Research On Separation Of Ground Objects Based On Multi-spectral Marine Remote Sensing Image

The Technology of marine remote sensing, which is an effective means of observingreal-time ocean data, has been widely used in a variety of fields, such as marine navigation,ship positioning and stereo marine monitoring system. Especially in the field of marinenavigation, because remote sensing satellites are characterized by extensive coverage andhigh timeliness of work, the timeliness and accuracy of the obtained area, marine navigationconducted by synchronized remote sensing images, and ship shield barrier are greatlyimproved. However, marine remote sensing images contain various ground objects. Therefore,accurately identifying ground objects, especially efficiently separating the land, reefs and seawater, is an important part in marine navigation and ocean data monitoring and processingareas.In this paper, multi-spectral remote sensing image is the object of study and from twobasic characteristics of image segmentation, the discontinuities of edge and the region’ssimilarity，separation of ground objects is studied. Under the use of characteristics ofmulti-spectral remote sensing image, one kind of land and water separation method, which isbased onNDPI i, is provided and it gets a good effect of separation.First of all, against different ground objects in the multi-spectral remote sensing imagecharacteristics, the study analysis and selects Worldview-2remote sensing image as theexperimental data in this paper. The study has the necessary pre-processing analysis forremote sensing image in order to satisfy the accurate requirement of follow up processinganalysis of remote sensing data. Pre-analysis include radiometric and geometric correction,and radiometric correction includes radiometric calibration and atmospheric correction.Next, against the existence of band overlap and redundancy in multi-spectral remotesensing image, by using standard principal components transformation, the study analysismain components and extracts the effective band as the experimental data for ground objectsseparation. As the same time, taking into account the impact of the noise in remote sensingimage on the edge of the ground objects, useing the piecewise linear stretching algorithm toenhance the effective band edge.Then, this paper does the in-depth study about the ground objects’ separation algorithmin remote sensing images. From the angle of the edge operator, this paper studies the reefsextraction algorithm that is based on the edge gradient operator and the second derivativeoperator, by edge detection, edge stacking and post-processing to achieve the extraction of reefs; against the region’s similarity, this paper provides the land and water separationalgorithm which is based on the improvement of index. According to the analysis ofNDPI iof images, differences of radiation between water bodies are enhanced, using Otsumethod to obtain adaptive threshold, achieving water and land segmentation. Then, throughedge blending and polishing, that can achieve the result of water and land segmentation.Lastly, in this paper, multiple sets of simulation data and real multi-spectral imagespropose land and water separation method validation and the experimental resultsdemonstrate the proposed algorithm is superior to edge detection algorithm and it is also hasgood robustness.