Li Shujing, Li Zhangfei, Li Qinghe, Zhang Mingyu, Li Linguo
School of Computer and Information Engineering, Fuyang Normal University, Fuyang, China.
School of Computer, Nanjing University of Posts and Telecommunications, Nanjing, China.
Front Plant Sci. 2023 Jan 26;14:1122788. doi: 10.3389/fpls.2023.1122788. eCollection 2023.
With the development and wider application of meta-heuristic optimization algorithms, researchers increasingly apply them to threshold optimization of multi-level image segmentation. This paper explores the performance and effects of Capuchin Search Algorithm (CAPSA) in threshold optimization. To solve problems of uneven distribution in the initial population of Capuchin Search Algorithm, low levels of global search performance and premature falling into local optima, this paper proposes an improved Capuchin Search Algorithm (ICAPSA) through a multi-strategy approach. ICAPSA uses chaotic opposite-based learning strategy to initialize the positions of individual capuchins, and improve the quality of the initial population. In the iterative position updating process, Levy Flight disturbance strategy is introduced to balance the global optimization and local exploitation of the algorithm. Finally, taking Kapur as the objective function, this paper applies ICAPSA to multi-level thresholding in the plant images, and compares its segmentation effects with the original CAPSA, the Fuzzy Artificial Bee Colony algorithm (FABC), the Differential Coyote Optimization Algorithm (DCOA), the Modified Whale Optimization Algorithm (MWOA) and Improved Satin Bowerbird Optimization Algorithm (ISBO). Through comparison, it is found that ICAPSA demonstrates superior segmentation effect, both in the visual effects of image segmentation and in data comparison.
随着元启发式优化算法的发展和更广泛应用，研究人员越来越多地将其应用于多级图像分割的阈值优化。本文探讨了卷尾猴搜索算法（CAPSA）在阈值优化中的性能和效果。为了解决卷尾猴搜索算法初始种群分布不均、全局搜索性能较低以及过早陷入局部最优等问题，本文通过多策略方法提出了一种改进的卷尾猴搜索算法（ICAPSA）。ICAPSA采用基于混沌反向学习的策略来初始化个体卷尾猴的位置，提高初始种群的质量。在迭代位置更新过程中，引入莱维飞行扰动策略来平衡算法的全局优化和局部开发能力。最后，以卡普尔熵作为目标函数，将ICAPSA应用于植物图像的多级阈值分割，并将其分割效果与原始的CAPSA、模糊人工蜂群算法（FABC）、差分土狼优化算法（DCOA）、改进的鲸鱼优化算法（MWOA）和改进的缎蓝亭鸟优化算法（ISBO）进行比较。通过比较发现，ICAPSA在图像分割的视觉效果和数据比较方面均表现出优越的分割效果。
