Research On Change Detection Technology Of Multi-temporal Remote Sensing Image

The continuous city urbanization and natural disasters have not only changed topography,climate,hydrology,ecology and other natural geographical environment,but also caused great changes to the natural,human geography environment,With the passage of time,these changes and influence not only have not been weakened,and continues to accelerate.In order to assess the impact of these changes on natural environment and social economy,we need a dynamic monitoring of these changes effectively.But artificial regional survey,statistical methods and other traditional methods waste a lot of manpower and material resources,and they have a long monitoring period and a small coverage area.These make that the monitoring results is difficult to meet the needs of the current actual monitoring.Due to its high temporal resolution,large real-time monitoring range,convenient acquisition and other advantages,remote sensing images have been widely used in dynamic monitoring of the natural,human geography environment change.Among them,multitemporal remote sensing image change detection technology is the key technology to realize the dynamic monitoring.Remote sensing images of different time phase are used in this technology for analyzing and identifying the characteristics and process of the natural and human environment change.According to the different test objects,the requirements of spatial resolution of remote sensing image are also different.Middle and low spatial resolution remote sensing image can meet the detection of large scale objects(city,river),but only high spatial resolution remote sensing image can meet the detection of artificial target,military targets and other features.Currently middle and high spatial resolution remote sensing images constitute the main data source of the natural and humanistic environment monitoring changes detection.In this paper,multi-phase and high spatial resolution remote sensing images were as the research objects,through a combination of intelligent optimization algorithm,image segmentation and object based image analysis method to explore the new method of detecting new multi-temporal remote sensing image change.The concrete research work were as follows:(1)Aiming at the differences of middle and high spatial resolution remote sensing images,method of direct objects comparison and pixel level change detection respectively were used for change region detection.And the research status,key technologies and existing problems of the two methods were analyzed and summarized;(2)A multi-temporal remote sensing images change detection method based on the minimum cross entropy and the simplified PCNN and a multitemporal remote sensing images change detection method based on firefly algorithm were proposed.The change threshold selection is the key and difficult points of pixel level change detection methods.In this paper,pulse coupled neural network(PCNN)and firefly algorithm were studied,and the two intelligent optimization algorithm were used to solve the problem of selecting the threshold of pixel level change detection.(3)Image segmentation is the key technology of change detection method based on direct comparison between the objects;one also is an important factor of restricting its development.In order to improve the accuracy of image segmentation,study on a multiresolution segmentation method of remote sensing image based on edge constrained.Spatial resolution improvement can be more precise identification of ground objects,but also because of its rich texture details,spectral variety to the object edge detection brings a lot of difficulties.In order to reduce the impact of ground texture information and internal details caused by edge detection,detect the feature contours reduce unreasonable,bringing in Markoff chain algorithm and the simplified PCNN model,proposed a edge detection method of high resolution remote sensing image based on Markoff chain algorithm and proposed a edge detection method of high resolution remote sensing image based on simplified PCNN model combination with adaptive Canny operator.(4)The spectral features,texture features and spatial features of remote sensing image were introduced and analyzed;the vector similarity and several kinds of commonly used vector similarity measure method were introduced;(5)On the basis of comprehensive analysis the ground features and vector similarity of remote sensing image,proposed a change detection method for multitemporal remote sensing images based on direct comparison between the objects and multi-feature combination with vector similarity.The main innovative points of this paper are as follows:(1)Proposed a change detection method for remote sensing images based on firefly algorithm.Study on the application of the firefly algorithm in change detection for multitemporal remote sensing images,proposed a change detection method for remote sensing images based on the firefly algorithm.Firstly,the difference image of multitemporal remote sensing images is produced by difference method;then,according to the difference in different types of remote sensing,dynamic initialization the parameter of firefly algorithm;calculated 2D-OTSU function value H(s,t)corresponding with each firefly,the objective function value as the brightness,the position of maximum brightness firefly obtained by sequence,output the maximum value of(s1,t1),the value(s1,t1)as the judging threshold to judged the pixels in difference image change or not;Finally,the change areas obtained by segmented the difference image.(2)Proposed a change detection method for multitemporal remote sensing images based on direct comparison between the objects and multi-feature combination with vector similarity.Study on change detection method based on direct comparison between the objects,proposed a change detection method for multitemporal remote sensing images based on direct comparison between the objects and multi-feature combination with vector similarity.Firstly,the edge information of remote sensing obtained by proposed edge detection method;then,the edge information as constraint condition,the image objects obtained by pixels merging in according with the "top-down" merging rules,so as to realized on the remote sensing image segmentation;In respectively obtained image objects of multi temporal remote sensing images,the image objects of multitemporal remote sensing images direct compared by logic operation model,in order to judged the objects change or not,chose the best ground features and vector similarity as judging factors.