Study On Classification Of High-Resolution Remote Sensing Image Base On Hierarchical Multi-scale Segmentation

In recent years,remote sensing science and technology have been significantly improved in the aspect of hardware and software.How to obtain more accurate and higher-quality ground information from images has always been a focus of attention by researchers at home and abroad.Feature information has always been an extremely important information in remote sensing imagery.For earlier satellite imagery,the image resolution is low,the details of the feature itself are blurred,the relationship between features is not obvious enough,and the acquisition of feature information is therefore limited.At present,remote sensing images have greatly improved in resolution,More high-resolution remote sensing images were acquired by people.In the high-resolution remote sensing image,the relationship between the details of the ground features and the ground features becomes clearer,which provides an excellent research basis for the extraction of ground object information in remote sensing images,and also reflects the research value of high-resolution remote sensing images.The high-resolution remote sensing images have plentiful and detailed information of the ground objects,but at the same time,the structure of the ground objects is more complicated and the interference information is more difficult to handle.The traditional feature extraction methods for medium and low resolution images are no longer suitable for high resolution images.Therefore,this paper studies the object-oriented technology based on hierarchical multi-scale segmentation and uses this technique to extract the feature information in high-resolution images.During the research process,using GF-1 image as the experimental image,the advantages of high spatial resolution of high-resolution remote sensing images are fully combined to improve the extraction accuracy of feature information.Get accuracy.Finally,this paper achieved a more ideal result in the extraction of feature information.The main contents of this paper are as follows:1)In order to improve the image quality of high-resolution remote sensing images,the fusion technology of remote sensing images was studied.Several common image fusion methods were tested,and the applicable scope and characteristics of several fusion methods were compared through experiments.Finally,the Gram-schmidt fusion method is used to preferentially fuse high-resolution panchromatic band images and the multi-spectral medium-resolution images.The results show that this fusion method can greatly enhance the quality of the fusion image.2)In order to improve the effect of high-resolution image segmentation,this paper compares and analyzes four commonly used edge detection operators through experiments,selects improved Canny operator which is the best one to detect image edges,and combines edge information with multi-scale segmentation.The edge information is used as a reference in the multi-scale segmentation,and better to utilize the shape features of the image.The results show that the segmented image is more in line with the original shape.3)Researched the important content of hierarchical multi-scale segmentation techniques,using object-oriented classification based on multi-scale segmentation as the main method,establishing rules for the five types: vegetation,buildings,roads,waters,and naked ground in the experimental area,then extract these types for classification.The classification results of the ground features were evaluated by calculating the confusion matrix.The evaluation results showed that the Kappa coefficient of the classification results reached 0.918,and a good classification effect was achieved.Finally,we compare the experimental results based on object-oriented standard nearest neighbor classification.The results show that the method of this paper is better than the standard nearest neighbor classification.