Development Of Detetion Device For Potato Late Blight Disease Based On Spectrum And Image Technology

Potato is rich in nutrition,which is the third biggest economic crop in the world.It is also one of the four main grains in China,but the average yield per mu is far below the world average.Late blight is an important factor restricting potato production.Late blight is widespread and destructive.When it breaks out,it often leads to crop failure,which causes immeasurable economic losses to farmers and the country.Rapid and accurate identification of diseases is one of the key technologies to control the development of late blight.Traditional late blight detection relies on experienced manual work,which is imprecise,complicated,destructive and difficult to popularize.In this study,a high precision,detection device for late blight was designed and developed.The main research contents and conclusion are as follows:Firstly,Early prediction method of late blight based on spectrum and stress-resistant enzymes was proposed.In the near infrared band(400-1100 nm),prediction models based on full spectrum and activity of stress-resistant enzymes were established by using support vector machine(SVM)and partial least squares(PLS).Prediction model based on the characteristic wavelength extracted by SPA algorithm and stress-resistant enzymes activity was established.The prediction accuracy was improved and the prediction time was reduced,the early detection of late blight by spectrum was realized.Secondly,an early prediction and classification method of late blight based on image was proposed.A 10-layer CNN model was constructed based on the characteristics of hyperspectral images of potato leaves,which was infected with late blight in different degrees.The recognition rate of the model was 92%.It makes up for the deficiencies of spectrum-based disease detection method,which needs to indirectly judge whether the plant is infected by the enzyme activity,and that there is no spot location and status image.Thirdly,a hardware system of disease detection device was established.According to the potato planting environment in Northwest China and the characteristics of potato late blight,a detection device was designed and developed.The detection device is divided into three modules: system control module,data acquisition module and power supply module.A crawler-type field mobile detection platform was designed.Stainless steel skeleton and crawler walking mechanism to meet the stability requirements of the detection device in the field.In the data acquisition module,the optical fiber spectrometer and RGB high-definition camera are used to collect the spectrum and image information of the blade in real time through remote control of the Web server.The 12 V lithium battery pack powers the motors and the L298 N control board,and the 220 V battery pack powers the optical fiber spectrometer and the raspberry pi.Raspberry Pi was used as the control board to receive wifi signals from hot spots of mobile phones.Realizing remote data acquisition and wireless transmission.Fourth,a software based on Python and the system was designed with hierarchical and modular design ideas.It is divided into five modules: user management,blade management,device control and information acquisition,data management and intelligent display.The system has a user-friendly image operation interface and good compatibility.It can complete the motion control of the detection device,and real-time transmission of spectral data and RGB image data through network communication.The automatic monitoring of potato growth status in the field,early detection and disease classification of late blight were realized,and the purpose of early warning of field diseases was achieved.Fifth,the function and performance of the detection device system were tested by building a real scene simulating the planting environment in on the campus of Northwest University of Agriculture and Forestry Science and Technology.The results showed that the system can perform its functions in an orderly manner according to the prescribed process.The system can login smoothly in different network environments,and the interface runs well.The detection device has good obstacle surmounting performance and can operate in the actual field environment.And the average response of common functions is no more than 3 seconds.In addition,the accuracy test results of the model show that the overall recognition accuracy of the model is as high as 89%.Among them,the accuracy of identifying the early state of late blight was 80%.