Low-altitude Remote Sensing Image Processing And Analysis Of UAV Based On Visible And Multi-spectral Images

Agricultural monitoring,such as nutrition monitoring,land monitoring and pest monitoring,is a monitoring of crop growth.More efficient implementation of agricultural monitoring allows decision makers to quickly obtain crop growth,crop pests and diseases and other information,so that decision makers quickly make better management decisions.Reducing the use of pesticides and fertilizers in the agricultural production process is the requirement of information agriculture and precision agriculture.At present,domestic and foreign applications on the monitoring of human resources in UAVs are generally focused on one aspect,and there is little research on the generation from the system design to the application guidance map.Therefore,in this study,low-altitude remote sensing images of visible and multi-spectral UAVs were studied,to build a low-altitude remote sensing image acquisition platform,and a series of studies,such as image correction,image mosaic,image classification and generation of prescriptions.The main results are as follows:1.To set up low-altitude remote sensing acquisition platform for UAV based on visible and multi-spectrum;according to the characteristics of visible and multispectral cameras,remote sensing image acquisition and control circuit were designed,and build image transmission module.2.The correction of visible light low-altitude remote sensing image is studied.By comparing the characteristics of two kinds of splicing correction algorithms,a simple image correction method of multi-plate method is selected.3.The stitching of UAV low-altitude remote sensing images is studied.According to the remote sensing image obtained by shooting,the relationship between the ground resolution and the shooting height of the two cameras is deduced by referring to the data and designing the experiment.The simple GPS conversion method is used to realize the low altitude based on the geographical coordinates at the same height remote sensing image splicing,and conducted error analysis.4.The feasibility of low-altitude remote sensing image acquisition platform is studied.By using the visible light low-altitude remote sensing image acquisition platform to collect the visible light image,after the correction,by analyzing the statistical characteristics of the visible light band,using four visible light vegetation indexes for vegetation extraction,the maximum extraction accuracy of up to 95.8%,the feasibility of low-altitude remote sensing acquisition platform for visible light is verified.5.The feature extraction of low-altitude remote sensing images of cotton visible and multi-spectral UAVs is studied.The extracted features include band statistical features,vegetation index characteristics,and Tamura texture features.Finally,through feature screening;A 9-dimensional classification feature vector is constructed for visible image;An 8-dimensional classification feature vector is constructed for multi-spectral images.6.The classification method of low-altitude remote sensing images of cotton visible and multi-spectral unmanned aerial vehicles is studied.According to the selected characteristics,we input into K subspace classifier,SVM support vector machine classifier and Adaboost lifting tree classifier respectively,the results show that the classification results of the Adaboost lifting tree classifier for low-altitude remote sensing image classification of UAVs are the best,and the classification accuracy is 83.8%.