Neural Network Drift Correction for Topographic Image Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D topography imaging systems, such as AFMs and profilometers, often suffer from directional drifts in recorded topographic images due to system imperfections or sample misalignments, making manual correction time-consuming and operator-dependent.
Innovation Solution
A device and method utilizing a trainable neural network to normalize and correct directional drifts in topographic image data, allowing for automated, efficient, and scalable correction of vertical or horizontal drifts by denormalizing processed values back to the original range.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual correction of directional drift is performed by an operator, then the correction can be adjusted based on experience and judgment, but the process is time-consuming and varies from image to image
Solution Approach 1:
The patent replaces the manual mechanical correction process with an automated computational system. A neural network model processes the topographic image data to automatically detect and correct directional drifts, substituting human operator actions with an automated algorithm that consistently identifies drift patterns and applies corrections without manual intervention.
Solution Approach 2:
The correction system performs self-service by automatically analyzing the topographic image data, identifying drift characteristics, and applying corrections without requiring external human input. The neural network independently processes the data through normalization, drift detection, correction calculation, and validation steps, making the system self-sufficient for the correction task.
2Adaptability or versatility
If manual correction is performed, then operator judgment can be applied, but the correction process lacks consistency and scalability across different images
Solution Approach 1:
The patent replaces manual correction operations with an automated neural network-based system that processes topographic image data. The system automatically normalizes values, detects directional drifts, calculates corrections, and validates results without human intervention, achieving full automation while maintaining correction quality through algorithmic consistency.
Solution Approach 2:
The system changes the operational parameters from manual human judgment to automated computational processing. By transforming the correction process into a series of defined computational steps (normalization, drift detection, correction calculation, validation), the system achieves scalability and consistency across different images while maintaining adaptability through the neural network's ability to learn from data patterns.
3Productivity
If automated correction using neural network is implemented, then processing speed and consistency are improved, but the system complexity increases
Solution Approach 1:
The patent segments the correction process into distinct modular steps: data normalization, drift detection, correction calculation, and validation. Each step is handled by specific computational routines within the neural network framework, allowing the complex task to be broken down into manageable, independently optimizable components that collectively achieve high processing speed and consistency.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a device (100) for correcting directional drifts in topographic image data, comprising a receiver (103) configured to receive the topographic image data in a raw data format, wherein the topographic image data comprises values representing spatial coordinates of a topography, and wherein the values are in a first range, a processor (105) configured to normalize said values of the topographic image data to a second range, and a trainable neural network (107) configured to receive the topographic image data with normalized values and to remove a directional drift of the normalized values, wherein, following removal of the directional drift by the neural network (107), the processor (105) is configured to denormalize the values, in particular to the first range, to generate corrected topographic image data.