Seed variety quality detection method and system based on hyperspectrum

Abstract

The invention belongs to the technical field of crop detection, and discloses a seed variety quality detection method and system based on hyperspectrum, and the method comprises the steps: extracting an average spectrum of a sensitive region, and constructing a grain variety prediction model based on a convolutional neural network through different spectrum features; and meanwhile, grain seed forms and surface texture information are combined to assist spectral data for joint discrimination, and rapid nondestructive detection is performed on grain seeds by using a hyperspectral imaging method. According to the invention, image features and spectral features are simultaneously obtained based on the hyperspectral image, and the form and surface texture information of the grain seeds are combined to assist spectral data for joint discrimination, so that the accuracy of grain seed identification is improved; and multiple preprocessing methods are selected to improve the accuracy and stability of the detection model. The nondestructive component detection of the grain seeds is realized based on the spectrum-component correlation model.

Description

technical field [0001] The invention belongs to the technical field of crop detection, and in particular relates to a method and system for quality detection of seed varieties based on hyperspectrum. Background technique [0002] At present, hyperspectral technology is a non-destructive testing technology that integrates spectral technology and image technology. Image information represents the visual characteristics of grains such as size, shape, and defects, and spectral information can fully reflect the differences in physical structure and chemical composition inside the sample.” "Mapping in one" has higher potential for analysis and detection. The image information can reflect the external quality characteristics of the sample. Since different components have different spectral absorptions, while imaging the spatial characteristics of the hyperspectral technology target, each spatial pixel is dispersed to form dozens or even hundreds of narrow bands. Continuous spectra...

AI Technical Summary

Problems solved by technology

[0018] ① When using grayscale image detection, it cannot reflect the difference between grayscale images of different bands

[0019] ② When using near-infrared spectral characteristics to detect a single grain, due to the small size of the seeds and individual differences in morphology, the spectral error between samples is large, which has a great impact on the modeling results

[0020] ③ When hyperspectral technology is used to detect seed quality, there are few studies on the combination of spectral information and image information

[0022] ⑤ There are few studies on identifying maize seed maturity and freezing damage by using images or spectra, and the recognition effect of using near-infrared spectroscopy to identify freeze-damaged seeds is not good

[0024] (1) The existing technology mainly uses a single spectral data or texture feature, and there is no method for multi-feature fusion detection; at the same time, the existing seed recognition method has poor recognition effect;

[0025] (2) The existing technology lacks the method of using hyperspectral data to detect seed components, and the existing detection methods of seed components will cause damage to seeds

Images