Image-based Brassica Napus L.study On The Detection Method Of The Severity Of Clubroot Disease

Rapeseed is an important oilseed crop in China.Brassica napus has strong disease resistance and high yield,and is widely cultivated at home and abroad.Brassica napus will be harmed by rhizobia during the growth process,affecting its growth,yield and quality.Accurately detecting the degree of root swelling disease in rapeseed(Brassica napus L.)can evaluate the damage of root swelling disease and variety resistance,and further provide theoretical basis for breeding resistant varieties and comprehensive disease control.At present,the severity of rape root swollen disease is mainly determined by artificial identification methods.This method can only be judged and prevented based on its own experience,so it is very subjective,time-consuming,labor-intensive,and easy to misjudge,and has certain limitations.With the rapid development of image processing technology,image processing technology can be combined with crop disease diagnosis.In this paper,we mainly study the method of detecting the disease severity of rapeseed rape root from the aspects of image segmentation,feature extraction and classification model,and finally we get a classification model of disease severity.Testing with samples,the average error between the expected output value and the network output value is 0.15833,the mean square error is 0.17761025,and the detection accuracy of the result is 90.00%.The main work of this article is as follows:(1)Preprocessing of Brassica napus plant images.The root segmentation effect of rapeseed is mainly affected by factors such as image sets.Therefore,pre-processing of the image is required to improve the quality of the image segmentation effect.The general method is to crop the images in order,replace the multi-plant images with single-plant images,and then perform the post-operation.In this paper,color space model is used to segment the rape root image: compared with common space models,the color components that can highlight the root of rape are superimposed,then threshold segmentation is performed,and the segmented image is binarized and denoised.(2)Feature Extraction of Brassica napus Plant Images.This feature extraction is mainly aimed at the shape features of Brassica napus plants.The shape features are extracted from two aspects.On the one hand,feature extraction is performed from the entire root of rapeseed.As feature parameters,the other part is to equalize the entire root and select segment features.Based on experimental experience,this paper uses 5 equal parts,and finally selects 12 feature parameter combinations for the design of disease classification models.(3)A classification model was established for the disease severity of rapeseed rape.A BP neural network is used to design a classification model of the disease severity ofrapeseed rape,and the category of the disease severity of rapeseed that needs to be detected is set as an output.The selected 12 characteristic parameters are set as the input of the network to establish the BP neural Network model.The trained classification model was used to detect the condition of rape root swollen disease.Based on the test samples,the mean square error between the expected output value and the network output value was0.17761025,and the accuracy of the classification result was 90.00%.