基于莱维飞行麻雀搜索优化的Otsu多阈值图像分割  

Otsu multi threshold image segmentation based on Levy flight sparrow search optimization

作　　者：马卫[1] 朱娴 李微微 Ma Wei;Zhu Xian;Li Weiwei(School of Hotel Management,Nanjing Institute of Tourism and Hospitality,Nanjing,Jiangsu 211100,China;Department of Computer Science,Zijin College,Nanjing University of Science and Technology;College of Computer and Information,Hohai University)

机构地区：[1]南京旅游职业学院酒店管理学院大数据技术教研室,江苏南京211100 [2]南京理工大学紫金学院计算机学院 [3]河海大学计算机与信息学院

出　　处：《计算机时代》2023年第4期77-85,共9页Computer Era

基　　金：江苏省高校自然科学基金(No.17KJB520013);江苏省高校哲学社会科学研究项目(No.2020SJA0794,No.2021SJA0782);江苏省高校“青蓝工程”学术带头人项目;国家文化和旅游部文化艺术职业教育和旅游职业教育提质培优行动计划“双师型”师资培养扶持项目(No.2021TZPYSS);江苏省社科应用研究精品工程课题(No.22SYB-117);江苏省职业改革研究课题(No.ZYB601);科研创新团队资助项目(No.2021KYTD04)。

摘　　要：群智能算法优化多阈值图像分割策略,易陷入局部最优,搜索精度不高。将莱维飞行扰动策略融入麻雀种群觅食的搜索优化过程,以增加图像分割空间搜索的多样性,从而提高分割精度,避免搜索过程陷入局部最优;同时嵌入Sin混沌搜索机制,改进种群初始化策略,加强搜索性能。最终实现多阈值图像分割的应用性能提升,在保持算法全局寻优能力的基础上大幅度提高收敛速度和求解精度。对经典的图像多阈值分割的实验结果表明,本文方法相比于传统的智能优化策略,在寻优率和分割精度方面提升显著,收敛能力强。The swarm intelligence optimization multi threshold image segmentation strategy is easy to fall into local optimization and the search accuracy is not high.Levy flight disturbance strategy is integrated into the search optimization process of sparrow population foraging to increase the diversity of image segmentation spatial search,so as to improve the segmentation accuracy and avoid falling into local optimization in the search process.At the same time,the Sin chaotic search mechanism is embedded to improve the population initialization strategy and strengthen the search performance.Finally,the application performance of multi threshold image segmentation is improved,and the convergence speed and solution accuracy are greatly improved on the basis of maintaining the global optimization ability of the algorithm.The experimental results of classical multi threshold image segmentation show that compared with the traditional intelligent optimization strategy,this method significantly improves the optimization rate and segmentation accuracy,and has strong convergence ability.

关 键 词：多阈值图像分割 麻雀搜索算法 莱维飞行扰动 群智能优化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]