Research Of Early Non-destructive Detection Of Cabbage Clubroot Based On Magnetic Resonance And Hyperspectral Imaging Technology

Cabbage(Brassica campestris ssp.chinensis Makino)is a cruciferous crop of the Brassica genus and is one of the common vegetables that originated in China.While cabbage obtains various nutrients and water through the root system.Root system can also infect root diseases and reduce plant yield when conditions are suitable.Root diseases are more difficult to achieve early detection than other diseases.This thesis takes the highly susceptible cabbage varieties as the research object,and takes nutrient soil,mixed soil and red clay as the growth medium.Magnetic resonance imaging technology and hyperspectral imaging technology are used for early non-destructive detection of cabbage clubroot.The main findings are summarized as follows:(1)The effects of different soil weight water content,different types of growth media,different data collection angles,and different growth periods on the spectral data of cabbage plants have been explored based on magnetic resonance spectroscopy technology.The peak position of the sample spectrum moved to the right as the soil moisture content increases.Soil moisture content between 50%-100% had the best effect on spectrum analysis.The inversion spectra of the three types of seedling trays,growth media and plant roots showed that the peak positions were quite different.The type of growth medium and the angle of data collection had no effect on the inversion spectrum curve of the sample.It showed that the spectral analysis technology of magnetic resonance imaging instrument can successfully distinguish the growth medium and plant samples.Analyzing the spectral data of healthy cabbage and diseased cabbage at different growth periods showed that there was difference between the relaxation time of diseased and healthy samples.The relaxation time of diseased samples was usually less than that of healthy samples.In the three growth media,as the plant grows,the relaxation time gradually approached the range of 10-100 ms.(2)Based on the imaging technology of the magnetic resonance instrument,the soil weight moisture content for the best imaging effect is determined.By studying the influence of growth media,soil moisture content,and averages on imaging result,a three-dimensional visualization model of the root system was established.Magnetic resonance tomography was performed on the samples under three growth media,and it was found that the growth media had no significant effect on the imaging of the samples.The imaging results of four gradients of soil moisture content showed that the higher the soil moisture content,the worse the imaging effect of the sample.The moisture content for the best imaging effect was between 50% and 100%.The number of averages was one of the important parameters in the imaging process.The greater the number of averages,the higher the clarity of the image and the better the evaluation of the root morphology.Finally,the root parameters in the 3D visualization model were compared with the measured values.It was found that the error between the model value and the measured value was within 3%.(3)Hyperspectral imaging technology combined with a variety of analysis methods had obtained the significant difference bands between healthy and diseased cabbage in multiple periods.Discriminant analysis models of clubroot with multi-phase and multi-type growth media was established based on traditional machine learning methods and deep learning models.In a continuous period of six weeks,the spectral data of each group of healthy and diseased plants were analyzed for significant differences in 350 bands.The bands with significant differences were: 4 bands for nutrient soil,97 bands for mixed soil and 5bands of red clay.The principal component analysis score image showed that there were obvious differences between plants of different health types.Among the selected vegetation indexes,physiological reflectance index(Ph RI),nitrogen reflectance index(NRI)and anthocyanin reflectance index(ARI)can better distinguish health samples from infected samples.The support vector machine(SVM)model and convolutional neural network(CNN)model established based on the pixel spectra of cabbage showed that the CNN model had better performance.