Predicting sensory evaluation of spinach freshness using machine learning model and digital images
Kento Koyama1, Marin Tanaka1, Byeong-Hyo Cho1
1Graduate School of Agricultural Science, Hokkaido University, Sapporo, Japan.
Plos One
|March 19, 2021
Summary
This study developed a smartphone-based image analysis technique to assess spinach freshness. Machine learning models accurately classified spinach freshness, offering a faster alternative to traditional sensory panel testing.
Area of Science:
- Agricultural Science
- Computer Science
- Food Science
Background:
- Consumer purchasing decisions for produce heavily rely on visual freshness cues.
- Traditional sensory panel testing for food freshness is resource-intensive and time-consuming.
Purpose of the Study:
- To evaluate an image processing technique for non-destructively classifying spinach freshness.
- To develop a machine learning model for automated spinach freshness assessment.
Main Methods:
- Spinach images captured via smartphone after varying storage durations.
- Image processing involved background removal, conversion to grayscale, CIE-Lab, and HSV color spaces.
- Extracted color features (mean, min, std dev) and local features (Oriented FAST, Rotated BRIEF) were used to train Support Vector Machine (SVM) and Artificial Neural Network (ANN) models.
Main Results:
- Color and local features showed significant correlations with sensory evaluation scores.
- The SVM and ANN models achieved high classification accuracies: 70% (4-class), 77% (3-class), and 84% (2-class).
- Model performance was comparable to individual human panel evaluations.
Conclusions:
- Image processing combined with machine learning (SVM, ANN) offers a viable, automated method for assessing spinach freshness.
- This technique has the potential to replace subjective and time-consuming manual freshness evaluations.
