Related Experiment Video
Updated: Aug 30, 2025
11:37
RGB and Spectral Root Imaging for Plant Phenotyping and Physiological Research: Experimental Setup and Imaging Protocols
Published on: August 8, 2017
16.3K
Detection of early decayed oranges by structured-illumination reflectance imaging coupling with texture feature
Zhonglei Cai1, Wenqian Huang2, Qingyan Wang2
1College of Mechanical and Electrical Engineering, Shihezi University, Shihezi, China.
Frontiers in Plant Science
|August 29, 2022
Summary
Early detection of fungal infection in citrus fruits is crucial to prevent economic loss. A novel structured-illumination reflectance imaging system successfully identified early decay in oranges with 96.4% accuracy using texture analysis and a PLS-DA model.
Area of Science:
- Agricultural Engineering
- Food Science
- Computer Vision
Background:
- Post-harvest fungal infections cause significant economic losses in citrus fruits.
- Early detection of decay is challenging due to invisible surface symptoms.
Purpose of the Study:
- To develop and validate a system for early detection of fungal infection in oranges.
- To identify optimal parameters for image acquisition and analysis.
Main Methods:
- A structured-illumination reflectance imaging (SIRI) system with LED illumination and a monochrome camera was employed.
- Sinusoidal illumination at varying spatial frequencies and three-phase-shifted imaging were used.
- Image demodulation to obtain direct (DC) and alternating (AC) components, texture feature extraction, and classification models (PLS-DA, SVM, LS-SVM, KNN) were applied.
Main Results:
- Decayed areas were clearly identifiable in AC and RT (AC/DC) images compared to DC images.
- The partial least square discriminant analysis (PLS-DA) model, using eight texture features from RT images, achieved the highest classification accuracy of 96.4%.
Conclusions:
- The developed SIRI system is effective for the early detection of fungal decay in oranges.
- Combining SIRI with appropriate texture features and classification models offers a promising solution for post-harvest quality assessment.
