Related Experiment Video
Updated: Sep 21, 2025
09:37
Combining Reflectance Confocal Microscopy with Optical Coherence Tomography for Noninvasive Diagnosis of Skin Cancers via Image Acquisition
Published on: August 18, 2022
2.5K
Convolutional neural network for apple bruise detection based on hyperspectral
Zhaodong Gai1, Laijun Sun1, Hongyi Bai2
1College of Electronics Engineering, Heilongjiang University, Harbin, Heilongjiang Province, China; Jiaxiang Research Academy of Industrial Technology, Jining, Shandong Province, China.
Summary
This study introduces SpectralCNN, a novel spectral analysis model for detecting apple bruises. The model achieves high accuracy, aiding the apple industry by reducing economic losses from damaged fruit.
Area of Science:
- Agricultural Engineering
- Food Science
- Machine Learning
Background:
- Apple bruises from impact and pressure cause significant economic losses in the fruit industry.
- Early detection of these defects is crucial for quality control and waste reduction.
Purpose of the Study:
- To develop and evaluate a spectral analysis model, SpectralCNN, for accurate apple bruise detection.
- To assess the impact of spectral preprocessing techniques on model performance.
- To identify key spectral features for bruise identification.
Main Methods:
- A one-dimensional convolutional neural network (SpectralCNN) was developed for spectral analysis.
- Six different spectral preprocessing methods were investigated.
- The successive projection algorithm was used to extract characteristic wavelengths.
Main Results:
- The SpectralCNN model demonstrated superior accuracy compared to traditional chemometric models.
- Model accuracy was largely independent of the spectral preprocessing method used.
- Extraction of 20 characteristic wavelengths yielded a 95.79% accuracy on the test set.
Conclusions:
- SpectralCNN is an effective tool for detecting apple bruises using spectral data.
- Characteristic wavelengths identified by the successive projection algorithm capture essential spectral features for bruise detection.
- The model offers a promising approach for automated quality assessment in the apple industry.
