Data Clustering And Image Segmentation Based On Quantum-inspired Evolutionary Computation

In the 90th year of the 20th century, quantum computing aroused widespread concern. Quantum computing is a research area which includes concepts like quantum mechanical computers and quantum algorithms. Based on the properties, such as superposition, entanglement and parallel features, a new quantum-inspired evolutionary algorithm (QEA) was proposed by researchers, which integrated the advantages of evolutionary computation and quantum computation. QEA is based on the concepts and principles of quantum computing, such as the quantum bit and the superposition of states. It can explore the search space with a smaller number of individuals and exploit the search space for a global solution within a short span of time. In view of these characteristics of quantum-inspired evolutionary algorithm, we focused on data clustering and image segmentation problems. The simulation experiments showed that our algorithms achieved good results. The content of this paper includes:◆A quantum-inspired evolutionary algorithm based data clustering method is proposed. In traditional quantum-inspired evolutionary algorithm (QEA), quantum rotate gate is used to update the quantum population. However, the choice of rotate angle is also discrete, not continuous, which makes the search of the problem easy to fall into local optimum. Therefore, a modified quantum rotate gate is proposed in this paper. The new gate uses adaptive method of calculation of rotation, which makes the population have a relatively good global search capability. At the same time, the probability amplitude is modified after rotation to enable population to jump out of local optimum. For data clustering problem, the method proposed makes better accuracy rate compared to the QEA before. Then for data sets of symmetrical distribution property, we use a symmetrical distance measure and achieved good results.◆An unsupervised image segmentation based on quantum-inspired evolutionary gaussian mixture model (QEAGMM) is proposed. In the image segmentation algorithms based on gaussian mixture model (GMM), the standard method for training model parameters is expectation maximization algorithm (EM). However, the EM algorithm is apt to fall into a local optimum and the result is sensitive to initialization. So the EM algorithm is embed into QEA and proposed a quantum-inspired evolutionary algorithm based EM algorithm (QEA-EM) to train GMM. Then we use the new algorithm for image segmentation with a lot of other preprocess procedures. The experimental results show that compared to gaussian mixture model clustering algorithm (GMMC), the proposed method is successfully applied to texture mosaic images and SAR images, and achieves overall improvement in performance.