Embedded Machine Learning Self-Calibration Using Captured Display Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Calibrating machine learning systems embedded within dedicated hardware is challenging due to the need for inputs to be provided via the hardware, which can limit the calibration process.
Innovation Solution
An apparatus and method that enable a device with an embedded machine learning system to display and capture calibration images using an imaging device, allowing for self-calibration of the machine learning system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning systems are embedded within dedicated hardware, then the system integration and reliability are improved, but the calibration process becomes problematic and difficult to perform
Solution Approach 1:
The machine learning system performs self-calibration by using its own embedded imaging device to capture calibration images and process them through the embedded machine learning system. This eliminates the need for external calibration equipment or manual intervention, allowing the system to maintain its embedded reliability while performing calibration autonomously.
Solution Approach 2:
The embedded imaging device serves multiple functions: it captures calibration images for system calibration, and can potentially serve other system functions. This multi-functionality resolves the contradiction by enabling calibration capability within the embedded system without requiring separate dedicated calibration hardware.
2Measurement precision
If calibration requires inputs to be provided via hardware, then the calibration process can be performed, but the device complexity and calibration difficulty increase
Solution Approach 1:
The system uses its own imaging device and processing capabilities to perform calibration autonomously, eliminating the need for external calibration hardware or complex input mechanisms. The embedded machine learning system processes the captured calibration images to adjust its parameters, simplifying the overall calibration system while maintaining precision.
3Extent of automation
If the imaging device captures calibration images displayed on the same device, then self-calibration is enabled, but geometric distortion and capture accuracy problems arise
Solution Approach 1:
The system uses feedback mechanisms to correct geometric distortions. The imaging device captures calibration images, the machine learning system processes these images to detect distortions, and then adjusts the calibration parameters to compensate for the distortions. This feedback loop enables automated self-calibration while maintaining image capture accuracy.
Solution Approach 2:
The system modifies parameters such as geometric distortion correction, image processing algorithms, and calibration models to account for the specific capture conditions. By dynamically adjusting these parameters based on the captured images and device configuration, the system maintains measurement precision while enabling automated self-calibration.
Data Source
AI summary
Examples of the disclosure relate to apparatus, methods and computer programs for calibrating machine learning systems by using images captured by an imaging device. The apparatus can comprise means for: enabling at least one of a first device or second device to display one or more calibration images, the first device comprising an embedded machine learning system wherein the one or more calibration images are for calibrating the embedded machine learning system. The apparatus can also comprise means for controlling an imaging device to capture, at least part of, the one or more calibration images, wherein the imaging device is provided within the first device, and using the one or more calibration images captured by the imaging device to calibrate the embedded machine learning system of the first device.