Identification of split mouth chestnut by hyperspectral image technique

In this study, a 400-1000nm hyperspectral camera was applied, and FS13, a product of Hangzhou Color Spectrum Technology Co., LTD., could be used for related research. The spectral range is 400-1000nm, the wavelength resolution is better than 2.5nm, and up to 1200 spectral channels can be reached. The acquisition speed can reach 128FPS in the full spectrum, and the maximum after band selection is 3300Hz (support multi-region band selection).

Chestnut is one of the edible nuts in China, high quality and low price, rich in nutrition, the annual output ranks first in the world. Schizorhynchus is one of the important indexes to evaluate the external quality of chestnut. Schizorhynchus is a kind of chestnut whose peel is cracked under natural production conditions or damaged by external forces such as mechanical damage. The exposed pulp of chestnut can easily lead to a series of food safety problems. At present, the split mouth chestnut mainly adopts manual sorting, which is subjective and has a high sorting error rate. Therefore, the study of an effective and applicable method for the detection of split mouth chestnut can lay a foundation for the rapid non-destructive detection and classification of chestnut. In view of the identification methods of defective chestnuts, the research group has done some research in the early stage, but there is no report on the identification methods of cracked mouth defects in defective chestnuts. Near infrared spectroscopy technology can quickly, non-destructive and effective detection of internal quality information of agricultural products, machine vision technology can well reflect the external characteristics of agricultural products, have been widely used in agricultural product quality detection, but both can not meet the requirements of detection of internal and external quality of agricultural products. With the rapid development of science and technology and the rapid development of computer technology, hyperspectral image detection technology, which integrates spectrum and image, has been paid more and more attention by researchers in the field of non-destructive testing of agricultural products. Hyperspectral images can record abundant quality information of agricultural products and can be used to detect both internal and external quality of agricultural products. Scholars at home and abroad have applied hyperspectral image technology to non-destructive testing of fruits, vegetables, tea and meat and achieved good results. However, there is no study on the detection of split mouth chestnut by hyperspectral image technology. In this paper, hyperspectral image technology is used to identify split mouth chestnut, extract and analyze the spectral curves of split mouth chestnut and qualified chestnut, select the characteristic wavelength, adopt the band ratio algorithm, extract the cooperative image through texture filtering, and combine with a series of mathematical morphology to complete the identification of split mouth chestnut, which can provide a new idea for online detection of split mouth chestnut.

In this paper, hyperspectral image technology was used to identify split mouth chestnut.

1) The characteristic wavelengths (477nm, 769nm and 923nm) were selected by principal component analysis, and the band ratio image obtained by different combinations of the characteristic wavelengths and the single band image at the characteristic wavelength were analyzed and compared, indicating that the 769mm/923nm band could best reflect the split nozzle region than the image, and was more conducive to the extraction of split nozzle features.

2) The image of 769nm/923nm band ratio was analyzed, the image based on collaborative texture filtering was extracted, and the target region was extracted by combining threshold segmentation and mathematical morphology. The correct recognition rate of cracked beak was 94.3%, the recognition rate of qualified chestnut was 96.8%, and the overall recognition rate reached 95.5%. The filter based on the filter type hyperspectral image detection system is designed to realize the on-line, rapid and non-destructive detection of split mouth chestnut. At the same time, it also provides a new method for the quality detection of other agricultural products.