TEM Image Classification for Automatic Electron Beam Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing transmission electron microscopes require manual adjustment of the optical system based on user experience, leading to suboptimal image display for system adjustment.
Innovation Solution
Implement an illumination system, imaging system, detector, and arithmetic unit that classify images based on average pixel values and standard deviations, enabling automatic adjustment of the optical system without relying on user experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual image classification based on user experience is used, then the optical system adjustment can be performed, but the adjustment precision is suboptimal and relies on operator skill
Solution Approach 1:
The patent replaces the manual mechanical process of image classification (based on user experience) with an automated computational system. The arithmetic unit calculates objective metrics (average value and standard deviation of pixel values) to automatically classify images, substituting human judgment with algorithmic processing. This resolves the contradiction by achieving higher precision through automation while maintaining ease of operation through the system's autonomous decision-making capability.
Solution Approach 2:
The system performs self-service by automatically classifying images without requiring user intervention or expertise. The arithmetic unit autonomously analyzes pixel values, compares them against predetermined thresholds, and determines the appropriate image type (beam image, vacuum image, or specimen image). This eliminates dependency on operator skill while maintaining operational simplicity.
2Extent of automation
If automatic image classification based on pixel statistics is implemented, then the extent of automation is improved, but the device complexity increases due to additional arithmetic processing
Solution Approach 1:
The patent transforms the complex task of image classification into simple parameter comparisons. Instead of analyzing complex image features, the system changes the approach to use only two statistical parameters (average pixel value and standard deviation) that can be computed efficiently and compared against predetermined thresholds. This reduces arithmetic processing complexity while achieving full automation.
Solution Approach 2:
The patent applies local quality analysis by examining specific regions of interest within the image (first pixels and second pixels) rather than analyzing the entire image globally. This localized statistical analysis simplifies the computational burden while maintaining classification accuracy, as only relevant pixel regions need to be processed.
3Reliability
If standard deviation calculation is used for image classification, then the reliability of classification is improved, but the loss of time increases due to additional computational steps
Solution Approach 1:
The patent performs preliminary action by pre-establishing classification thresholds and criteria before actual image processing. The system has predetermined values for average pixel intensity and standard deviation thresholds that guide classification decisions. This preparation work is done in advance, allowing rapid real-time classification without extensive computational analysis during operation, thus reducing processing time while maintaining reliability.
Data Source
AI summary
A transmission electron microscope includes an illumination system that illuminates a specimen with an electron beam; an imaging system that forms an image using the electron beam that has been transmitted through the specimen; a detector that captures the image formed by the imaging system; and an arithmetic unit that classifies the image captured by the detector. The arithmetic unit: acquires a first image captured by the detector; calculates an average value and a standard deviation of first pixel values of a plurality of first pixels composing the first image; acquires information of an average value and a standard deviation of second pixel values of a plurality of second pixels composing a second image captured in a state in which no electron beam is detected; and classifies the first image based on the average value and the standard deviation of the first pixel values and second pixel values.