Research On Object-oriented Land Use Classification Technology Based On Multi-source Image Fusion

Accelerating the development of new-type urbanization is an inevitable requirement for the economic and social development of the Corps.Using remote sensing data to obtain land use information and agricultural profiles of the farm field can not only provide the basis for government decisions,but also a necessary condition for the development of agricultural informatization.It can improve efficiency,save manpower and resources,and make agriculture develop more scientifically.With the continuous improvement of the quality of satellite products,the information contained in high-resolution satellite images is becoming more and more abundant.The performance of a single sensor satellite data in terms of time,spectrum,and spatial resolution.Due to the different characteristics of the sensor itself,multi-source image fusion can synthesize the favorable information of different satellite sensors,so that the fused image can improve the utilization of the original image,can also increase the scope of application of remote sensing images.In the past,the method of image classification was to complete the classification based on the spectral characteristics of the image pixels.This method ignored other information contained in the image,not only the low utilization rate of the image data,but also generated great waste.Aiming at the problems existing in the process of multi-source image fusion and object-oriented classification,this paper is based on the preprocessing of ZY-3 multi-spectral imagery and the full-color remote sensing image of KZ-1,multi-source image fusion method,image segmentation parameter determination,Research on classification and accuracy evaluation,the main work is as follows:(1)Research on different parameters,characteristics and image resolution of different remote sensing satellite sensors.The pre-processing work of the multi-source remote sensing image data is systematically studied.The pre-processing work of image radiation calibration,atmospheric correction,and image registration is mainly completed to ensure the consistency of the pixels between the original images and provide a data basis for further image fusion.(2)Research on multi-source sensor image fusion methods.In order to explore a multi-source data fusion method suitable for the ZY-3 multispectral image and the KZ-1 panchromatic image.In this paper,six common pixel-level fusion methods,such as PCA fusion,Brovey fusion,Gram-Schmidt fusion,NNDiffuse fusion,Subtractive fusion,and Pansharp fusion are used to perform fusion experiments on resource three multispectral image and the fast boat one panchromatic image.And improve the super-resolution Bayesian algorithm to improve the image quality after fusion.After a comprehensive comparative analysis of visual interpretation and quantitative evaluation(spatial information integration,spectral information retention,and clarity)indicators.The improved super-resolution Bayesian algorithm is more suitable for multi-source fusion of resource three remote sensing image,which not only maintains the spectral information of the image,but enhances the detailed texture characteristics of the remote sensing image.(3)Research on object-oriented image segmentation experiments.The object-oriented classification method is used to make full use of the spectral feature information and spatial feature information of ground features in remote sensing images.Using the remote sensing images of the 125 mission research area acquired by the drone to generate sample vector files to verify the experimental classification results.The results show that the overall classification accuracy of the experimental area is 84.75%,the Kappa coefficient is 0.8248,the overall classification accuracy of the experimental area II is 84.42%,the Kappa coefficient is 0.8145.The overall classification accuracy and Kappa coefficient are higher than 80%,which can effectively understand the study area Land use profile.In this paper,ZY-3 and KZ-1 remote sensing images are used as data sources.Based on multi-source image data,through pixel-level image fusion,a fusion algorithm suitable for ZY-3 multi-source image fusion is discussed.Using the object-oriented image segmentation technology to extract the ground feature category of the fusion image,it effectively extracted the land cover utilization in the research area,which can provide a theoretical basis and scientific guidance for land resource allocation and water resource management.