| Alfalfa is an important forage crop in China’s animal husbandry,and its leaf area and leaf number directly determine its yield.Since alfalfa leaf area can directly reflect its yield,leaf area estimation technology is the key to estimate alfalfa yield.Based on the rapid and accurate estimation of alfalfa leaf area,leaf area prediction can provide effective guidance for field management strategies.This paper takes alfalfa as the research object,and carries out research and discussion on the key issues of alfalfa leaf area estimation and prediction.The research content includes the following four aspects.(1)In response to the current lack of public datasets on the growth environment and surface data of alfalfa in Ningxia Yellow River Diversion Irrigation Area,this paper carried out alfalfa planting work of multiple different varieties and different water treatments in the agricultural reclamation test site of Military Horse Farm in Xixia District,Yinchuan City,Ningxia.At the same time,indoor and outdoor comparative experiments were conducted.By setting up a Tianqi meteorological station and a ZhiShang soil moisture monitoring instrument,an experimental environment for collecting alfalfa growth environment data was established.Between May 21,2022 and October 4,2022,data from three crops of alfalfa at different growth stages were measured.A total of 19,584 pieces of alfalfa growth environment data and 64,116 pieces of alfalfa leaf data were measured and the corresponding data was sorted to form a complete dataset.(2)In view of the problem of low efficiency and accuracy of existing alfalfa leaf area estimation,methods for estimating single leaf and single plant leaf area were respectively carried out to achieve rapid and accurate measurement of leaf area.According to the characteristics of alfalfa leaves,in the estimation of single leaf area,the relationship between leaf area and leaf length,leaf width and some environmental characteristics was analyzed.Based on the results of correlation analysis,it was concluded that there was a strong correlation between leaf length and width and leaf area.Therefore,leaf length and width were used as input features of the model,and linear regression and ANN models were used to estimate leaf area.The estimation accuracy of linear regression method was 99.6%,while that of ANN was 99.9%,which was 0.3%higher than that of linear regression method.For single plant leaf area estimation,a single plant leaf area estimation method incorporating plant height was proposed,which improved the estimation accuracy by 1.1%compared with the model without plant height.(3)In view of the problem that the low efficiency and large workload of alfalfa leaf area collection result in a large time interval of leaf area data,which in turn affects the accuracy of leaf area prediction,a leaf area prediction method based on GAIN data interpolation was proposed.This method uses an improved GAIN network to interpolate and add data.For data interpolation results with large errors,data was re-interpolated.Through experimental comparison,it was found that the accuracy of data supplementation using GAIN was 96.8%,which was 8.3%higher than that using linear regression(88.5%).The accuracy of leaf area prediction using supplemented data and LSTM model was 99.7%,which was 2.1%higher than that predicted by LSTM without supplemented data(97.6%).(4)A system based on leaf area estimation and prediction was developed.The system mainly includes leaf area estimation and prediction modules.Plant data manually collected by humans,including leaf length,leaf width,leaf area,nodule height,number of leaves,stem thickness and other data are uploaded to the system,and visual development of related functional interfaces such as estimation and prediction is completed.Experiments show that all data collected can be successfully uploaded to the system,all functions can be used normally,and users can intuitively and clearly display the predicted results of leaf area. |
PDF Full Text Request