Research On Extracting Method Of Reef Feature Based On Ocean Remote Sensing Data

As an effective means of observing ocean data real-time, marine remote sensing technology has been widely used in the fields of intelligent navigation, marine stereo monitoring and ship precise positioning. Especially in the marine navigation field,because of the characteristics of wide coverage and high working timeliness of marine remote sensing,marine remote sensing technology can greatly improve the timeliness and accuracy of marine navigation when applied to get the large area and synchronous sea remote sensing image for the marine navigation and ship barrier. However,there is too much information in marine remote sensing images containing large numbers of kinds of objects,making the accurately identifying of the object especially the island features become the precondition of applying the marine remote sensing technology to marine navigation,marine barrier and marine data monitoring.This paper takes the multispectral remote sensing image as the object of study,doing clustering analysis the image data starting from the image processing. As multispectral remote sensing images have the features of large scale,multi band,rich content and multi-scale,the original feature extraction algorithm shows its shortage. Two improved clustering analysis algorithms are proposed for image feature extraction and the correctness of the algorithm is verified by experiments.Firstly,the necessary preprocessing analysis of remote sensing image is carried out.The preprocessing stage includes two parts: the radiation correction and the geometric correction.The reasons of correction and the methods of the two parts are included in the paper respectively.Secondly, the commonly used feature extraction algorithm is introduced about the spectral feature, edge feature and shape feature of remote sensing images, providing a direction for improving subsequent algorithm.Thirdly,fuzzy kernel clustering algorithm is deeply studied.It is improved aiming at the shortcomings of fuzzy kernel clustering algorithm in the using process. The data processing speed is improved by analyzing the principal component of the raw data. The accuracy of data processing is improved by using Markov clustering instead of Euclidean clustering.The clustering center is obtained by using Schrodinger equation to solve the problem of local optimal solution.And the experimental simulation is carried out on this basis,experimental results show the correctness of the improved algorithm.Last,the support vector clustering algorithm is deeply studied.lt is improved aiming at the shortcomings of the support vector clustering algorithm in the using process.In order to reduce the running cost of the algorithm,this paper proposes the concept of optimal data subset to ensure that the original data set loss is minimal.In order to improve the accuracy of clustering algorithm,this paper presents a new clustering method to avoid the occurrence of random error of calculating the adjacency matrix and connected components.And the experimental simulation is carried out on this basis,experimental results show that the improved algorithm is better than the support vector clustering algorithm,and the improved algorithm has better robustness.