Textile Augmented Reality Pattern Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional augmented reality systems struggle to detect patterns on non-flat surfaces, accommodate surface variations and distortions, track moving patterns, and recognize patterns in woven and printed materials due to low contrast, poor resolution, and surface reflections.
Innovation Solution
Integration of textiles with augmented reality systems that incorporate aperiodic markers within ornamental designs, enhancing pattern detection by improving contrast and resolution, allowing for the use of textiles with unique and unpredictable patterns that can trigger audio-visual content display.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional image recognition capabilities are used, then the system can detect patterns on flat surfaces, but it cannot detect patterns on non-flat surfaces or accommodate surface variations and distortions
Solution Approach 1:
The system changes the parameters of pattern detection by using deep neural networks that can process distorted and varied patterns. The network is trained to recognize patterns under different transformations (wrinkles, stretches, folds) by adjusting detection parameters to be invariant to these changes, enabling reliable detection on non-flat surfaces.
Solution Approach 2:
The patent combines conventional image recognition with deep learning techniques to create a hybrid system. The deep neural network processes the image data to identify patterns while accommodating surface variations, effectively compositeing traditional methods with advanced AI capabilities to handle varied surface geometries.
2Adaptability or versatility
If conventional image recognition capabilities are used, then the system can detect patterns on flat surfaces, but it cannot track displacement of detectable environmental patterns
Solution Approach 1:
The system implements feedback mechanisms where the deep neural network continuously processes image data to track pattern displacement. The network receives feedback about pattern positions and adjusts its tracking accordingly, enabling precise measurement of pattern movement over time while maintaining adaptability to various surface conditions.
3Ease of manufacture
If conventional textile manufacturing techniques are used, then repeating patterns can be created, but spatially aperiodic designs with unpredictable variations cannot be generated
Solution Approach 1:
The patent uses digital copying and rendering techniques to create textile patterns. Instead of relying solely on traditional manufacturing processes, the system digitally generates and reproduces both repeating and aperiodic patterns, allowing for unlimited design variety while maintaining manufacturing feasibility through digital printing and rendering technologies.
Solution Approach 2:
The system introduces dynamic design capabilities where patterns can be generated with unpredictable variations and adaptability. The deep learning models enable dynamic pattern creation that can vary spatially and temporally, transforming static textile manufacturing into a dynamic process capable of creating unique, aperiodic designs.
4Difficulty of detecting and measuring
If conventional augmented reality systems are used, then graphical overlays can be projected onto flat surfaces, but the system cannot detect patterns in woven and printed materials due to low contrast and surface reflections
Solution Approach 1:
The system changes the detection parameters by using deep neural networks that are specifically trained to handle low-contrast and reflective surfaces. The network adjusts its detection thresholds and feature extraction parameters to accommodate the optical characteristics of woven and printed materials, significantly improving pattern detectability in these challenging conditions.
Solution Approach 2:
The deep neural network acts as an intermediary between the camera and pattern recognition. It processes the raw image data through multiple layers, filtering out noise and enhancing subtle patterns that conventional systems miss. This intermediary processing enables accurate pattern recognition in textured, low-contrast, and reflective textile surfaces.
Data Source
AI summary
This disclosure relates generally to augmented reality, and more particularly to augmented reality systems and methods using textiles. In one embodiment, a processor-implemented textile-based augmented reality method is disclosed. The method may comprise capturing, via one or more hardware processors, a video frame including a depiction of an aperiodic marker included in an ornamental design of a textile fabric. Via the one or more hardware processors, the presence of the marker may be identified using one or more image-processing marker detection techniques. The identified marker may be associated with one or more audio-visual content files. Finally, data from the one or more audio-visual content files may be displayed as part of an augmented reality presentation.