Research And Application Of Calculation Techniques Of Grape Planting Area Combining Remote Sensing Data With Meteorological Data

In recent years,with the continuous launch of remote sensing satellites,more and more remote sensing data has been applied to the extraction of crop acreage.The key to extracting crop acreage using remote sensing data is to identify crops.The selection of the best crop phase can greatly improve the accuracy of crop information extraction.Optimal phase refers to the period when the crop growth season corresponding to the crop phase can be distinguished to the greatest extent when it is classified based on crop spectral information.The period of crop growth and development can be calculated from the accumulated temperature experienced by the crop development process.Therefore,using the meteorological data,the period corresponding to the best crop of the year can be obtained by the method of estimation of the accumulated temperature of the crop development period,and the remote sensing image within this period can be deduced.The date of transit,so as to quickly determine the remote sensing image data within this date range,this is the best time remote sensing image data.In terms of crop planting area extraction,traditional methods have problems such as large human interference,low accuracy and extraction efficiency,and incompatibility of indicators between different time-phase images.In this thesis,the wine grape planting area in the eastern foot of Helan is taken as an object.Using remote sensing data combined with meteorological data,a deep learning and transfer learning fusion algorithm is constructed to classify crops from remotely sensed images of wine grape growing areas and then extract the area.Finally,based on C #language and Arc Engine,the planting area extraction algorithm and software development were implemented.The main work of this paper is as follows:(1)Confirmation of the best phase.This paper uses the MODIS data to extract the NDVI index and combines the accumulated temperature data in meteorology to build a NDVI accumulated temperature inversion calculation model,which can quickly obtain the extracted wine grape information.The best time.(2)Remote sensing imagepreprocessing,the remote sensing image obtained in the best phase is subjected to a series of processing such as preprocessing,sample labeling,segmentation and data enhancement to construct a remote sensing classification data set of grape growing area as the target of deep migration learning.Domain dataset.(3)Construction of migration learning model,a remote sensing image classification algorithm based on full convolutional neural network is proposed.Combined with migration learning,the first thirteen layers of VGG-16 network are used as feature extraction of the full convolutional network model constructed in this paper.Layer,and using the model trained on the large source dataset to migrate to the deep neural network model constructed in this paper,training the remote sensing classification data set of the grape growing area(2),and then testing the trained network.Extract the grape growing area.(4)Planting information extraction,constructing the sample,using the manual visual interpretation of the actual grape growing area in the sample,and using the deep migration learning constructed in this paper to predict the grape growing area within the sample,and then extracting the area of the two results.And then the area extraction accuracy is obtained.(5)Platform development,using the C# language combined with ArcEngine to develop the grape planting area extraction system on the Visual Studi2012 software platform.