High Resolution Remote Sensing Image Information Extraction Method For Multi-Scale Segmentation And Case-based Reasoning

Object-oriented image analysis technology was suitable for high-resolution remote sensing image extraction,But because of its own segmentation and classification process involves more manual participation and human-computer interaction,high-resolution image of the large amount of data,the high resolution remote sensing image information automatically extract the automation and intelligent development was limited.Remote sensing image extraction automation and intelligent,less manual participation,is the goal of remote sensing application has been pursued.Based on case-based reasoning,this paper studies the multi-scale segmentation and case-based high-resolution remote sensing image ground cover classification information extraction method,focusing on multi-scale segmentation with multi-classification-level and multi-scale-level schame,reasoning case building and case matching search strategy.Combined with multi-scale segmentation and image masking technology,the structure of the case database and the structure of the database are designed,The high-resolution remote sensing image surface coverage system with multi-scale segmentation and case-based reasoning is realized,and good results are achieved.The main contents and achievements of the paper are as follows:(1)Multi-scale segmentation and case-based information extraction system for high-level image surface coverage classification was established.The system includes image segmentation scale and parameter determination,case expression model,case library structure,case feature selection and feature weight,similarity measure and case search strategy.The establishment of this system provides a new method for object-oriented high-resolution remote sensing image classification.Through the self-learning ability of the case,the intelligent extraction of remote sensing image surface cover classification information is realized.(2)Multi-scale segmentation scheme with mult-classification-level and multi-scale-level was designed.Based on the image mask technology,multi-scale segmentation scheme with multi-classsification-level and multi-scalse-level,the surface coverage classification objects with different classification level and different area levels can be divided into pure image objects.(3)Multi-scale case study of surface cover classification reasoning was constructed.According to the image features of different sensors,different spatial resolution and different scales,the image features which can best represent the characteristics of the surface cover classification are selected as the matching characteristics of the case and calculate the weight of the image.Metadata information and case matching information,as a supplementary feature of the case,the final formation of multi-scale surface coverage classification reasoning case library.(4)The multi-scale segmentation and case-based high-resolution remote sensing image surface coverage classification information extraction process was realized.Based on the multi-scale segmentation of eCognition software,C#and ArcEngine were used to extract the information of surface cover classification information based on multi-scale segmentation and case-based reasoning,which could realize case building construction,feature weight calculation,case feature selection,case matching Information extraction,image mask,case training and learning functions,to complete the high-resolution remote sensing image surface coverage classification information semi-automatic extraction.WorldView-2 satellite remote sensing image data and SWDC-4 aerial remote sensing image data were used to analyze the surface coverage information classification,and the experimental results were analyzed and evaluated,which proved the effectiveness of the high-resolution remote sensing image extraction method.