Research On Object-oriented Remote Sensing Image Change Detection Method

The change detection of remote sensing image is a technology for determining the change of surface coverage in a region by algorithmic recognition of remote sensing images at different times in the same region.The change detection method of remote sensing images based on object-oriented analysis has been used by more and more scholars in the field of change detection of high-resolution remote sensing image owing to it can make full use of texture information of high-resolution remote sensing images because of its object-based nature.Therefore,this paper conducts an in-depth study on the change detection methods of remote sensing image based on object-oriented analysis.The research content of this paper mainly includes the following three parts:(1)In the change detection method based on object-oriented,the image segmentation scale is uncertain,and the image quality obtained by segmentation is not high.In order to solve the above problems,an improved optimal segmentation scale solving model has been proposed.Firstly,the model establishes the segmentation scale quality evaluation model to obtains a series of scale quality evaluation indexes,and then gets the “segment scale-segment quality” function model by the way of polynomial fitting method.Finally,the peak value of the model is solved to obtain the optimal segmentation scale.The improved optimal segmentation scale solving model not only introduces spectral features into the evaluation,but also utilizes the texture features of the image by using the gray level co-occurrence matrix,which can fully take advantage of the rich texture information of high-resolution remote sensing images.(2)In the change detection method based on object-oriented,the quality of image objects is uneven.For solving the problem,a Relief-PCA feature selection method based on object-oriented has been proposed.At the optimal segmentation scale,fisrtly,the superimposed two-phase remote sensing images are segmented to obtain image objects,and the spectral features and texture features of the image objects in each band are extracted.Then,the features of the image objects are optimal dimensionality reduction by Relief-PCA method,then the CVA change intensity of the features after dimension reduction is calculated,and the change detection results are obtained by fuzzy C-means clustering algorithm.This method not only filters out high-quality image features through the Relief algorithm,but also eliminates the feature redundancy between the selected features by using the PCA algorithm,and can utilize various features of the image object more effectively.(3)In order to prove the superiority of the proposed method,four existing methods are introduced to compare with the proposed method.Compared to the traditional models,the experimental results show that the improved optimal segmentation scale model for introducing texture features is much better,and using Relief-PCA algorithm to select high-quality features can improve the accuracy of change detection results effectively.