Research On Classification Method Of High Resolution Remote Sensing Image And Application In Landscape Pattern Analysis

With the development of computer hardware and remote sensing science and technology,the spatial resolution and temporal resolution of remote sensing images have gradually increased.High-resolution remote sensing images are widely used in land resource planning and ecological environment protection because of their complex features and spatial information.However,the rich texture and spatial information content and the complexity of the complexities make the processing and analysis of remote sensing images more complicated.At the same time,there are few high quality training sample data in the field of remote sensing classification and it is difficult to obtain.Therefore,how to use existing empirical data effectively and explore new classifiers to improve the classification accuracy of high-resolution remote sensing images is an inevitable requirement to meet the urgent needs of current land use and urban planning.The main research contents of this article are as follows:(1)Combining high-resolution remote sensing images of Baohe District in Hefei City in 2016,using ensemble learning method which popular in recent years,such as Bagging,AdaBoost,and Random Forest,to classify high-resolution remote sensing images based on pixel,and compare it with traditional individual classifiers such as support vector machines(SVM)and decision trees to evaluate the effectiveness of various classification methods with the accuracy of results.It is proved that the multi-classifier integration scheme is indeed superior to a single classifier in the classification of high-resolution remote sensing images.(2)Based on the classification results of the ensemble learning for the classification of the remote sensing imagery in Baohe District,Hefei City.Using the traditional landscape pattern index to analyze the changes and researching its driving factors.(3)Combining high-resolution remote sensing image datasets of Baohe District in Hefei City in 2016,a scene classification experiment based on convolutional neural network migration learning was conducted.By using the pre-trained Inception-v3 model to migrate the scenes of the ImageNet image annotation data set in the study area,the great advantage of the method based on the high-resolution remote sensing image data set was verified.(4)Analogy between the scene classification results and the pixel classification results,and discuss the application of scene classification results in landscape pattern analysis.