Thermal Image Interpolation Correction for 3D Printing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Low-resolution thermal cameras in 3D printing systems provide limited insight into the printing process, making it difficult to detect defects and anomalies, and the resulting high-resolution images often contain undesirable gradients and interpolation artifacts that hinder accurate defect detection.
Innovation Solution
A system comprising a camera, interpolation engine, and correction engine that captures low-resolution thermal images, upsamples and interpolates them, and enhances fine details using a machine learning model to produce accurate, artifact-free thermal images, allowing for improved defect detection and process optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If low-resolution thermal images are upsampled and interpolated to produce high-resolution images, then image resolution is improved, but interpolation artifacts and gradients are introduced that reduce image quality
Solution Approach 1:
A machine learning model is introduced as an intermediary between the low-resolution thermal image and the final high-resolution output. The model learns the complex mapping relationship and generates high-resolution images without traditional interpolation artifacts, effectively mediating the resolution enhancement process while preserving image quality.
Solution Approach 2:
The system changes the parameter space by training a machine learning model on pairs of low-resolution and high-resolution thermal images. The model learns optimal transformation parameters and features that enable resolution enhancement while avoiding interpolation artifacts, fundamentally changing how resolution upscaling is achieved.
2Ease of manufacture
If a low-resolution thermal camera is used in the 3D printing system, then device cost and robustness are improved, but defect detection capability deteriorates
Solution Approach 1:
The machine learning model creates a virtual copy or representation of what a high-resolution thermal image would look like, based on the low-resolution input. This computational copy enables defect detection capability equivalent to high-resolution cameras without the associated cost and complexity.
Solution Approach 2:
The system replaces the mechanical solution of using a physically high-resolution thermal camera with a computational approach using machine learning. The ML model substitutes for the physical resolution enhancement, eliminating the need for expensive high-resolution hardware while achieving the same defect detection capability.
3Productivity
If traditional interpolation methods are used to enhance thermal images, then processing speed is improved, but image accuracy deteriorates due to artifacts
Solution Approach 1:
The machine learning model is trained in advance on large datasets of paired thermal images. This preliminary training phase enables the model to perform real-time enhancement with both high speed and high accuracy, as the complex computational work is already done during training rather than during actual image processing.
Data Source
AI summary
An example three-dimensional (3D) printer may include a camera to capture a low-resolution thermal image of a build material bed. The 3D printer may include an interpolation engine to generate an interpolated thermal image based on the low-resolution thermal image. The 3D printer may also include a correction engine to enhance fine details of the interpolated thermal image without distorting thermal values from portions of the interpolated thermal image without fine details to produce an enhanced thermal image.