Research And Application Of Teaching-learning-based Optimization Algorithm

Swarm intelligence optimization algorithmhas received more and more attention from scholars. It has the advantage of simple operation, fast and better global convergence. The research on swarm intelligence optimization algorithm is more and more mature now. Teaching-Learning-Based optimization algorithm is a new swarm intelligent optimization algorithm proposed by R.V.Rao. The algorithm imitates the process of teaching and learning between teachers and learners in real life. Through the teaching by teacher and the learning among learners, it improves the grade of the class and enhances the ability of finding the global optimal solution.In the paper, we analyze the problems existing in the algorithm and the areas to be improved, such as the ability of global search and the speed of convergence. Moreover, the algorithm is easy to fall into local optimal solution for solving the problem, especially for large complex multimodal functions. We propose the improved algorithm based on K-means for the problems as mentioned above, and the improved algorithm is applied to the image segmentation.The detailed work of this paper is given as bellow:First, aiming at the shortcomings of teaching-learning-based optimization algorithm, the TLBO algorithm based on K-means(KTLBO) is proposed. By the K-means clustering, entire class is split into different groups of learners and individual teacher is assigned to individual groups of learners. The algorithm introduces more than one teacher for leaners to teach learners in accordance of their aptitude. Another modification is related to the teaching factor(TF) of the TLBO algorithm. Thus in the KTLBO algorithm the teaching factor varies automatically during search. Automatic tuning if TF unites both the requirements of optimization algorithm, i.e. fine and quick search and hence improve the performance of the algorithm. Furthermore, the mutation phase can maintain the diversity of population and avoid precocious. The algorithm can avoid trap in local optimum and can obtain the global optimal solution by the mutation. By comparing with the existing intelligent optimization algorithms and the TLBO, we can find that the algorithm proposed in this paper can obtain better results and can gain a better global optimal solution.Second, by combining KTLBO and the maximum entropy multi-threshold selection techniquethe, we propose the image segmentation based on KTLBO and get better maximum entropy and can achieve better image segmentation results. According to the application of cell image segmentation and counting in medical image, we design and implement the cell image segmentation and counting system based on KTLBO. This system mainly has the function of image input, image preprocessing, image grayscale, image segmentation, image marking and counting, and the result of image segmentation can be obtained by adjusting the parameters of the actual demand. The segmentation and counting results getting from the system, has important significance for 3D visualization, positioning and computer-aided diagnosis.