Research On Method Of Building Information Extraction Based On High Resolution Remote Sening Image

With the rapid development of satellite remote sensing imaging technology,the resolution of the acquired ground image is getting higher and higher,and it can get all-weather,all-round,real-time data.Therefore,the use of high resolution remote sening image for feature information extraction,can overcome a cumbersome field survey,greatly improve work efficiency.Building is an important artificial object in remote sensing images,it's information extraction has important significance and practical application on the establishment of the digital city,geographic information update,military target reconnaissance and other areas.Based on the previous work,the thesis studies building information extraction method based on high resolution remote sensing image for building remote sensing image enhancement,building remote sensing image segmentation,building edge detection and building region recognition.The main work is as follows.Firstly,a method of building remote sensing image enhancement based on contourlet transform and L₁ norm for edge-preserving image smoothing is proposed.The contourlet transform is performed to decompose the building remote sensing image into several high-frequency components and a low-frequency component.Then the high-frequency components are transformed by a nonlinear transform,through which the edges and details are enhanced.The edge-preserving image smoothing based on L₁ norm is utilized to enhance the low-frequency component.Finally,the result image is reconstructed from the contourlet inverse transformation.Experimental results show that,compared with four enhancement methods put forward in recent years,The proposed method has obvious advantages in terms of subjective visual effects and objective evaluation indicators.Next,a thresholding method for remote sensing images of building based on two-dimensional Tsallis cross entropy using chaotic cuckoo search optimization is proposed.Using the characteristics of Tsallis cross entropy to comprehensively consider difference and correlation of pre-segment and post-segment image information,the formula of 2-D Tsallis cross entropy threshold selection based on the histogram is derived.The cuckoo search algorithm is utilized for precise optimization of thresholds based on the 2-D Tsallis cross entropy.In order to reduce run time further,logistic chaotic map is used in the cuckoo search algorithm so as to improve the convergence rate of cuckoo search algorithm.Lastly,realize the threshold segmentation of building remote sensing images with optimal threshold.The experimental results show that,compared with 2-D reciprocal cross entropy thresholding method,2-D Tsallis entropy thresholding method,2-D Tsallis gray entropy thresholding method based on chaotic particle swarm optimization and so on,the building object in the images segmented by the proposed method is more accurate,the details are more explicit,in addition,its running time is shorter.Then,a method of building edge detection for remote sensing images is discussed based on modulus maxima of complex Shearlet domain and Hough transform.Firstly,the building remote sensing image is multi-scale decomposed through complex Shearlet transform,resulting in a low-frequency component and a series of high-frequency components.Detecting the low-frequency component by Hough transform to get the low-frequency edge extraction result,the modulus maximum detection is performed for each sub-band of high-frequency components and get the high-frequency edge image with rich texture.Finally,the above two parts are combined by weighted summation,and perform pseudo edge suppression to finally detect the complete building edge.The experimental results demonstrate that in subjective visual effect,peak signal to noise ratio?PSNR?,the proposed method has a better performance compared with Canny method,Wavelet based modulus maxima method and Shearlet based modulus maxima method.Finally,selecting median robust extended local binary pattern?MRELBP?as texture feature operator and Franklin moment as shape characterization,a remote sensing image building area recognition method based on MRELBP feature,Franklin moment and optimized support vector machine?SVM?by cuckoo search is proposed.To calculate the texture feature vector of the image with MRELBP texture feature operator and use Franklin moment invariant obtain the shape feature vector,the texture feature vector and shape feature vector are combined into a comprehensive feature vector.Then train the SVM with training image samples,meanwhile,to use cuckoo search to optimize the kernel function parameter as well as the penalty factor.Lastly,input the recognizing image into the SVM to get the the result of building area recognition.Experimental results show that compared with the the classification method based on RGB and SVM,the classification method based on LBP and SVM and the classification method based on Zernike moment and SVM,the accuracy of the building area of the remote sensing image identified by this method is higher.