Study On Classification Method Of Main Crops Based On Irrigation Area Of UAV

The cultivated area of the crops in Sheng Wu Village,Huinong,Ningxia District,is about 3,000 mu.It is time-consuming and labor-intensive to calculate the distribution information of cultivated crops through artificial field statistics.Based on this,UAV remote sensing technology is used to classify the main crop types in the study area.In order to improve the accuracy and processing efficiency of crop classification,the data dimension is reduced by the best index method and feature extraction method,and the classification rule set is constructed to establish supervision.The classification model classifies the main crop types in the irrigation area.The main research contents and conclusions of this paper are as follows:(1)To solve the problem of high spatial resolution,redundant information,and strong correlation between the wavelengths of UAV multi-spectral images,combined with spectral information,texture information,and elevation information of UAV remote sensing images,the best index method and the maximum minimum value normalization method are used.Statistical analysis of each image,The best band combination of Red,Nir,Green spectral features,NDVI index features,red band mean texture features,near infrared band information entropy texture features,green band correlation texture features and elevation features were selected as the best band combination for drone remote sensing image classification.Crop classification in research areas.(2)Using pixel-based supervision method,combining crop spectral information,texture information and elevation information in images,constructing classification rule sets,establishing maximum likelihood classification model(MLC),support vector machine(SVM)classification model,BP neural network Network classification model.The results show that the overall accuracy of the classification model established using different feature combinations is higher than 75%,and the Kappa coefficient is higher than 0.61,which has a good classification effect on each crop in the research area.Under the combination of spectral+texture+elevation characteristics,the overall accuracy and Kappa coefficient of the classification results were the highest.Taking into account the overall classification accuracy,Kappa coefficient,and computer processing time of each classification method under the combination of spectral+texture+elevation characteristics,Determining the maximum likelihood classification method based on spectral+texture+elevation feature combination can be used as the best classification method for major crops in the research area.(3)The object-based multi-scale segmentation method is adopted.Under the eCognition workspace,the segmentation parameters are continuously changed,and the visual discrimination is used as the standard.The multi-spectral data of the UAV is determined to be 15 and the shape weight is 0.05.The weight of the weight is 0.5,which is the best segmentation parameter.The local objects are completely segmented to eliminate the salt and pepper phenomenon.The confusion matrix is used to evaluate the accuracy of the classification results.The results show that the overall scale of the multi-scale classification method based on the combination of spectral+texture+elevation features The accuracy is 92%,the Kappa coefficient is 0.8891,and the high-performance computer processing takes 0.79h.The classification accuracy is much higher than that of the pixel-based classification method.(4)Comparing the area of each crop under each classification result with the area of each crop under visual interpretation,the error of each crop area in MLC,SVM and multi-scale segmentation based on spectral+texture+elevation feature combination is lower than 20%indicates that drone multispectral image data and digital surface model(DSM)data can be used to extract irrigation crop types.