Establishing And Verifying Of Nutrient Expert System For Citrus Fertilization Recommendation

In view of the fact that the current nutrient expert system is not applicable to citrus and other fruit crops,a nutrient fertilization recommendation expert system has been constructed to improve fruit quality and yield with fertilizer reduction and nutrient efficiency improvements.In this study,10a citrus reticulata cv.’Dekopon’from the Citrus demonstration base in Suining National Agricultural Science and Technology Park,Sichuan Province,and 10-12a citrus grandis cv.‘Longan’from the Modern Agricultural Industrial Park of Longan Pomelo,Guangan City,Sichuan Provinc,were taken as the research object.3 kinds of citrus orchard yield monitoring algorithms,based on three acquisition methods:mobile phone images,mobile phone videos,and panoramic images,were carried out.The indexes were tracked,including the tree mineral nutrition at the physiological fruiting state fruit and fruit enlargement stage and,the monitoring yield at the fruit color turning stage,the fruit quality at the fruit maturity stage.The original database was subsequently established.Algorithms were constructed,including the yield monitoring algorithm based on fruit recognition,the nutrient criteria generation of high-quality leaf at the fruit expansion stage and high-yield leaf at physiological fruiting state,the correction fertilization of deviation from the optimum percentage(DOP),according to the standard of"yield quantification,the total quantity control,and staging control",and the compound fertilizer formula recommendation using the local fertilizer and agricultural materials supply database.Six functions were subsequently implemented,namely yield monitoring,automatic generation of high-quality leaf nutrient criteria during fruit expansion and high-yield leaf nutrient criteria during physiological fruiting state,nutrient assessment of citrus orchards,recommendation of available nutrient formulations,and recommendation of compound fertilizer formulations.The main results of the study are as follows:1)The yield monitoring algorithm of"Dekopon"based on fruit recognition was constructed.Three yield monitoring algorithms based on mobile phone images,mobile phone videos,and panoramic cameras are constructed.Finally,the optimal yield monitoring scheme based on smart phone with shooting on both sides of each plant and 5 meters panoramic camera was determined.The accuracy rates were 89.6%and 92.1%,recall rates were 92.1%and 83.3%,F1values were 90.8%and 89.7%,model determination coefficient R~2 reached 0.63 and 0.55.Forty-four mobile phone and three panoramic images were collected from 80 plants in the validation orchard,with a collection time of approximately 11 min and 6 min,and the relative errors between the monitored and actual yields were-3.07%and 5.49%,which could meet the demands of farmers and research institutes.2)Nutrient expert system for citrus fertilization recommendation has been Established.The system was distributed to the growers,government agencies,and system administrators with the different permissions using role-based access control(RBAC).The comprehensive fruit quality evaluation algorithm constructed by our team was introduced into the study,and CND inflection point value method was used to divide the characteristics of high-yield and high-quality plots based on yield or fruit comprehensive quality and leaf mineral nutrient response,combining with the original database of mineral nutrition and fruit quality of‘Dekopon’and Longan pomelo.The research was conducted at the Suining National Agricultural Science and Technology Park,where high-quality leaf criteria in the fruit expansion phase and high-yield leaf criteria in the physiological fruiting state were established,and at the Longan Pomelo Modern Agricultural Industrial Park in which high-quality leaf production in the fruit expansion phase and high-yield leaf production were established.The DOP bias index was used to construct the DOP algorithm-based nutrient state evaluation and effective nutrient correction formula fertilization algorithm for orchards.Combined with the compound fertilizer formulation algorithm constructed by our team,the feasibility of the proposed expert system for citrus nutrient fertilization has been verified.3)Complete effect verification of system recommended fertilization.To Quhe township,the effectiveness validation results of the recommended fertilization expert system show that the average middle-yielding plantations obtained by the monitoring model 24.65 t/hm~2 as the target yield is the best modified fertilization scheme.Nitrogen,potassium,and phosphorus fertilizer inputs were reduced by 4.6 percent,16.7 percent,and 54.5percent,respectively.Total soluble solids and fruit maturity index increased by 6.3 percent and13.8 percent,titratable acid and citrus color indices reduced by 8.0 percent and 8.9 percent,and the comprehensive fruit quality improved significantly.To Chongwang and Longan township,the results showed that the treatment of the two orchards performed best with the average middle-yielding plantations 35.71 t/hm~2,particularly as recommended for fertilization.Specifically,compared to the chemical empiric fertilization scheme,the treatment of longan was reduced by 40.3 percent,43.6 percent and 1.3 percent in nitrogen,kalium and potassium inputs,Total soluble solids and fruit maturity index increased by 5.2 percent and 14.5 percent,titratable acid and citrus color indices reduced by 13.1 percent and 90.6 percent,respectively.In contrast,treatment with RF2 of Chongwang reduced N,K and P inputs by 27.1 percent,30.8percent,and 78.8 percent,Single fruit weight,total soluble solids and fruit maturity index increased by 19.1 percent、9.3 percent and 12.4 percent.At the same time,the treatment can be expected to be effective in avoiding delays in fruit pulp coloring due to empirical organic fertilizer applications.Therefore,two orangeries significantly improved the comprehensive fruit quality of Longan Pomelo,providing case support for the subsequent system promotion.