Research On A Frequency Domain Change Detection Method For Target-level High-resolution Remote Sensing Images

As the spatial resolution remote sensing image is increasing day by day,its application is increasingly wide,many industry for high resolution remote sensing image change detection put forward higher requirements,but in the process of practical application has encountered some difficulties.On the whole,there are two points: one is that the target of the change detection is not strong,and the other is the poor anti-interference ability of the change detection.The root cause of this study is to put the transformation information into the whole background,which not only leads to low detection efficiency,but also amplifies the influence of non-changing information on the change detection.In addition,the influence of "isomorphic spectrum" and "same spectrum foreign body" makes it difficult to fully cover the characteristics of the change information.The results show that the accuracy of the low change detection,the high omission rate and the high error rate.In view of the above mentioned problems,this paper takes buildings as an example and make the research work of the following aspects:1.Change detection pretreatment.Based on the law of image in frequency domain and airspace,the periodicity characteristics and directional characteristics of the ground objects are analyzed,and the image is segmented with the premise of ensuring the integrity of the ground.Further,the change degree of the image block is measured with the change texture rule degree,and the scope of the change information is determined.2.The target level change information detection based on frequency domain is carried out.Paper embarks from the Fourier change,discusses the distribution features of the frequency domain energy,on the one hand,through the section method,the first access to specific filter parameters,then build the 2-D Gabor filter bank complete low frequency information detection;On the other hand,using tensor wavelet analysis and harmonic window to obtain high frequency components in different directions.Finally,the information of the two aspects is superposed and integrated,and the target information is extracted,and the change information detection in the frequency domain is realized.Throughout the research process,the main contribution of this paper is:(1)It improves the pertinence of change detection.By means of video frequency domain energy to explore the image partition method,then the child in the image block feature change,will change to move information from the whole background to the study of relatively controllable environment,reduce interference of the change information from the source to detect the quality of ascension.(2)It strengthens the validity of change detection.Spectrum is verified by experiment and the advantages and disadvantages of low frequency band,high frequency band by combining the two complement each other,not only an accurate calibration of the features of the target range,more detail information to make a complete description of target feature.The test results are accurate and readable.