Related Experiment Video
Updated: Aug 6, 2026
08:27
Image Recognition and Parameter Analysis of Concrete Vibration State Based on Support Vector Machine
Published on: January 5, 2024
Prediction of compost physicochemical properties using image-based feature extraction and machine learning
Anshuman Nayak1, Lewis S Rowles2, Thomas Echols2
1Agricultural and Food Engineering Department, Indian Institute of Technology Kharagpur, Kharagpur, India.
Waste Management (New York, N.Y.)
|July 23, 2026
Summary
This study uses smartphone photos and machine learning to quickly predict compost quality, including pH and electrical conductivity. This low-cost method offers a rapid screening approach for compost assessment.
Area of Science:
- Agricultural Science
- Environmental Science
- Computer Science
Background:
- Compost quality assessment is crucial for effective soil amendment and environmental management.
- Traditional methods for analyzing compost physicochemical properties can be time-consuming and resource-intensive.
- Developing rapid, non-destructive methods for compost analysis is essential for widespread adoption and quality control.
Purpose of the Study:
- To evaluate the use of standard digital photographs combined with automated image segmentation and machine learning for predicting compost physicochemical properties.
- To develop and validate a rapid, non-destructive method for assessing compost quality indicators such as pH, electrical conductivity (EC), volatile solids (VS), ash content, and bulk density (BD).
Main Methods:
- Collected 230 compost samples from diverse locations in the USA.
- Acquired digital images using a smartphone (iPhone 13 Pro Max) under controlled lighting.
- Utilized a U-Net convolutional neural network for automated image segmentation of compost regions.
- Extracted image-derived color, texture, and morphological features.
- Employed K-nearest neighbors (KNN) regression for predicting physicochemical properties, comparing it with nine other algorithms.
Main Results:
- K-nearest neighbors (KNN) regression demonstrated optimal predictive performance for compost properties.
- Strong internal validation results were achieved: EC (R² = 0.97), pH (R² = 0.95), VS (R² = 0.92), ash content (R² = 0.94), and BD (R² = 0.87).
- Encouraging external validation results were observed: EC (R² = 0.69), pH (R² = 0.75), VS (R² = 0.81), ash content (R² = 0.70), and BD (R² = 0.63).
Conclusions:
- Smartphone-based imaging, coupled with U-Net segmentation and KNN regression, offers a rapid and low-cost screening approach for visually linked compost quality indicators.
- The findings support the potential for field-scale deployment of this technology.
- Further validation across diverse feedstocks, composting systems, and environmental conditions is recommended to broaden applicability.
