Based On High Way Of Remote Sensing Image Information Rapid Extraction Technology Research

Because of its high resolution remote sensing image contains rich geographic information, so the high resolution remote sensing image in basic geographic data update,the information of land and resources survey and land use change detection is widely used,etc. With significantly improve image interpretation technology, explore how to quickly and accurately in high resolution remote sensing image to extract target feature information,as people are constantly thinking and exploration direction. Road as an important national categories and basic geographic entities, road can accurately extract to build national and geographical spatial database related geographic information industry to lay a solid foundation. Road extraction is mainly based on the traditional way of target pixels, to obtain the feature information, it is just using the spectral information of image, and was not used to feature space structure, texture, and other information. Based on high resolution remote sensing image to target features the reaction feature details, target feature of the texture, color, brightness and other characteristics, adopting the object-oriented method in the study area within the scope of the path information. This article chooses Songhai Economic and Technological Development Zone GF-1 satellite remote sensing image as data sources, adopting the object-oriented method to extract the road in the information.First of all, through trial and error method to determine the scale parameter of a variety of scale segmentation, after appropriate parameter, use a variety of scale segmentation method to get several attributes with a consistent image area; Then according to the spectrum, shape, texture, such as characteristic information, obtain the knowledge base of road; Finally under the rules of the knowledge base is set up, use a computer to automatically classify the segmented regions, and the experimental results are obtained.Through the experimental study summarizes the test results are as follows:(1)Based on eCogniton software research platform, a variety of scale segmentation,quadtree segmentation method, the partition board method three segmentation methods such as comparative analysis, obtained by different segmentation method of target feature area after field verification and artificial extraction and data space analysis, found the highest a variety of scale segmentation accuracy, better than the other two methods.Through many times of test area segmentation contrast, determine the way of case study area is 60, the optimal segmentation scale shape difference factor is 0.3, compact shape factor of 0.4.(2)According to the experiment to determine the optimal segmentation scale, scale segmentation research area, and classify image segmentation. In this paper, we adopt the method of fuzzy classification and fuzzy classification was introduced in detail in the adjacent taxonomy and classification of membership function two classification methods,according to the comparison and analysis, finally chooses the classification methods to classify subordinate function. For road on the images reflect the mean different wavelengths, brightness, length-width ratio of characteristics, to establish a road image region classification rule set, with the help of relevant rules for automatic classification of roads with high score finally remote sensing image are extracted, and achieved good results.(3) Case, according to the path information extraction based on object oriented image road extraction work, illustrates the concrete operation of the method, multi-scale segmentation and fuzzy classification in the concrete utilization, and compared with manual extraction result analysis, on the basis of comparative analysis, object-oriented image information automatic extraction method, the advantages and disadvantages, finally summarizes the shortcomings of this research and to carry out the research direction in the future.