Research On Automatic Extraction Algorithm Of Enclosure Culture Area In Yangcheng Lake Based On ZY-3 Image

With the rapid development of China fishery,enclosure culture of shallow lakes expands very quickly in Yangcheng lake,which increases economic benefits greatly.However,the high density of enclosure culture area in shallow lakes has also caused serious damage to lake's environment.In this context,it is key work for lake management to plan and control the lakes' enclosure culture area reasonably.According to information of temporal and spatial distribution,the lake management can make measures on improvement of water's quality and optimization of ecosystem rationally.Owing to the remote sensing technology having the advantages of real-timing,large scale and objection,it becomes the mainstream method of obtaining spatial information of enclosure culture area gradually.At present,The method of identification of enclosure culture area in shallow lakes is mainly manual visual interpreting remote sensing image of shallow lakes.This method not only costs time,but the benevolent see benevolence and the wise see wisdom.Although there are some researchers studied the feasibility of machine interpretation of enclosure culture area to further improve the interpretation efficiency,but most of them use only enclosure culture area's texture features,and need threshold characteristics to identify the enclosure culture area,what's more,the method can not be integrated.In this paper,an automatic identification method of enclosure culture area in shallow lakes based on remote sensing image from ZY-3 satellite is designed.In order to find the enclosure area and identify it,the pattern recognition theory of image is applied in recognition process of the shallow lakes' enclosure culture area,including image denoising,signal enhancing,image segmentation,image feature extraction and image classification.The main contents are followed:(1)Enhance and Segment remote sensing image based on image processing.First,analyze the necessity of enclosure area's boundary enhancement,and enhance boundary information by using gradient transformation and gray transformation.The results show that this method can effectively detect the mutational point and line in image which is equivalent to using high-pass filter;In the next place,describe filtering principle and steps of the adaptive filtering algorithm,and use the adaptive filtering method to eliminate gaussian noise and impulse noise in images;finally,study the basic principle of threshold segmentation,using wavelet transformation to find the accurate segmentation's threshold,and the result shows that when the threshold is 130 and 180,the algorithm can accomplish the task of segmentation perfectly.(2)Extract feature of in shallow Yangcheng Lake based on Fourier transformation.Firstly,through the analysing spectral information of enclosure culture area,make the most separable band of the culture area and water being near infrared clear;Then,Transform the segmented image using two-dimensional Fourier transform based on the near infrared band of ZY-3 image.Choose the frequency spectrum centered on the DC signal as characteristics of enclosure culture area;finally decompose it by using Singular value decomposition to reduce the dimension of the original feature vector.Take singular value of the frequency spectrum as the image characteristics of the segmented image.(3)Identify enclosure culture area based on nearest neighbor classification.First of all,through analysing advantages and disadvantages of different classification methods,regard the nearest neighbor classification as region identification method.Secondly,study the basic principle and steps of the nearest neighbor classification and apply the nearest neighbor classification to recognizing the Yangcheng Lake's enclosure culture area.finally,verify the accuracy by the confusion matrix comparing the manual visual interpretation,whose accuracy is 86%.