The Extraction Types Of Land Use/land Cover With High-resolution Remote Sensing Image

With the rapid development of the remote sensing technology, people get more andmore spatial information, especially for high spatial resolution remote sensing image,has become a research field of city environment information related to an importantremote sensing information source. The land use/land cover types of extraction ofremote sensing image information extraction is an important part in. But the traditionalland use types extracted from the most use the manual interpretation method. Thismethod is not only time-consuming and laborious, and by the interpreter professionalknowledge and subjective judgments of the impact is bigger, has been increasinglyunable to meet the processing needs, so the object oriented remote sensing informationextraction methods emerge as the times require.In order to further demonstrated in remote sensing image information extraction inobject oriented method is better than the pixel-oriented method, so this paper selectedDistrict of Chongqing area as the study area, in2009September by acquiringCBERS-02B star HR and multi-spectral image fusion as the research object, the imageresolution is2.5m. Using eCognition, Erdas and ArcGis9.3software, respectively, onthe land use/land cover types by using the pixel method and object oriented method toextract, and compare results.In this paper, facing the pixel method is mainly used Erdas software to undertakesupervised classification, which is used in the maximum likelihood method;unsupervised classification is used in iterative self-organization method. Object orientedmethod through comparing the different segmentation scale, shape index andcompactness index under the segmentation results and determine the final segmentationscale; established the extracted feature knowledge base; for the study area to classifyextracted; finally the extraction results are accuracy evaluationThrough the comparison of two methods for the extraction of land use/land covertypes of results, found object oriented method in process of extracting the knowledgebase into the extraction rules, the results of the extraction precision. The water,construction sites, roads, bare land and green space to extract correct rates were:92.54%,89.68%,88.70%,78.33%and98.17%. While the supervised classificationaccuracy, the pixel values of the percentage of comparison, water and soil extractionprecision is higher, respectively91.01%and3.06%. Thus, can see the object orientedmethod in the extraction of land has very good practicality. Through the further analysis, found the object oriented method in process ofextracting the knowledge base into the extraction rules, the results of the extractionprecision. The pixel extraction process only for imaging spectral characteristics tocomplete information extraction, so the extraction precision is not ideal. Therefore, theobject-oriented method either process or final extraction accuracy, are superior to thoseof pixel-oriented technique, is a kind of promising approaches to information extractionmethod.