| Inland waterway, the main vehicle of inland navigation, affects the development of industrial belt along the river as well as the development of our country. It is crucial for the inland navigation to provide accurate channel information. At present, the acquisition of inland channel information mainly depends on manual measurement which is not only time and energy-consuming, but also of low efficiency and instantaneity. With the advent of information era, it is necessary to realize channel information management. The traditional method of acquiring channel information cannot meet the need, thus it is urgent to find a more efficient and accurate way to acquire channel information.Remote-sensing image refers to the information of objects from the target areas which is recorded by remote sensing system with the help of remote sensor. It mainly refers to satellite photograph and aerial photograph. Remote-sensing image provides real-time and comprehensive ground information of monitoring areas which would provide decision-maker favorable data support. Remote-sensing image enjoys high instantaneity, wide coverage, short update cycle. With solid processing technology,remote-sensing image can reflect the dynamic change of the target area. Thus, the present dissertation aims to acquire inland channel information of the areas covered by remote-sensing system through the processing of remote-sensing image.The main works are as follows:First, preproccess the remote-sensing image and remove the noise of fog and cloud in it. On the basis of analyzing remote-sensing image cloud degradation model, remove the fog and cloud with dark channel prior. The present research makes an attempt to improve dark channel priority and gains a good result.Second, analyze the spectral characteristics of all ground objects like: water,vegetation, rick-soil, urban architecture and roadway in remote-sensing images.Information of water body can be extracted by its spectral characteristics which are different from other ground objects. After a contrast of all effects of all kinds of extraction, the study chooses normalized difference water index to extract information about water body. Then, the result of extracted water body information has been modified by mathematical morphology. This is a preparation for the extraction ofchannel information.Third, choose appropriate geometric eigenvalues as decision attribute of decision tree algorithm after an analysis of inland channel’s characteristics. The study proposes a new automatic generate algorithm of decision tree, based on the inland channel’s threshold. Then, the image of water body is classified by this decision tree and then a map of inland channel has been produced.Finally, the inland channel can be marked on remote-sensing images. |
