With the improvement of living standards, consumers are increasingly concerned about the quality and safety of fruits and vegetables. Issues such as black and white spots on the surface, internal rot, or damage caused by transportation significantly affect consumer health. Therefore, developing a rapid and effective method to identify these defects is crucial for ensuring food safety and maintaining market value.
Hyperspectral imaging technology combines spectral analysis with image processing, making it an ideal tool for detecting both external and internal characteristics of fruits. Researchers like Zhao Jiewen have used this technique to detect minor fruit damage with an accuracy rate of 88.57%. Jasper G. Tallada applied hyperspectral imaging to assess surface damage in strawberries, apple blemishes, and mango ripeness. Other studies, such as those by Wang Yutian using fluorescence spectroscopy, Hu Shufen with laser technology, and Xue Long focusing on navel oranges, have explored the detection of pesticide residues. These efforts highlight the versatility of hyperspectral imaging in agricultural quality control.
This study uses hyperspectral imaging to detect black and white spots and damaged areas on different fruits, aiming to develop a fast and accurate identification system. The research includes detailed experimental setups, data acquisition procedures, and image processing techniques to analyze spectral reflectance and identify defects efficiently.
The experiment involved oranges as test subjects, where black and white spots and bruises were intentionally created. Data was collected using the GaiaSorter hyperspectral sorting system, which includes a V10E hyperspectral imager, CCD camera, light source, and other components. Spectral parameters such as scanning range (400–1000 nm), resolution (2.8 nm), and collection interval (1.9 nm) were recorded for reference.
Image processing was conducted using software like SpecView and ENVI/IDL, including preprocessing steps such as mirror transformation and black-and-white calibration. The Minimum Noise Fraction (MNF) transform was applied to reduce noise and extract key features from the hyperspectral data. Analysis showed that certain eigenvalues could better distinguish between normal, damaged, and spot regions, particularly in the 530–1000 nm range.
A vegetation index (NDVI) was also constructed to enhance defect recognition, combined with threshold segmentation for faster and more accurate results. The findings suggest that while MNF provides detailed information, its complexity makes it less suitable for industrial applications. In contrast, the vegetation index algorithm is simpler and efficient, requiring only four bands for quick identification of spots and damage.
Overall, hyperspectral imaging demonstrates great potential in automating fruit quality inspection, offering a reliable and scalable solution for modern agricultural practices.
Tungsten Molybdenum Fabricated
Shaanxi Xinlong Metal Electro-mechanical Co., Ltd. , https://www.cnxlalloys.com