Study On Pixel Entanglement Theory For Imagery Classification

Remote sensing technology is a modern scientific observation technology which is based on the electromagnetic wave to study the distribution and space-time change of the earth objects.Remote sensing image classification is the foundation of remote sensing technology,which is widely used in thematic mapping of remote sensing image,land cover analysis,digital city and other fields.At the same time,the classification processing of remote sensing image is a complicated and difficult process of mass data processing workflow,which will be affected by many objective factors.For example,the selection of remote sensing images and complex surface information still face great challenges.Quantum mechanics,especially quantum entanglement theory,provides a new perspective for remote sensing image processing.Therefore,this paper fully refers to and summarizes some basic concepts and main basic principles of quantum entanglement theory,and proposes a self-organizing pixel entanglement neural network remote sensing image classification method.In this paper,based on the basic concept and principle of quantum entanglement,the pixel entanglement theory is studied around the actual remote sensing image classification technology:(1)Aiming at the classification problem of single-band remote sensing images,this paper proposes a single-band remote sensing image classification algorithm based on self-organizing pixel entanglement neural network.In the self-organizing pixel entanglement neural network(SOQENN),the pixels in the remote sensing image are first treated as quantum particles on the spatial array.Then,the quantum entanglement coefficient is proposed to obtain the quantum entanglement relation of quantum particles on the spatial array.Finally,a self-organizing neural network was established to simulate the quantum entanglement behavior between quantum particles,and the classification process of single-band remote sensing images was transformed into the self-organizing quantum entanglement process of quantum pixels in the state configuration space.(2)In this paper,the classification algorithm of self-organizing pixel entanglement single-band remote sensing image is extended to multi-band remote sensing image,and the state superposition principle of multi-spectral quantum pixel in feature space and the classification algorithm of self-organizing pixel entanglement neural network multi-spectral remote sensing image are proposed.Firstly,the quantum pixels in the single-band remote sensing image are superimposed into a multi-spectral quantum pixel,and each single-band quantum pixel is regarded as the feature vector of the multi-band quantum pixel,so that a multi-spectral quantum pixel contains the quantum pixel information of each band.(3)In order to verify the correctness of self-organizing pixel entanglement theory,SOQENN has been used for remote sensing image classification in four test areas.In this paper,the remote sensing image classification results of Sopen N are compared with unsupervised classification methods such as ISODATA,K-means,SOM and supervised classification methods such as SVM.The results of this paper show that the classification accuracy of the proposed method is good,and the relative accuracy of the four test areas are 97.56%,73.14%,96.45% and 91.85% respectively,which meet the classification requirements of remote sensing images on the whole.In addition,the classification accuracy of the proposed method is better than that of the traditional SOM in the other three test areas,except that the first test area is equal to the traditional SOM.Compared with ISODATA and K-means,the proposed method improves the classification accuracy by 21.21%,4.88%,11.06% and 11.31%,respectively.The results show that the self-organizing pixel entangled neural network proposed in this paper can effectively improve the classification accuracy of remote sensing images.