| In recent years, a new upsurge has been set off in the development of marine economy in Chinese coastal areas. Reclamation is the hot means of sea usage in the coastal zone. It brings significant social and economic benefits for coastal zone, as well as it may result in serious negative impacts on nearshore marine environment. Adopting the scientific technical means to investigate of sea usage can provide scientific basis for the managers who keep these on the right lines. So remote sensing technology is introduced to the marine resources survey, and has become the powerful technical means for coastal investigation. By comparison with traditional methods, the remote sensing technique has the macroscopic, fast, timely and dynamic characteristic, as well as wide applicability.On the other hand, remote sensing technology has developed continually. Accordingly, its imaging capability is increasing which performs in higher spectral resolution, higher spatial resolution, higher temporal resolution and higher radiance. The high-resolution RS image contains a wealthy information on spatial structure and geographic features. It also has more noise and spectral confusion. The traditional classification method based on pixel can’t do well with high-resolution image and its precision can’t meet the demand. In order to improve the classification accuracy, the object-oriented image analysis technique is used to do the classification work. The object-oriented technique has stronger capability to deal with noise, and it can introduce more abundant image features and expertise knowledge in analyses. The first and most important step of the object-oriented technique is image segmentation, which segments an image into many visual homogenous parcels. Based on the parcels, more features can be extracted, such as shape, spectra and spatial relation, which facilitate the succeeding image interpretation.A new two-level classification method is designed and applied for high-resolution RS image in the coastal zone. This method is also based on object-oriented image analysis technology. Our work includes two parts as follows:(1) Combining with the characteristics of ground objects and the place where the reclamation located in, a classification system has been established for reclamation projects. Based on the classification system, a proper and two-level classification workflow has been designed.It is applied to the extraction of the target within the reclamation project area. The process as follows:Supervised classification is done with spectrum feature as the leading property, which is the first-level classification. The classification result includes water, building land, tidal flat, vegetation and non-classified field. Then, based on the front classification result and features extracted from parcels, knowledge rules are established and used to extract aquiculture water and dam further. This is the second-level classification.(2) In the second stage of classification, an approach is designed for the extraction of dam in the combination with line and parcel. The detailed introduction of the approach as follows; Firstly, taking first-level classification result as the background, the knowledge rules (a combination of spectral features, shape features and spatial relationships, etc.) are applied to extract dam.Secondly, the dam parcels collection of pre-extraction is expanded by the adjacent parcels and the new dam has the constraints of main direction and width to avoid tidal flat being extracted as dam. Thirdly, taking the dam parcel extracted as the starting point, the edge is tracked and the new dam parcels are found. The edges of image are detected with Canny operator. Experiments indicate that the accuracy of the approach proposed above is much higher and the dam extracted is more continuous. |
PDF Full Text Request