Estimation Of Feeding Intensity Based On Track Data Of Grazing Sheep

In order to quickly and efficiently grasp the spatial and temporal distribution of the feeding intensity of grazing sheep,this thesis proposed to use the Global Positioning System(GPS)to design a Positioning collar and data acquisition System to obtain the track data information of grazing sheep.BP neural network method and regression analysis method were used to analyze the factors affecting the intake intensity,and the advantages and disadvantages of the two methods were compared.The temporal and spatial distribution of feeding intensity of sheep at different locations and different grazing time were determined by grid analysis and buffer analysis,aiming to obtain the regularities of grazing behavior of sheep through the temporal and spatial distribution of feeding intensity.It mainly includes the following research contents:(1)Design of GPS collar.The function of the GPS positioning collar required by the test was studied,and the GPS positioning collar was made.The GPS module was used as the data acquisition contact point to collect the positioning data of free-range sheep.The GPRS module is used as the data transmission unit,and the collected data information is transmitted to PC to realize the visualization of grazing information.(2)Overview analysis of the test area.Through the analysis of the terrain condition by the elevation data,it is found that the slope of the test area is mainly in the range of0～6.5 degrees.The altitude of the test area is between 1380 and 1450 meters above sea level.The normalized vegetation index(NDVI)of the test area was obtained by atmospheric correction and radiometric correction of the remote sensing data.(3)Study on influencing factors of feeding intensity of sheep based on BP neural network and linear regression method.To establish the BP neural network model and linear regression model,will be the duration of the track section,and only walking distance,sheep sheep’s weight,experimental zone of slope,slope direction,elevation,temperature and weather conditions,vegetation index as a feature vector,through the analysis of single factor and multiple factors,using the BP neural network algorithm and linear regression algorithm to explore its internal relations,By measuring the correlation coefficient(~2)and mean square error(MSE),the correlation between each feature vector and feeding intake intensity was analyzed,and the main factors affecting feeding intake intensity were studied.The analysis proved that:Vegetation index,track duration,sheep body weight and ambient temperature were the most important combinations affecting feeding intensity.In linear regression analysis,the performance of this combination was~2=0.90,MSE=1.46;In BP neural network analysis,the performance of this combination is~2=0.97,MSE=0.0048.After analyzing the correlation coefficient and mean square error of the two methods,it is concluded that BP neural network analysis is the most suitable method for this study.(4)Study on the distribution of feeding intensity of free-range sheep.Buffer analysis and mesh partitioning were carried out on the track data by system tools.The behavior of sheep was identified and classified by kernel density analysis method,and the track points that would not affect the feeding intensity of sheep were deleted.The distribution information of feeding intensity of sheep was established on each independent grid.The neural network was used to fully consider the combination of factors that had the greatest influence on the feeding intensity of sheep,and the existing feeding intensity was revised.The range of feeding intensity of weekly intake of sheep in the experimental area was obtained from 0 to 133.19g/m~2.The intensity of monthly intake in the experimental area ranged from 6.81 to 730.12g/m~2.In this study,the primary and secondary relationships of the factors influencing the feeding intensity of sheep were analyzed,and the temporal and spatial distribution of the feeding intensity of sheep was obtained by using the grazing track data of sheep,which provided a theoretical basis for the balance of grazing and livestock in the pasture.