Research On The Change Detection Method Of Specific Region In Multi Temporal Optical Remote Sensing Image

Optical remote sensing image change detection is an important means to obtain real-time situation information in key areas.It has a wide application value in many fields such as natural disaster early warning,urban development monitoring and battlefield damage assessment.With the rapid development of optical remote sensing satellite technology,there are more and more high-resolution optical remote sensing satellites in orbit in China,which can obtain multitemporal remote sensing images with higher spatial resolution and temporal resolution in the same area.The change detection ability can be improved by using these multitemporal temporal series data with high spatial resolution.Most of the existing methods are aimed at medium and low resolution remote sensing images,and image change information is described by pixel as the analysis unit.However,ground object targets in high resolution remote sensing images are often composed of a large number of pixels,so the change detection method based on pixel is difficult to adapt to the overall information description of ground object change.At the same time,the multi-temporal data with shorter revisit period contains more abundant information of temporal change of ground object scenes,which brings new technical difficulties to change detection technology.Aiming at the above problems,this paper studies the typical target change detection method in a specific area of multitemporal optical remote sensing image,and carries out experimental verification.The main research work is as follows:(1)Analysis of spatial-temporal evolution characteristics of multi-temporal remote sensing images.Based on the mechanism of space-based optical remote sensing imaging and the characteristics of multi-temporal optical remote sensing data,the physical factors affecting the spatial-temporal evolution characteristics of multi-temporal optical remote sensing images are analyzed,which mainly include the changes of ground objects themselves,the changes of illumination intensity,the changes of solar altitude Angle,the changes of satellite observation Angle and attitude Angle.On this basis,the temporal and spatial evolution characteristics and their differences between the background and typical objects in a specific region of multi-temporal remote sensing images are analyzed,which provides a theoretical basis for the research of remote sensing image change detection methods.(2)Research on image preprocessing method for change detection.Combined with the spatiotemporal evolution characteristics of multi-temporal remote sensing images,in order to eliminate the "pseudo-change" of ground objects caused by the coupling effect of the changes of ground objects themselves and the changes of imaging conditions,the image preprocessing methods were studied for the geometric distortion and radiation distortion of multi-temporal images caused by the changes of imaging conditions.In order to solve the problems of redundancy of feature points and mismatching of feature points in the image changing region existing in traditional registration methods,a feature point screening strategy based on Delaunay triangulation was introduced.In view of the limitation of the relative radiometric correction method based on pseudo-invariant features requiring manual participation,an adaptive estimation strategy is introduced to make full use of the time-series variation characteristics of the background in a specific region.(3)Research on the change detection method of specific region based on background modeling.On the basis of image preprocessing,a change detection method of specific region based on background modeling is proposed by taking into account the difference between the background and typical objects in specific region and the characteristics of the time series change of the background.In order to improve the accuracy and efficiency of background modeling,an adaptive learning rate and parameter updating strategy were introduced to solve the problems of low learning efficiency,modal residue and shadowing in traditional mixed Gaussian background modeling.In order to solve the problems of fracture in the extracted change area,morphological processing method was introduced into the extraction of change information to improve the integrity of change information on the object scale.Finally,in order to overcome the false alarm problem caused by the change of ground object shadow,a change information discrimination method based on multi-dimensional feature fusion is introduced to improve the accuracy of change detection.(4)Experimental verification and analysis.Using Orb View-3,IKONOS and other high-resolution optical remote sensing satellite measured data and multi-temporal optical remote sensing image simulation data as input,verify the correctness and effectiveness of the multi-temporal optical remote sensing image change detection method in a specific area,and combined with the change detection performance evaluation method,The performance of the proposed method for change detection in a specific area and the accuracy of change detection under shadow interference are analyzed.The experimental results show that the proposed method can achieve high precision change detection in airport and port areas of multi-temporal optical remote sensing images,and to some extent,it can overcome the interference of change detection accuracy caused by the change of ground object shadow.