熵最优与改进SCA的图像分割及其图像识别应用
Image segmentation based on entropy optimization and improved sine cosine algorithm and image identification application
孙博玲 1孙博文2
作者信息
- 1. 哈尔滨学院土木建筑工程学院,黑龙江哈尔滨 150086
- 2. 哈尔滨理工大学计算机科学与技术学院,黑龙江哈尔滨 150080
- 折叠
摘要
针对传统图像分割效率低、精度差的不足,提出一种混合变异正余弦算法的多阈值图像分割方法.为提高SCA算法的寻优性能,设计拉丁超立方种群初始化改进种群多样性;以非线性转换因子动态调节算法搜索能力;融入惯性权重机制提升算法全局寻优;结合高斯和拉普拉斯分布混合变异对个体扰动,使算法跳离局部最优.将Cross熵作为适应度函数,利用HMSCA求解分割阈值.实验结果表明,该算法可以提升图像分割精度和效率.将其应用于火灾图像识别,能够实现火焰源与背景分离,得到更好的分割效果.
Abstract
Traditional image segmentation algorithms have low efficiency and poor segmentation accuracy.Aiming at this prob-lem,a multilevel thresholds image segmentation algorithm based on hybrid mutation sine cosine algorithm HMSCA was pro-posed.To improve the optimization performance of SCA,a Latin hypercube population initialization method was designed to rea-lize the diversity of the initial population.A nonlinear conversion factor regulating mechanism was presented to balance the search ability.An inertia weight mechanism was proposed to promote global optimization ability.A hybrid mutation strategy in combination with Gaussian and Laplace distribution was designed to avoid the local optimum.Cross entropy was used as fitness function,and HMSCA algorithm was used to search the optimal image segmentation multilevel thresholds.Experimental results confirm that the proposed algorithm can enhance the accuracy of image segmentation and segmentation efficiency.The algorithm was applied to fire image segmentation.Results verify that the algorithm can realize the segmentation between fire source and background and obtain a better segmentation outcome.
关键词
图像分割/正余弦算法/拉丁超立方/混合变异/多阈值/图像熵/火灾图像Key words
image segmentation/sine cosine algorithm/Latin hypercube/hybrid mutation/multiple thresholds/image entropy/fire images引用本文复制引用
基金项目
黑龙江省重点教研基金(SJGZ20170056)
国家自然科学基金(61702140)
出版年
2024
NETL
NSTL
维普
万方数据