the joint spectral clustering method achieving an overall accuracy of 94.7% in identifying foreign objects.
Implementation Method 1
The RGB-P image is captured when an unpolarized light from a controlled light source falls on an object of interest, and reflected light, which becomes polarized is captured using a polarization camera from multiple polarization filter angles and estimate Stokes vector parameters
Data Source
Figure 1
Figure 2a~2b
Figure 3a~3b
AI summary
Impurity material detection in industrial processing is important to ensure quality of output, and also to save industrial machines from wear and tear caused by such impurity materials. State of the art systems in this domain rely on background subtraction related approaches, which fail to identify the impurity materials correctly. The disclosure herein generally relates to image processing, and, more particularly, to a method and system for impurity detection using multi-modal image processing. This system uses a combination of polarization data, and at least one of a depth data and an RGB image data to perform the impurity material detection. The system uses a graph fusion based approach while processing the captured images to detect presence of the impurity material, and accordingly alert the user.