基于改进遗传算法的多目标截面投影图像特征分割方法  

Feature Segmentation Method of Multi-target Cross-section Projection Images Based on Improved Genetic Algorithm

作　　者：莫兴福 李沙 MO Xingfu;LI Sha(School of Information,Guangdong Nanfang Institute of Technology,Jiangmen 529000,China;Jiangmen Key Laboratory of Artificial Intelligence,Jiangmen 529000,China)

机构地区：[1]广东南方职业学院,信息学院,广东江门529000 [2]江门市人工智能重点实验室,广东江门529000

出　　处：《微型电脑应用》2025年第2期24-28,共5页Microcomputer Applications

基　　金：2020年度广东省普通高校青年创新人才项目(自然科学类)(2020KQNCX254);2020年度江门市基础与理论科学研究类科技计划项目(2020JC03030);广东省大学科技园与智能制造产教融合创新平台项目(2022CJPT029)。

摘　　要：在多目标截面投影图像处理过程中,受到特征分割阈值的影响,分割后区域一致性较低,因此,提出基于改进遗传算法的多目标截面投影图像特征分割方法。利用多目标截面投影图像构建拓扑结构模型,对图像特征像素进行分块处理。基于极大类间方差准则,设计以分割阈值为核心的图像特征分割模式。依托于改进遗传算法设计阈值优化算法,获取更加合理的分割阈值。在人工记忆模型的作用下,求解出最终图像特征分割方案。实验结果表明,所提方法的图像特征分割区域一致性较比较方法分别提升了8.37个百分点、11.32个百分点,具有较高的区域一致性。In the multi-target cross-section projection image processing,the consistency of segmented region is low due to the influence of feature segmentation threshold.Therefore,a multi-objective cross-section projection image feature segmentation method based on improved genetic algorithm is proposed.A topological structure model is constructed by the multi-target cross-sectional projection image,and the image feature pixels are divided into blocks.Based on the criterion of maximum inter-class variance,an image feature segmentation mode with segmentation threshold as the core is designed.Based on the improved genetic algorithm,the threshold optimization algorithm is designed to obtain a more reasonable segmentation threshold.Under the action of artificial memory model,the final image feature segmentation scheme is solved.The experimental results show that the regional consistency of image feature segmentation of proposed method is improved by 8.37 percentage points and 11.32 percentage points,respectively,compared with the comparison methods,which has a higher regional consistency.

关 键 词：改进遗传算法 截面投影 图像分割 人工记忆模型 滤波算法 分割阈值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]