Study On The Classification Model Of Flavonoids Content In Acanthopanax Senticosus Leaves Based On Hyperspectral Technology

Nowadays,a number of people is pay attention to the trace elements and minerals they need while paying attention to their own diet.Flavonoids,as trace minerals necessary for daily health care of the human body,have the effects of reducing the incidence of cardiovascular and cerebrovascular diseases and improving sleep conditions.They are widely present in wild plants,especially the Acanthopanax senticosus plants account for the highest overall proportion.However,due to many restrictions such as the growth conditions of wild Acanthopanax senticosus,it is difficult to obtain it.Therefore,this paper proposes a classification study based on the content of hyperspectral Acanthopanax senticosus flavonoids under large-area planting.From the perspective of model classification,many months of comparative analysis and different fertilization methods is carried out,and the most suitable fertilization methods and picking methods are selected.time.Its significance is:fast,real-time and accurate classification and estimation of flavonoid content in Acanthopanax senticosus leaves through hyperspectral non-destructive imaging technology,while avoiding environmental impact and chemical inspection damage to the leaves,avoiding waste of labor and inspection costs.The main research of this paper is as follows:(1)The experiment was carried out in two parts in August and mid-September 2019,with a period of 7 days.The test site is divided into two areas,A and B,in which area A is set as a fertilizer application area,and area B is set as an organic application area.The sampling in each area is repeated three times.A hyperspectral imager was used to photograph the canopy leaves of Acanthopanax senticosus,and an ultraviolet spectrophotometer was used to determine the content of flavonoids in the leaves.ENVI 5.3 software was used to extract the leaf reflectivity.In the500～600nm band,the overall absorption peak contrast range of 8-A,8-B,9-A,and 9-B is0.05～0.18,0.05～0.13,0.15～0.25,0.04～0.18,the average flavonoid content was 0.103,0.111,0.098;0.085,0.086,0.079;0.122,0.127,0.132;0.093,0.106,0.102.(2)Five processing methods of MSC,SG,SNV,FD,and SD are selected through the Unscramble 10.4 software to perform spectral preprocessing.Because the conditioning effect of the SD method is low.,the article only deals with the data combined with the FD processing method.analysis.Among them,R_c~2and R_p~2of SG-FD,MSC-FD,and SNV-FD are 8-A:0.8869,0.8431;0.8095,0.7688;0.8007,0.7613 under different variables.8-B:0.7625,0.7104;0.7217,0.6423;0.7170,0.6747.9-A:0.9521,0.8790;0.8674,0.8207;0.8017,0.7204.9-B:0.8151,0.7678;0.7963,0.7533;0.7616,0.6970.In comparison,SG-FD is finally selected as the best pretreatment method.And use SPA and PCA as the method of characteristic wavelength extraction,respectively select 13,13,15,9 and 17,16,15,23 characteristic wavelength.(3)KNN-Adaboost algorithm model and GMM-BP neural network algorithm model are established respectively,and KNN and GMM are used as classifiers to classify the data.After introducing the iterative parts of the Adaboost algorithm and the BP neural network into the KNN and GMM models,respectively,significantly increased the accuracy of model classification and reproduction.,which solves the problem of fewer iterations and poor classification accuracy of a single classification model.The classification accuracy range and number of iterations of the KNN-Adaboost algorithm model and the GMM-BP neural network algorithm model are98%～99.7%and 96,80,55,68;95%～97.7%and 14,10,10,14.(4)Introduces the number of wrong sample points,number of sample points,boundary,R~2coefficient of determination,and squared error of the RMSE root mean.,and the number of iterations.These five model evaluation indicators are used to comprehensively compare the pros and cons of the model.And select the KNN-Adaboost algorithm model as the optimal algorithm model in this paper,and determine the final optimal picking time and artificial planting method as the growth conditions for fertilizer application in September.And provide a feasible theoretical reference for the planting of Acanthopanax senticosus.