An Apple Fungal Infection Detection Model Based on BPNN Optimized by Sparrow Search Algorithm

Area of Science:

• Agricultural Science
• Food Science
• Computational Biology

Background:

• Fungal infections pose a significant threat to apple quality and yield.
• Rapid and non-destructive detection methods are crucial for effective disease management in the apple industry.

Purpose of the Study:

• To develop and evaluate an electronic nose system for the rapid detection of fungal infections in apples.
• To compare the performance of various machine learning models for classifying infected and healthy apple samples.

Main Methods:

• Utilized a portable electronic nose to capture volatile organic compound (VOC) profiles of apples.
• Applied data preprocessing techniques including smoothing filtering, dimensionality reduction, and outlier removal.
• Trained and evaluated multiple classification models: KNN, RF, SVM, CNN, BPNN, PSO-BPNN, GWO-BPNN, and SSA-BPNN, using 10-fold cross-validation.

Main Results:

• The Sparrow Search Algorithm (SSA) effectively optimized the Back Propagation Neural Network (BPNN) model.
• The optimized SSA-BPNN model demonstrated a high recognition accuracy of 98.40% for detecting fungal infections in apples.
• All tested models showed varying degrees of success, with the SSA-BPNN outperforming others.

Conclusions:

• The electronic nose combined with the SSA-BPNN model offers a promising solution for non-destructive, rapid, and accurate detection of fungal infections in apples.
• This approach provides a valuable tool for quality control and disease surveillance in the agricultural sector.
• Further research can explore broader applications of this technology for other fruits and diseases.