灰狼优化算法的改进及其在图像分割中的应用

发布时间：2018-05-16 12:49

本文选题：智能优化算法 + 灰狼优化算法　；参考：《河南师范大学》2017年硕士论文

【摘要】：灰狼优化(Grey Wolf Optimization,GWO)算法是一种新颖的元启发式智能优化算法,其模拟了大自然中灰狼种族特有的等级制度和集体狩猎行为。GWO算法由于结构简单、参数少,收敛速度快等优点在实际工程优化问题中得到广泛应用,但由于该算法提出时间较晚,其理论基础和算法应用方面的研究都不完善,算法本身也存在着诸多不足,如面对复杂优化问题时存在求解精度低、易早熟收敛等缺陷。为有效提高GWO算法的性能,拓展算法应用领域,本文通过对其算法理论和进化方式进行研究和分析,提出了两种改进方案,并将改进算法简单应用于多阈值图像分割问题,主要研究内容如下:(1)详细描述了GWO算法的思想来源,算法原理以及实施步骤,分析讨论了GWO算法的优缺点,并归纳了目前国内外对于GWO算法的各种改进思路,同时对GWO算法的应用领域进行了总结。(2)基于等级制度对于狼群狩猎影响的深入分析,提出了一种强化狼群等级制度的灰狼优化算法。该算法中的灰狼个体具有两种狩猎模式:一种是跟随狩猎模式,一种是自主探索模式。这两种狩猎模式既能体现高等级灰狼对低等级灰狼的引领作用,又能在充分挖掘种群位置信息的基础上发挥个体的自主能动性,提高种群的多样性避免算法陷入局部极值。仿真结果表明:该算法具有更强的全局勘探能力和更高的寻优精度。(3)针对灰狼优化算法和差分进化算法各自在应用上的优势和不足,提出了一种灰狼优化和差分进化的混合算法,实现了算法之间的优势互补,获得了一种全局搜索能力和局部搜索能力兼顾的高效混合优化算法,并将混合算法用于解决复杂高维函数优化问题,实验分析表明,该混合算法具有更好的收敛速度和优化性能,更适用于求解各种函数优化问题。(4)基于对上述混合算法特性的分析,将其应用于解决最大熵多阈值图像分割法中存在的阈值选择不准确、分割速度慢等问题,提出了一种新型多阈值图像分割算法。实验结果表明,该方法能够快速、准确的找到图像分割的最优阈值组合,进行有效分割。
[Abstract]:Grey Wolf Optimization (GWO) algorithm is a novel meta-heuristic intelligent optimization algorithm. It simulates the class system and collective hunting behavior of the gray wolf race in nature because of its simple structure and few parameters. The advantages of fast convergence rate have been widely used in practical engineering optimization problems. However, due to the late development of the algorithm, the theoretical basis and the application of the algorithm are not perfect, and the algorithm itself has many shortcomings. For example, in the face of complex optimization problems, there are some defects such as low accuracy and premature convergence. In order to effectively improve the performance of GWO algorithm and expand its application field, through the research and analysis of its algorithm theory and evolution mode, this paper puts forward two improved schemes, and applies the improved algorithm to multi-threshold image segmentation problem. The main research contents are as follows: (1) this paper describes in detail the origin, principle and implementation steps of the GWO algorithm, analyzes and discusses the advantages and disadvantages of the GWO algorithm, and summarizes various improvements to the GWO algorithm at home and abroad. At the same time, the application field of GWO algorithm is summarized. (2) based on the deep analysis of the effect of rank system on the hunting of wolves, an optimization algorithm is proposed to strengthen the hierarchical system of wolves. The gray wolf individuals in this algorithm have two hunting modes: one is following hunting mode and the other is self-exploring mode. These two hunting modes can not only reflect the leading role of the high-grade gray wolf to the low-grade gray wolf, but also give full play to the autonomous initiative of the individual on the basis of fully excavating the information of the population location, and improve the diversity of the population to avoid the algorithm falling into the local extremum. The simulation results show that the algorithm has stronger global exploration ability and higher searching accuracy.) aiming at the advantages and disadvantages of the grey wolf optimization algorithm and differential evolution algorithm in application, A hybrid algorithm of gray wolf optimization and differential evolution is proposed, which realizes the complementary advantages between the algorithms, and obtains an efficient hybrid optimization algorithm with both global and local search capabilities. The hybrid algorithm is used to solve the complex high-dimensional function optimization problem. The experimental results show that the hybrid algorithm has better convergence speed and optimization performance. It is more suitable for solving various function optimization problems. Based on the analysis of the characteristics of the hybrid algorithm mentioned above, it is applied to solve the problems of inaccurate threshold selection and slow segmentation speed in the maximum entropy multi-threshold image segmentation method. A new multi-threshold image segmentation algorithm is proposed. The experimental results show that this method can find the optimal threshold combination of image segmentation quickly and accurately, and can effectively segment the image.
【学位授予单位】：河南师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;TP391.41

【参考文献】