Near-Infrared Spectral Sensor with Machine Learning Wavelength Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Industrial spectral sensors for near-infrared spectroscopy face challenges in speed and integration efficiency, as they require extensive measurement of full spectra for learning and often lack the ability to automatically select relevant wavelengths for faster, accurate object detection in operating modes.
Innovation Solution
A spectral sensor designed to operate in both learning and operating modes, where it determines a reduced second spectrum using machine learning, allowing for faster measurements and integration by selecting relevant wavelengths independently, potentially with external computing support, and incorporating a machine learning unit with a classifier and gating filter for efficient object recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the spectral sensor measures a full first spectrum with many wavelengths in learning mode, then the accuracy of object recognition is improved, but the measurement time increases and productivity decreases
Solution Approach 1:
The spectrum is segmented into two modes: a comprehensive first spectrum for learning mode and a reduced second spectrum for operating mode. The sensor selectively measures different spectral ranges based on the operational phase, allowing full spectral information to be captured during learning while using only essential wavelengths during operation, thus resolving the contradiction between accuracy and speed
Solution Approach 2:
The sensor performs preliminary measurements of the complete first spectrum during learning mode to gather comprehensive spectral data. Based on this preliminary action, the system pre-determines which wavelengths are most relevant for object recognition, enabling the subsequent operating mode to focus only on these critical wavelengths and achieve faster measurements without sacrificing accuracy
2Extent of automation
If the spectral sensor uses machine learning methods to determine relevant wavelengths, then the integration into industrial processes is simplified and automation is improved, but the device complexity increases
Solution Approach 1:
The spectral sensor incorporates a machine learning unit that enables the device to automatically determine relevant wavelengths independently during operation. This self-service capability allows the sensor to autonomously adapt to different objects and materials by selecting optimal wavelengths without external intervention, improving automation while the integrated design keeps complexity manageable
Solution Approach 2:
The spectral sensor is designed with multi-functionality, serving both as a measurement device and as a system that performs machine learning analysis. By integrating the machine learning unit directly into the sensor, it can handle multiple tasks including data acquisition, wavelength selection, and object recognition, reducing the need for separate external systems and simplifying overall integration into industrial processes
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly reduces measurement time in operating mode without compromising accuracy, enhancing throughput and simplifying integration into industrial processes by automating the selection of relevant wavelengths, enabling quick and reliable object detection.
Implementation Method 1
spectral sensor for near-infrared spectroscopy, which serves to distinguish and/or recognize objects and/or materials
Implementation Method 2
Materials usually differ in the wavelengths they absorb and can therefore be differentiated from one another
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a spectral sensor for near-infrared spectroscopy, which serves to distinguish and/or detect objects and/or materials, wherein the spectral sensor is configured to operate in a learning mode and in an operating mode, wherein the spectral sensor is configured to perform one or more measurements of intensity values of a first spectrum of wavelengths in the learning mode and at least one measurement of intensity values of a second spectrum of wavelengths in the operating mode, wherein the second spectrum is completely contained in the first spectrum but comprises fewer wavelengths than the first spectrum, wherein the spectral sensor is configured to determine the wavelengths of the second spectrum itself, wherein the spectral sensor is configured to select the wavelengths of the second spectrum using a machine learning method.