Research On Feature Extraction And Change Detection Of Island And Coastal Region

Island and coastal region is rich in natural resources and has a developed economy. At the same time, due to sea development and urban expansion, island and coastal region changes rapidly in land use. Therefore, it is of great significance for ensuring sustainable development and utilization of coastal region to grasp coastal resource distribution timely and precisely. Remote sensing technology has an advantage of a large area of simultaneous observation, objective and comparable data, time-sensitive, and high economic and social characteristics. And it has been widely used in the field of coastal resource monitoring. Not only has it become a hot topic of concern workers, but also the starting point of this article to extract feature of island and coastal region from remote sensing images.In this paper, it adopted three SPOT5 satellite images in Qingdao Red Island area as original data, data of SRTM3 and tide data of Huangdao Port as auxiliary data. Based on object-oriented and knowledge classification, it extracted feature accurately and conducted dynamic change analysis. The main research work was as follows:(1)In order to improve results of image segmentation, it vectorized road information from three remote sensing images in 1:3000 scale, and regarded road as thematic layer to participate in multiresolution segmentation.(2)Multiresolution segementation, one method of object-oriented, was used for images segmentation optically. In this paper, to gain appropriate parameters, it set various parameters, such as shape weights, compactness and scale parameter to segment images. From segmented images, it concluded that it was suitable to establish three split layers, road layer, layer with scale of 500 and layer with scale of 120, under the condition of using thematic layer of road, shape weigh being 0.1, compactness being 0.5.And road layer would be used for extracting road, layer with scale of 500 would be used for extracting seawater, layer with scale of 120 would be used for extracting grass, wetland, buildings and so on.(3)It adopted classification based on knowledge to classify the three remote sensing images. Through a series of experimental studies, knowledge of various land types on remote sensing images was established. Road can be extracted by density, seawater can be extracted by GLCM Homogeneity or MNDWI, wetland can be extracted by MNDVI or MNDWI, vegetation can be extracted by NDVI, buildings can be extracted by GLCM Dissimilarity and Mean SNIR. After accuracy assessment, it is obvious that experimental method is of high precision and classification results are consistent with visual interpretation. The overall accuracy of classified image is up to 97.80 percent and the lowest reaches 92.80 percent. Kappa coefficient is up to 0.970 and the lowest reaches 0.900.(4)Combined with tide data of Huangdao Port, it derived coastlines from waterlines. First, it obtained waterlines from classified remote sensing images according to object-oriented and knowledge classification. And it is clearly drawn that waterlines have continuous boundary, which separates wetlands and sea water well. Then, considered about tide data and waterlines, it concludes that coastline of sandy coast is the borders of ponds adjacent to sea water, and coastline of bedrock coast is the waterline.(5)By a variety of methods of change analysis, such as aggregate analysis, rate change analysis and transition matrix, it conducted change analysis of coastlines and classified images from remote sensing images and produced charts and tables from dynamic change analysis. The result showed that total length of coastlines increased, and it increased 727.9 meters from 2003 to 2007, 1957.79 meters from 2007 to 2011, length of sandy coast constantly decreased while bedrock coast constantly increased from 2003 to 2011; area of road and constructing area of buildings rose while wetland declined, and wetland mainly converted to buildings and other feature, which were caused by human factors; and values of comprehensive change rate of land use is large, and rate change from 2003 to 2007 reached 1.44 percent, while rate change from 2007 to 2011 reached 1.95 percent.