Silicon Ingot Diameter Measurement via Probabilistic Image Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for measuring the diameter of a growing silicon ingot during the Czochralski process are inaccurate due to reliance on partial scan images, assumptions about pixel locations and brightness, and are prone to noise from reflections and camera distortions, leading to errors in growth rate control and intrinsic properties of the ingot.
Innovation Solution
A method and apparatus that uses all meniscus pixels from captured images, applies a probabilistic filter to eliminate improbable pixels, and corrects for camera perspective distortion to accurately calculate the ingot diameter, employing high-pass filtering, auto-thresholding, and statistical analysis with weight factors to attenuate noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional methods use partial scan images and assumptions about pixel locations to measure diameter, then the measurement process is simplified, but measurement precision deteriorates due to missing pixels and errors
Solution Approach 1:
The patent extracts and uses ALL meniscus pixels from the captured image rather than relying on partial scan images. By extracting every pixel that represents the meniscus interface between the silicon ingot and melt, the system eliminates the need for assumptions about pixel locations and achieves complete utilization of available measurement data, thereby resolving the contradiction between simplicity and precision
Solution Approach 2:
The patent replaces conventional image processing algorithms with a probabilistic filter that statistically determines which pixels represent the meniscus. This substitution of the measurement approach allows the system to accurately identify meniscus pixels without relying on geometric assumptions or partial scanning, achieving high precision while maintaining computational efficiency
2Device complexity
If conventional methods rely on assumptions about brightness levels and pixel locations, then the processing algorithm is simpler, but reliability deteriorates when system setup is incorrect or brightness changes
Solution Approach 1:
The patent employs a probabilistic filter that dynamically adapts to changing conditions during the measurement process. Rather than relying on fixed assumptions about brightness levels or pixel locations, the algorithm continuously evaluates the probability that each pixel represents the meniscus based on the actual image data, allowing it to maintain reliability even when system setup varies or brightness conditions change
Solution Approach 2:
The patent changes the approach from using fixed parameters (assumed brightness levels, fixed pixel locations) to using probabilistic parameters that are determined from the actual image data. This parameter transformation allows the system to adapt to varying conditions and maintain reliability without requiring precise system setup or stable brightness conditions
3Productivity
If conventional methods use search algorithms to locate meniscus pixels, then the measurement can be performed, but measurement precision deteriorates due to noise from reflections and camera distortions
Solution Approach 1:
The patent replaces conventional search algorithms with a probabilistic filter that statistically identifies meniscus pixels. This substitution transforms the measurement approach from actively searching for pixels (which is susceptible to noise and distortions) to probabilistically determining which pixels represent the meniscus based on the actual image characteristics, thereby achieving higher precision while maintaining measurement capability
Solution Approach 2:
The probabilistic filter acts as an intermediary between the raw image data and the diameter measurement. It processes the image data through statistical evaluation, filtering out noise from reflections and camera distortions while identifying the true meniscus pixels. This intermediary processing step protects the measurement from the effects of noise and distortions present in the original image
Data Source
AI summary
The present invention provides a method and apparatus for controlling the growth of a silicon ingot in which the diameter of the growing silicon ingot can be accurately measured. A camera captures an image of the interface ring between the growing silicon ingot and the silicon melt. An image processor extracts local intensity maxima from the captured image, which are then digitized into an image data which comprises attributes of the pixels forming the local intensity maxima. An analyzer statistically analyzes the image data to derive parameters of an equation statistically simulating the interface ring. A probabilistic filter conducts the statistical analysis on the equation in which the respective pixels are weighted by their weight factors. The weight factor functions to attenuate the effect of noises caused by pixels which do not represent the interface ring. The statistical analysis may be repeated, using the renewed parameters, to progressively attenuate the effect of the noises to thereby obtain a satisfactorily accurate diameter of the silicon ingot.