Identifying Of Rapeseed(Brassica Napus L.) Leukoplakia Based On Spectrum

As one of the world’s four major oil crops, rapeseed is planted widely. Leukoplakia of Brassica napus L. is a ordinary disease occuring at all areas of rapeseed in our country, especially in rainy season and weak plant in growth status, which results in rapeseed yield loss and bad quality. It can be monitored and identified by spectrum rapidly, macroscopically and dynamically. Controlling the disease outbreak in time is significant to reduce the yield losses.In 2013 and 2014, the spectrum data of rapeseed leaf leukoplakia was collected in the experimental Farm of Jiangsu Academy of Agricultural Science. Disease index (DI) was defined on a single leaf level. The agronomic parameters for disease leaf, leaf nitrogen content, leaf moisture content, and leaf SPAD value, were determined. The relationships between these parameters and the reflectance spectra were studied, the main findings are as follows:1. Analysis of changes in agronomic parameters and disease index of rapeseed leaf leukoplakia based on spectraFirstly, the common distinctive bands of the disease and the health were found by comparing reflectance spectrum of leaves in the field and under black background. The results showed that with the progress of the growth period, reflectance of disease leaves decreased earlier than healthy leaves. It was the best period to identify rapeseed leukoplakia from 11 days after early anthesis to 9 days after finish flowering to identify rapeseed leukoplakia in the field due to during this period the reflectance of healthy leaves remained at 35% while the disease diseased to 30%. The sensitive band was 760-1080nm. The correlation among disease index (DI), agronomic parameters, and the reflectance of the disease samples were analysed, and the results showed that there were high correlations between DI, and agronomic parameters and reflectance, e.g., the correlation between the leaf moisture content and the reflectance in 460nm,550nm,650 nm,710 nm,760nm,1480 nm, and 1600 nm, between the leaf nitrogen content and the reflectance in 810nm,870nm,1080nm,1280nm,1320nm,1540 nm,1600 nm,1650nm, and 1700nm, and between the SPAD value and the reflectance in 1200 nm,1280 nm, and 1540 nm had significance with p<0.01. The quantitative models of agronomic parameters based on reflectances were got by stepwise regression, principal component analysis, and curve fitting. The data of rapeseed leukoplakia in 2013 and rapeseed virus in 2014 were used to test. The results showed that in the same disease test, the quantitative models of moisture content based on reflectance were fit well. In the different disease test, the quantitative models were fit badly except the model of moisture content. The moisture content of these two diseases leaves can be quantified by the model of moisture content. The model accuracy of nitrogen content was low. Therefore, the research of quantifying the agronomic parameters of rapeseed leukoplakia based on VI still need to be improved in future.2. Spectrum-based identification of rapeseed diseaseThe VI (Vegetation Indices) with the largest distance between the disease and control were screened out by transforming the reflectance of sensitive band. Then by comparing the various cluster analysis methods to the disease samples, the best clustering method, K-Means, was chosen to be used in clustering the samples in each period. The dataset of rapeseed leukoplakia in 2013 and rapeseed virus in 2014 were used to test, and the results showed that under the single leaves with black background, these two diseases could be identified completely by R810/R650, R.870/R650, R1080/R650, and R1200/R650, and R1200/R460 can identify rapeseed leukoplakia completely, while the recognition rate of the rapeseed virus is only 70%. R1200/R460 can distinguish these two diseases. Under the leaf with field background, the disease of rapeseed virus can be identified completely by R1200/R650, and the overall recognition rate was over 85%, but the recognition rate of rapeseed leukoplakia was low. While R1280/R460 could indetify leukoplakia completely, virus disease was low. This showed that R1200/R650 could be used to indetify the virus disease, R1280/R460 could be used to indetify leukoplakia of Brassica napus L. in the field.DI was quantitied by the reflectance of 1540nm and 650nm. After tested, this model could be used to quantify the severity of rapeseed leukoplakia leaves in the field. It could provide reference for automation of accurately identifying rapeseed disease and its prevention and control in the future.