Font Size: a A A

Research On Information Extraction Of Mountainous Rivers And Bridges In Xinjiang Based On Gf-1 Image

Posted on:2016-04-07Degree:MasterType:Thesis
Country:ChinaCandidate:Y H LiFull Text:PDF
GTID:2308330476450281Subject:Geography
Abstract/Summary:PDF Full Text Request
The starting star of Chinese high-resolution earth observation system-“GF-1” launched successfully on April 26, 2013, which open a new era of Earth observation for China. Therefore, it has scientific and practical significance to strengthen research on GF-1 remote sensing image. Relative to the low-resolution images, high resolution images have the advantages of that feature’s geometry is more obvious, position and layout clearer, texture and size information more precise. Using the traditional information extraction method based on the pixels is unable to meet the application requirements of high-resolution images at the aspects of image information utilization, processing efficiency, extraction accuracy, acquired result information, and so on. Currently, using object-oriented methods to improve the accuracy and efficiency of the high-resolution image information extraction has become a hot topic in the field of remote sensing. The processing object of object-oriented information extraction method is the object that has a variety of semantic characteristics and relationship characteristics, namely "homogeneous" polygon, which is not an isolated pixel, so as to effectively avoid the influence of the "salt and pepper noise". What’s more, the object-oriented method can make full use of the rich information of high resolution image, such as spectrum, shape, texture, topology, and so on. Therefore it can effectively overcome the phenomenon of "the same spectrum of foreign object" and "different spectrum same thing".In this paper, some key technologies of object-oriented method is improved based on the understanding of object-oriented method, and applies this method to GF-1 information extraction. Tekes river basin in the mid-section of Tianshan mountains in Xinjiang is choosed as the typical research area. Aimed at the problem of data deficiency in investigating difficultly area in Xinjiang, and object-oriented method was used to extract the various project elements, mainly include the mountains small rivers and bridges. This article main research is as follows:1、Image segmentation is the base of object-oriented classification method, the segmentation quality directly affects the accuracy of information extraction. In this paper, multi-scale segmentation of object-oriented method is improved by combining Canny operator in the edge segmentation and multi-scale segmentation. The Canny segmentation result is involved in the multi-scale segmentation as a layer. The results show that the obtained image segmentation object conforms to the actual boundary of the features.2、Selecting scale parameter in Multi-scale segmentation has a direct effect on classification results. On the basis of the previous studies, this paper put forward a index considering the weight of each layer- Modified Ratio of Mean Difference to Neighbors(ABS) to Standard Deviation(MRMAS),to resolve common problem of "Less divided" and "over-segmentation" in object-oriented classification. Through the measure MRMAS, it can determine the optimal segmentation scale of a particular feature.3、For the prevalence of mountain shadow and water mixed problem in high resolution images, the object-oriented method is used to detect mountain shadow area. Taking advantages of multi-band combination information, put forward the Shadow Water Index(SWI), which separates shadows and water successfully, with a simple, fast, high-precision advantage. According to the characteristics of the mountain shadow, using three methods of the linear correlation correction method, linear correlation stretch based on principal component transformation and bands regression model to compensate information of shadow area. Through the comparison and analysis, the results show that the linear correlation correction method has the most advantage for the mountain shadow information compensation. It can not only enhance the brightness of shadow area, highlight the details of shadow area, and color is most consistent with the original image.4、Rule-based object-oriented method was used to extract the small water body in mountainous area to resolve the difficult problem of interpreting small linear rivers that develop in the special climate and geomorphic environment in Xinjiang. The method can consider abundant image information such as the spatial structure, topology relationship in addition to the spectral. Besides, supplemented by 30m×30m ASTER DEM data, it effectively eliminate the interference of mountain shadow and dark features to keep small water continuous and integrity. The results showed that the proposed methodology is able to integrallty extract small water bady of hilly area with overall accuracy more than 90% and Kappa coefficient more than 85%. Finally, applying the morphological dilation filtering and Pavlidis asynchronous thinning algorithm to reprocess the extracted tiny water for getting the small river vectorization. Therefore, this study could provide some scientific information for the development and application of China-made “GF” remote sense image processing system.5、High-precision extraction of bridge information is of great significance for civil, military and commercial. For the difficult problem of high-precision extracting the bridge information, using rule-based object-oriented method and combined with mathematical morphology filtering algorithm, realized the accurate extraction of the bridge target. The results showed that the method is successful to extract Bridge information with accurate positioning and high precision. Therefore,it is able to provide certain scientific reference for river basin water resources investigation, flood control and disaster mitigation, or water conservancy planning work.
Keywords/Search Tags:object-oriented method, GF-1, shadow detection, small river, bridge
PDF Full Text Request
Related items