Rice leaf blast disease resistance identification grading method based on multi-scale hyperspectral image processing

Abstract

The invention discloses a rice leaf blast disease resistance identification grading method based on multi-scale hyperspectral image processing. Hyperspectral images of rice leaves with different resistance grades infected by rice blast are colleted by a hyperspectral imaging system. Spectral features of rice leaf blast disease spots and a normal position area of interest are analyzed at leaf scale, two wave bands with greater differences are obtained, two-dimensional scatter plot analysis of the two wave bands is made, and hyperspectral images only containing the disease spots are extracted. And then principal component analysis (PCA) is made at a disease spot scale, a principal component image which is beneficial for segmentation of brown disease spots and grey disease spots is obtained, and the grey disease spots are segmented out through an OTSU method. Finally, rice leaf blast disease resistance grading is conducted according to two parameters of elongation rate and suffered rate. With the rice leaf blast disease resistance identification grading method based on the multi-scale hyperspectral image processing, workload of resistance identification can be reduced, accuracy of resistance evaluation is improved, reasonable promotion and use of new disease-resistant varieties are supplied with scientific basis, and detection of rice leaf blast disease degree in the field is supplied with research foundation.

Description

technical field [0001] The invention relates to a method for identifying and grading rice leaf blast resistance, in particular to a method for identifying and grading rice leaf blast resistance based on multi-scale hyperspectral image processing, which belongs to the field of agricultural technology. Background technique [0002] Rice blast is one of the most serious rice diseases in my country's northern and southern rice-growing regions. It is also known as the three major diseases of rice together with sheath blight and bacterial blight. At present, the identification and identification and classification of rice blast disease are mainly done manually through picture comparison or visual inspection based on text descriptions. Judgment errors affect the identification and identification results. [0003] In recent years, scholars at home and abroad have mostly used spectral analysis or multispectral imaging technology to detect rice blast. However, spectral technology on...

AI Technical Summary

Problems solved by technology

However, spectral technology only obtains the average spectral information of a single leaf or within the rice canopy. Although these average spectral information can be used to diagnose rice blast and identify the degree of disease, due to the lack of reflectance spectra in terms of spatial location differences , making the early identification and detection of rice blast more difficult, especially in the detection and identification of small disease spots in the early stage of the disease, and its detection accuracy is low; multispectral imaging technology generally only has a few bands in the visible and near-infrared regions, and the obtained spectral data information Less, in previous studies on the grading method of leaf blast disease, it was mostly divided into 4 grades according to the GB / T15790-1995 standard

When the leaves are moderately infected and the disease spots are large and many, the classification accuracy is high; but when the leaves are slightly infected and the disease spots are small and small, the recognition accuracy is often low

Images