Multi-angle Identification Of Potato Late Blight Based On Multi-spectral Imaging

Potatoes are highly productive and are the world's fourth largest economic crop.Their production status and pest monitoring are an extremely important part of potato production development.However,the late blight caused by Phytophthora infestans is a devastating injury for potato production,which seriously threatens farmers' income and affects the healthy and sustainable development of the economy.Therefore,explor ing methods for rapid detection of late blight has important significance.The method which based on the traditional human eye vision and RGB digital co lor equipment to detect potato disease is often inefficient and unfaithful.Multi-spectral images containing spectrum is the characteristic of Spectral imaging technology whic h solves the problems "Multi-spectral images can not be combined with spectrum" an d "the same color have different spectrum" and other issues in the traditional disciplin es.However,the use of spectral imaging technology for plant disease detection still exist in the extraction of plant disease characteristics of information is not comprehensive,low reliability and other issues.Therefore,it is necessary to increase the angle dimension information in the traditional single-angle observation mode to obtain more spectral imaging information of the dimension.In view of this,this paper based on spectral imaging technology of potato late blight detection,made the following research:(1)A multi-spectral imaging system consisting of a Spectrocam camera and a computer was developed using a combination of A light source and fifteen narrowband filters.The system was used to collect multi-spectral images of healthy potato leaves and leaves of late blight with different illumination angles of 30 °,60 °,90 °,120 ° and 150 °.(2)Using the band index method,the characteristic bands of healthy potato leaves were extracted for false color synthesis.The experimental results show that channel 9(680nm),5(558nm),2(475nm)and channel 12(800nm),5(558nm),2(475nm)are the best combination of false color.(3)Using the band index method,the characteristic bands of healthy potato leaves at illumination angles of 30 °,60 °,90 °,120 ° and 150 ° were selected.The experimental results show that the characteristic bands of healthy potato leaves with illumination angle of 30 ° and 150 ° are 425 nm,558nm,578 nm,800nm,832 nm.The characteristic bands are 425 nm,509nm,558 nm,832nm,850 nm at illumination angles of 60 ° and 120 °.When the illumination angle is 90 °,the characteristic bands are 475 nm,558nm,680 nm,717nm,850 nm.(4)According to the selection results of the characteristic bands of healthy potato leaves at different illumination angles,the characteristic band extraction was carried out for the multi-spectral data of light angle 30 °,60 ° and 90 ° for the leaves of late blight.The results showed that the characteristic bands are 515 nm,558nm,578 nm,750nm and 832 nm when the light angle was 30 °.When the illumination angle is 60 °,the characteristic bands are 558 nm,620nm,650 nm,717nm and 800 nm,and the characteristic bands are 509 nm,578nm,620 nm,800nm and 832 nm when the illumination angle is 90 °.(5)From the Euclidean distance method and the correlation coefficient method,the characteristic bands of healthy potato leaves and leaves of late blight were classified at 30 °,60 ° and 90 °.For the healthy potato leaves,the experimental results show that the classification accuracy is high when the illumination angle is 60 ° in the Euclidean distance method,but the correlation is better when the illumination angle is 90 °.For the late blight leaves,the Euclidean distance method is better when the light angle is 30 °,and the correlation coefficient method is ideal when the illumination angle is 60 °.(6)The classification model of potato late blight based on LS-SVM algorithm was established,and the recognition rate was 87.92%.