Research On Change Detection Method Of High Resolution Remote Sensing Image

The high-resolution remote sensing images are important geospatial data for the country’s sustainable development,providing new ideas for solving social problems in land,resources,and environment.Change detection refers to the extraction of change information on features in different time phases and in the same geographic area.With the continuous launch of high-resolution remote sensing satellites,the rapid development of aviation technology,and the maturity of sensor technology,the image data has increased exponentially,and the acquired high-resolution remote sensing images have more abundant surface information.Change detection is of great significance to my country’s research in the fields of urbanization,disaster assessment,and land change.At the same time,the accompanying phenomenon of ‘same object with different spectrum,foreign object with same spectrum’ is also becoming common.The change detection method for a single feature often cannot meet the needs of high-resolution image change detection,and the traditional pixel-based change detection algorithm is easy to be disturbed by noise and produce ‘salt and pepper phenomenon’.This paper combines image spectral and texture features to detect changes in high-resolution remote sensing images.The main contents are as follows:(1)This paper has developed a high-resolution remote sensing image management platform,which realized the functions of fast display of high-resolution images,image segmentation,image preprocessing and fast retrieval of related images,the management platform has a friendly user interface;(2)The pre-processing of change detection mainly includes: firstly,use the image retrieval strategy based on the vocabulary tree to realize the fast retrieval of the multitemporal images of the image blocks after the splitting;then the multi-temporal images are registered;finally the dark channel-based defogging operation is performed on the image to make the image features more obvious and improve the accuracy of change detection.(3)Use object-oriented multi-feature fusion change detection algorithm to detect changes in images.The algorithm first performs difference change detection based on a single feature at the pixel level,and uses the EMD value to measure the degree of separation between change and non-change;then calculate the feature weight according to the EMD value,and fusion of multiple features at the object level in adaptive weighting method;finally,calculate the object feature difference,and divide the image into changed and unchanged categories using the Otsu algorithm.Finally,compare the change detection algorithm of this article and the traditional algorithm and test it.In the test based on the Tiszadob data set,the overall accuracy of the algorithm in this article is improved by 3.2% compared with the SP-Deep Labv3 algorithm,and the Kappa coefficient is improved by 0.04;in the test based on the Arcadia Lake data set,The overall accuracy of the algorithm in this paper is 9.9% higher than that of the SPDeep Labv3 algorithm,and the Kappa coefficient is increased by 0.31,which proves that the detection accuracy of the algorithm in this paper is higher than that of the comparison algorithm.This paper also compares and tests typical areas in detail: mountains,coasts,cities,and cultivated land,which proves the applicability of the algorithm in this paper to different types of ground features.Finally,the algorithm was tested through multi-temporal highresolution satellite images before and after the Tianjin explosion,which can effectively detect the changes of ground objects before and after the explosion,which proves the practicability of the algorithm in this paper.