Classification Of Crops Based On UAV Remote Sensing Images

Comprehensive and accurate access to crop classification information is of great significance for the government to formulate food policy management production,adjust agricultural industrial structure,and ensure national food security.And it's important for crop acreage,crop growth monitoring and yield estimation,agricultural disaster monitoring and eco-environmental information monitoring.In addition,obtaining high-precision crop classification has gradually become one of the important scientific issues in agricultural land system research.With the development of UAV platform and photogrammetry technology,UAV remote sensing provides new way for crop classification survey.At small and medium scales,UAVs can give full play to their advantages.UAVs have many advantages,such as flexible operation,simple operation and labor saving,etc.It can obtain ultra-high resolution remote sensing images,which is of great significance for the development and application of crop detection technology,and also provide ground verification for satellite remote sensing.Therefore,it is necessary to obtain crop classification information using remote sensing images of UAVs.In this paper,a method for extracting accurate classification information of crops using the combination of spectral,texture and spatial features of UAV images is studied,and the influence of uav image resolution on crop classification accuracy and efficiency is analyzed.First,we conducted research area selection and ground truth survey,used UAV remote sensing system for visible image acquisition.Secondly,based on visible light image and digital surface model,the spectral,texture and spatial optimal classification features of crops were extracted.The support vector machine was used to classify the remote sensing images of UAV.Finally,based on the support vector machine classification method,the crop information of different resolution images is obtained,and the classification accuracy and classification speed are compared,and the influence of uav image resolution on crop classification accuracy and efficiency is analyzed.The following contents are studied and realized in this paper:1,The acquisition and preprocessing of the UAV remote sensing image were carried out in the study area.The fixed-wing UAV was used to collect visible light images,and the collected images were preprocessed to obtain the spliced visible orthophotos and digital surface model images.2,Crop classification research based on multi-features of UAV remote sensing images.Firstly,we confirmed the classification of crops in the study area and calculated the visible vegetation index and texture filtering for visible light remote sensing images.Secondly,we analyzed the characteristics of digital surface model(DSM)data and performed differential processing on the two phases of DSM data to obtain the difference digital surface model(DDSM).The model data was used to extract the height information of the crops,and the difference digital surface model was filtered according to the canopy characteristics of the crops.Finally,the feature optimization and combination are carried out,and the SVM method is used to classify the crops,and finally the classification features were determined as RGB,R-band contrast,G-band second moment,B-band variance,DDSM,DDSM variance and DDSM contrast.The accuracy of classification using spectral,texture,and spatial features increased from 71.86% to 92.30%.3,Preliminary analysis of the influence of uav image resolution on crop classification accuracy and efficiency.Firstly,re-sampling was used to generate remote sensing images of different resolutions.Secondly,the support vector machine classification method was used to study the crop classification.The crop classification accuracy and time of different resolution images were calculated.The results show that with the decrease of image resolution,the classification accuracy of crops is reduced,and the effect of texture and spatial features on classification accuracy is reduced.The necessity of spectral,texture and spatial combination features in crop classification based on UAV remote sensing high-resolution images was clarified,and the spectral,texture and spatial combination characteristics were verified to improve the classification accuracy of crops.