Design Of Forage Quality Evaluation System Based On Multi-Source Heterogeneous Data

Chinese grassland resources are very rich,and the rich grassland resources provide a solid material basis for the development of Chinese animal husbandry.Among them,forage quality determines the utilization rate of forage,which is directly related to the production of animal husbandry.Forage quality is also an important indicator for evaluating the value of grassland resources.At present,the laboratory anatomy method and spectral inversion method are the main methods for forage quality detection.Although these methods have high detection accuracy,they have high detection cost and long detection period,so they are not suitable for routine methods of forage quality detection.In recent years,with the development of deep learning and multi-source heterogeneous data fusion technology,great progress has been made in the processing of multi-source heterogeneous data.Aiming at the above problems,a forage quality evaluation system based on multi-source heterogeneous data was researched and developed.The main research work is as follows:(1)Design an Image-Based Categorical Hierarchical Perception Model.The maximum pooling layer is replaced by the global average pooling layer,the structure of the VGG16 Network is improved,and the parameter amount of the fully connected layer is reduced.The activation function is placed before the last convolutional layer to improve the structure of the Res Net34 Network to ensure the integrity of the original information.Using two improved networks and categorical hierarchical method to train the model to determine forage categories and quality grades.The experimental results show that the improved network has higher training accuracy and lower loss than the original network,and the categorical hierarchical method is 4% higher than the model accuracy of the direct classification method.(2)Design a Meteorology-Based Forage Quality Perception Model.A fuzzy logic inference algorithm based on meteorological information is proposed,a fuzzy rule base for meteorological data and forage quality grades is established,membership functions are determined,and the inference of forage quality grades is realized;a sparse autoencoder data fusion algorithm based on clustering is proposed,which uses The meteorological data features are extracted from the encoder,the centroid of the data features is selected,the Euclidean distance between the meteorological data and the centroid is calculated,and the K-Medians clustering algorithm is used to cluster the meteorological data to complete the classification of forage quality.The results showed that the two forage quality perception models based on meteorological information could both determine the forage quality grades through meteorological information.(3)Design a Heterogeneous Feature Fusion Algorithmic Forage Quality Evaluation Model.A fuzzy neural network was designed to combine the output results of the categorical hierarchical perception model with the output results of the meteorology-based forage quality perception model according to different weights to perform decision-level fusion,output the decision results,and complete the comprehensive evaluation of forage quality.The experimental results show that the loss value of the heterogeneous feature fusion algorithmic forage quality evaluation model is lower,and the forage evaluation results are more accurate.On this basis,an intelligent monitoring and evaluation system for artificial grassland grass quality is developed,and the above model is embedded in the system to realize the system function.In the actual test,the accuracy rate can reach 94.6%,which has good practical value.