Research On Registration And Change Detection Technology For Remote Sensing Images Of Coral Reefs In South China Sea

Facing the demand of detecting the changes to the coral reefs in South China Sea,the present paper studies the automatic registration method in dealing with pixel-level accuracy of remote sensing images at different scales by using the data from high-resolution images and moderate-resolution images for remote sensing images ofcoral reefs,so as to achieve the identification of changing coral reefs in large areas of Nansha Islands and the change detection of land cover of individual coral reef,and provide support for detecting and monitoring the changes to the coral reefs in South China Sea.With the development of earth observation,a considerable amount of remote sensing images of coral reefs,located in the same area but captured under different imaging conditions can be obtained,which provides adequate data for remote sensing monitoring of coral reefs.However,it also brings about a great deal of challenges.The traditional image registration method,handcraft method,hinders the full use of these numerous remote sensing images of coral reefs on account of its high cost,low efficiency and long response time.Thus to automatize the image registration of coral reefs possesses a great significance in the further detection of the changes to the coral reefs.In recent years,the region of South China Sea was invaded repeatedly and the coral reefs in South China Sea were occupied illegally by neighboring countries.To achieve the change detection to the coral reefs in South China Sea is of great importance in maintaining marine rights and interests as well as safeguarding national security.Taking these challenges into consideration,the present study,based on the remote sensing images of coral reefs with moderate resolution and high resolution,selects the Southwest Cay and Thitu Island,with the part of the Nansha Islands,as the cases to study,and puts forward a technical scheme to automatic registration and change detection by using different scales of remote sensing images.The contents and conclusions of the present study are as follows:(1)The automatic registration of moderate-resolution images of coral reefs.For the lack of texture features and local geometric distortion existing in moderate-resolution images,the paper studies from two stages.The first stage is global rough registration,and the second stage is local fine registration.In the first stage,the global rough registration is achieved through the geometric profiles of coral reefs so as to avoid the difficulties in registration caused by inadequate texture features of the images.In the second stage,the local fine registration is achieved through constructing local geometric transformation model of coral reefs determined by the corresponding local grey scale information of coral reefs so as to correct the results caused by local geometric distortion of images.(2)The automatic registration of high-resolution images of coral reefs.Since the automatic registration of high-resolution images has the characteristics of unstable texture features and over-concentration of texture features,the present study employs"controlling of spatial distribution—controlling of overlapping ratio" that hierarchical controlling filtrate strategies.In the first stage,removing those matching pairs which are inconsistent with the spatial distribution through calculating spatial model and the spatial dispersion quality of matching pairs.In the second stage,according to the overlapping ratio of the corresponding coral reefs,selecting the optimal threshold value automatically to randomly sample and filtrate so as to ensure that the precision of geometric transformation models constructed for registration is high.(3)The change detection and analysis of coral reefs.On the basis of automatic registration of remote sensing images of coral reefs,moderate-resolution imagery for remote sensing images of coral reefs is employed to identify the Southwest Cay and Thitu Island,in which there are marked changes among the various coral reefs,through change detection of the two coral reefs according to their differences in grey scale.In addition,high-resolution imagery for remote sensing images of coral reefs is used to classify the two coral reefs and compare their change detection so that the specific changes to the coral reefs can be found and analyzed.In conclusion,the present study focuses on some coral reefs in the South China Sea.To begin with,the algorithm proposed for the automatic registration of moderate-resolution images of coral reefs can deal with the difficulties that the images of coral reefs captured by this registration method lacks adequate texture features and possesses local geometric distortion.Adopting the combination of global rough registration and local fine registration can achieve the high precision of automatic registration for moderate-resolution images of coral reefs.Moreover,the algorithm put forward for the automatic registration of high-resolution images of coral reefs can do with the difficulties that the texture features obtained by such registration is unstable and these texture features are too concentrated.Conducting two filtrates—the controlling of spatial distribution and the controlling of overlapping ratio of the coral reefs,to achieve the high precision of automatic registration for high-resolution images of coral reefs.Finally,the current study successfully detects the overall changes to coral reefs through registration of moderate-resolution images and then specializes in detecting more specific changes to the land types of those coral reefs,.That is "automatic registration of remote sensing images-identification of changing coral reefs—retrieval of information about the changes",which possesses reaching potential to be applied in this field.