Food Classification via Optical Feature Extraction and Confidence Thresholding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for classifying food objects, especially those with grown or irregular structures, face challenges in throughput and reliability due to manual classification, and existing automatic systems are limited in handling diverse food types and may require additional identification carriers or have high error rates.
Innovation Solution
A device with an image acquisition unit, evaluation unit, and data input/output units that captures optical and depth data to extract feature values, assigns classes based on mathematical algorithms, and uses a confidence threshold to differentiate between core and edge areas for automatic classification, optimizing recognition rates and reducing manual intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual classification is used, then classification accuracy can be maintained, but throughput is limited and costs increase
Solution Approach 1:
The patent replaces the manual mechanical classification system with an automated optical imaging and image analysis system. The device captures images of food objects and automatically extracts features to determine classification, eliminating the need for manual visual inspection and significantly increasing throughput while maintaining accuracy.
Solution Approach 2:
The classification system performs self-service by automatically analyzing images and determining food object classifications without requiring human intervention. The system independently extracts features from images, compares them against reference data, and generates classifications, making the process autonomous and highly efficient.
2Productivity
If automatic classification using identification carriers is used, then throughput increases, but additional costs and manual preparation are required
Solution Approach 1:
The patent extracts and utilizes naturally occurring visual features of food objects directly from their appearance, eliminating the need for external identification carriers. By focusing on inherent characteristics like color, shape, and texture visible in images, the system avoids adding complex identification hardware or manual carrier attachment steps.
Solution Approach 2:
The image analysis system serves multiple functions: it captures images, extracts features, compares against reference data, and performs classification all in one integrated process. This universal approach handles diverse food objects without requiring different identification methods for different product types.
3Productivity
If automatic classification of industrial goods is used, then throughput increases, but reliability decreases for irregular structures
Solution Approach 1:
The patent changes the analysis parameters by using multiple image capture conditions (different lighting, angles, or processing stages) and analyzing multiple feature types (color, shape, texture). This multi-parameter approach enables reliable classification of irregular food structures that vary in appearance, overcoming the limitations of single-parameter systems.
Solution Approach 2:
The system performs partial classification by identifying and analyzing only the most relevant features for each food object type, rather than requiring complete characterization. This selective feature extraction maintains high reliability while processing throughput, focusing computational resources on discriminative features.
4Productivity
If higher throughput is achieved, then productivity increases, but misclassification risk increases due to human error
Solution Approach 1:
The classification system incorporates feedback mechanisms by comparing extracted features against stored reference data and using the comparison results to refine classifications. The system learns from previous classifications and can adjust its analysis, reducing errors and maintaining high accuracy even at increased throughput levels.
Data Source
AI summary
The invention relates to a device for the classification of a food item (1) of an organic or irregular structure comprising an image capture unit (2), an evaluation unit (3), a data input unit (4) and a data output unit (5), wherein the evaluation unit (3) is connected to the image capture unit (2), the data input unit (3) and the data output unit (4), and wherein the food item (1) can be captured as optical data by means of the image capture unit (2) and the optical data can be provided to the evaluation unit (3) in a transferable manner and wherein characteristic values for the food item (1) can be extracted from the optical data by means of the evaluation unit (3), wherein the characteristic values can be assembled into a characteristic value tuple for the food item (1) and wherein the characteristic value tuple of the food item (1) can be automatically assigned to a characteristic value tuple range, wherein the characteristic value tuple range is formed by one or a plurality of characteristic value tuples and wherein a class can be assigned to said characteristic value tuple range and wherein the class can be assigned to the characteristic value tuple range by means of the data input unit (4). The device is particularly characterised in that the characteristic value tuple range can be divided up into a core range and a marginal range by means of the evaluation unit (3) and in that the assignment of the characteristic value tuple for the food item (1) can be made separately by core range and marginal range. The invention further relates to a method, particularly a method that can be carried out with the device, for the classification of a food item (1) of organic or irregular structure.