GAN-Based Image Boundary Detection for Semiconductor Elements
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting image boundaries in semiconductor elements using transmission electron microscopy are subjective and prone to noise, making accurate boundary analysis challenging.
Innovation Solution
A generative adversarial network (GAN) model is trained with image pictures to automatically detect image boundaries, utilizing a generator and discriminator network, with loss parameters adjusted to enhance accuracy and reduce manual errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If subjective judgment by user is used to measure grain boundaries, then measurement can be performed, but measurement precision deteriorates due to unclear boundaries and noise
Solution Approach 1:
The patent replaces the manual mechanical measurement process with an automated image processing system. The processor automatically detects grain boundaries by comparing images taken at different focal depths, eliminating the need for subjective human judgment and improving measurement precision through objective algorithmic analysis.
Solution Approach 2:
The patent introduces an intermediary processing system that captures images at multiple focal depths and uses image comparison algorithms as a mediator between the raw image data and the final grain boundary measurement. This intermediary process filters out noise and enhances boundary clarity before measurement.
2Productivity
If traditional image processing is used, then processing can be performed, but productivity deteriorates due to inability to handle noisy boundaries
Solution Approach 1:
The patent performs preliminary image capture at multiple focal depths before the actual boundary detection. By pre-acquiring a series of images with different focus settings, the system prepares enhanced data that makes subsequent boundary detection more reliable and faster, as the comparison process can directly identify boundaries without struggling with noise.
Solution Approach 2:
The patent employs dynamic image comparison where the processor iteratively analyzes images at different focal planes. The system dynamically adjusts by comparing multiple focal depth images and selecting the clearest boundary information, adapting to varying noise conditions in different regions of the image.
Data Source
AI summary
A method for detecting an image boundary using a generative adversarial network (GAN) model is provided. The GAN model includes a generator network and a discriminator network. The method includes the following steps. The GAN model is trained using a plurality of image pictures to obtain a weight value, an average value of a generator network loss parameter, and an average value of a discriminator network loss parameter. A generator network is updated with a first loss parameter, wherein the first loss parameter is a weighted sum of the average value of the generator network loss parameter and the average value of the discriminator network loss parameter. The discriminator network is updated with a second loss parameter, wherein the second loss parameter is the average value of the discriminator network loss parameter.