Research On High Resolution Remote Sensing Image Registration Technology

Recently,with the development of remote sensing technology,high-resolution remote sensing satellites have substituted spacecraft carried with sensors,because spacecraft has the disadvantages of long shooting period,high price,low image resolution and difficult processing.Currently,high-resolution remote sensing has been widely used in agriculture,forestry,urban planning,military and other fields.Image registration technology is the basis of these applications.Therefore,high-resolution remote sensing image registration technology has attracted the attention of scholars.Compared with the traditional remote sensing images with medium and low resolution,the high-resolution remote sensing images can better display the details and texture features of the obj ects.However,at the same time,the small deformation of the local images will be enlarged,which brings new challenges to the registration of high-resolution remote sensing images.Firstly,the high-resolution images have spatial information.The feature matching of near-ground objects is more susceptible to the interference of texture feature similarity.Secondly,high-resolution remote sensing images have high spatial resolution and large width,which brings inconvenience to image storage,transfer and processing.Thirdly,high-resolution remote sensing satellites usually carry a lot of important information and are vulnerable to attacks in the process of image registration.Based on the analysis of the above challenges,this paper will conduct in-depth research,followed by the related achievements and research ideas.(1)In order to solve the similarity interference,this paper proposes a descriptor which combines local features with global context information,namely local and global of scale-invariant feature transform(LG-SIFT).Then,the experimental simulation and comparison with other classical registration algorithms show that the proposed algorithm can reduce the interference of texture similarity to remote sensing image registration and improve the accuracy of image registration.(2)In view of the large storage space and difficult processing of high-resolution remote sensing images,compressed sensing technology is proposed to compress and sample images,thus reducing the transmission bandwidth of images.At the same time,image registration process is completed through the cloud,using cloud computing resources to improve image processing speed and facilitate image storage and transfer.(3)In order to reduce the information of remote sensing image from attack,the pseudo-randomness of lower chaotic map and the sensitivity of initial value of chaotic sequence are used to process the image,and two chaotic matrices are used as keys to enhance the security of the method.Combining the above three challenges,this paper proposes a chaotic compressed sensing cloud remote sensing image registration scheme based on finite-state.Firstly,remote sensing images are compressed and encrypted by chaotic compression sensing method with finite state,which reduces the transmission bandwidth of image and encrypts image information.Then,we decrypt the processed images in the cloud and adopt the cloud distributed processing method,which not only improved the efficiency of image processing,but also saved the storage space of the image and facilitates the transfer of the image.Finally,remote sensing image registration method combining local features and relative shape context performed,and the registration effect is compared with the current relatively new algorithms.The experimental results show that the proposed scheme can reduce the interference of similar texture features and improve the registration accuracy and robustness of high-resolution remote sensing images.