Research On Online Diagnostic Technology And System Of Citrus Huanglongbing Based On MobileNet

Citrus is one of the most important fruit trees in southern China,and the citrus industry has become an indispensable part of the southern agricultural economy.Huanglongbing(HLB)is considered to be a cancer of citrus,which spreads quickly and is harmful.The citrus trees infected with Huanglongbing will be seriously affected or even die.Once the outbreak of Huanglongbing,it will cause huge economic losses to the fruit farmers and related industries.Until now,no drugs have been able to cure citrus Huanglongbing.Therefore,the prevention of citrus Huanglongbing must be detected early and isolated.The current diagnostic methods for citrus Huanglongbing,such as PCR detection technology,DNA probe hybridization and serological diagnosis,are difficult to promote in actual production.With the development of deep learning and artificial intelligence,more and more researches have been conducted on the detection of plant diseases using convolutional neural networks.Aiming at the problem that the diagnosis period of citrus Huanglongbing is long and difficult to promote in actual production,this paper studies the method of detecting citrus Huanglongbing on the mobile phone based on the mobile-derived deep learning network.Main responsibilities include:1.In order to explore the feasibility of this study,we first collected the images of individual citrus leaves,and then used the migration learning method to realize the identification of Huanglong disease in a simple background based on the MobileNet model,and implanted the diagnostic model into the Android phone(Hisilicon Kirin 960,RAM 6G)was tested.The model’s occupancy rate on the mobile phone CPU is 25%-30%,and the detection time of each frame image is 30-250 ms.2.Further,using MobileNet as the basic network of SSD target detection algorithm,the multi-leaf image collected in the citrus orchard was modeled and analyzed,and the Mean Average Precision(m AP)reached 85%,and the CPU occupancy rate was 40%-50%,the image detection time per frame is 320ms-360 ms.3.In order to make the model better applied to mobile phones,the citrus Huanglongbing online detection APP was designed,which enables users to select photos on the APP for local diagnosis of citrus Huanglongbing while the mobile phone is offline.At the same time,the function of the expert system has been added,and the expert detection of the citrus leaves can be performed to ensure the reliability of the diagnostic system.The online diagnosis technology and system of citrus Huanglongbing based on MobileNet developed by this subject can diagnose the citrus Huanglongbing in real time,promptly guide the farmers to diagnose Huanglong disease,and take preventive measures.The use of mobile APP for the detection of Huanglong disease is more time-saving than the traditional method,and the diseased plants can be found more conveniently,which provides a reference for the field diagnosis of citrus.