Edge Defect Detection in Ophthalmic Lenses Using Image Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current contact lens defect inspection systems face inefficiencies due to the large size of high-resolution images, which negatively impact neural network performance and productivity, despite advancements in hardware, as additional algorithms are needed for stringent defect detection, leading to increased processing time and potential data degradation.
Innovation Solution
A computer-implemented process that optimizes high-resolution images by identifying and discarding irrelevant pixel data, rearranging the image to a square format with overlapping segments for faster processing, and applying Generative Adversarial Networks to enhance defect feature extraction and classification without image compression, utilizing high-performance CPUs, GPUs, and deep learning modules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high resolution images are used for defect inspection, then detection accuracy is improved, but processing time increases and productivity decreases
Solution Approach 1:
The patent divides the high-resolution image into multiple smaller image segments of predetermined sizes. These segments are then processed independently by deep learning modules, allowing parallel processing that maintains detection accuracy while reducing overall processing time and increasing inspection throughput.
2Speed
If image compression algorithms are applied to reduce image size, then processing speed is improved, but inspection quality deteriorates due to degradation of critical defect data
Solution Approach 1:
The patent extracts only the critical regions and features from the high-resolution image using edge detection and feature extraction algorithms. By taking out only the essential defect-related information rather than compressing the entire image, the system maintains inspection quality while reducing the data volume for processing.
3Measurement precision
If additional image processing algorithms are implemented for stringent defect detection, then inspection accuracy is improved, but inspection time increases
Solution Approach 1:
The patent performs preliminary image processing actions including edge detection, feature extraction, and image segmentation before the actual defect inspection. By preparing and pre-processing the image in advance to highlight critical areas, the subsequent inspection process is accelerated while maintaining high accuracy through the use of deep learning modules trained on the pre-processed data.
Data Source
AI summary
The invention is a computer implemented process directed towards a Deep learning neural network architecture to create object detection model using high resolution images. More specifically, the invention is directed towards enhancing the classification accuracy and reliability of edge inspection in contact lenses. The invention is a Computer implemented process to represent a software architecture comprising software components and their inter dependencies that represents the core functional modules of an application. The system and method of the invention captures a high resolution image, transforms the circular edge of the lens to a Horizontal line representing the circular edge, restricting the image size by eliminating the pixel information around the edge, dividing the horizontal edge image into overlapping portions, and stacking the extracted images vertically to form a single high-resolution image that is ideal to be processed and analysed by Convolution Neural networks after augmenting the original image dataset with new images generated by Generative Adversarial Networks, to enable accurate classification of the defects.