高分辨率全色遥感图像多级阈值分割  

Multi-level threshold segmentation of high-resolution panchromatic remote sensing imagery

作　　者：杨蕴 李玉[1] 赵泉华[1] YANG Yun;LI Yu;ZHAO Quan-hua(Institute for Remote Sensing Science and Application, School of Geomatics, Liaoning Technical University, Fuxin 123000, China)

机构地区：[1]辽宁工程技术大学测绘与地理科学学院遥感科学与应用研究所,辽宁阜新123000

出　　处：《光学精密工程》2020年第10期2370-2383,共14页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.41301479,No.41271435);辽宁省自然科学基金资助项目(No.2015020090)。

摘　　要：针对模糊熵多级阈值分割方法存在模糊特性不足、计算量大、自动性差等问题,提出一种基于区间二型模糊熵的高分辨率全色遥感图像多级阈值分割方法。首先,利用岭型模糊隶属度函数构造区间二型模糊集,由构造的模糊集和阈值个数,在多级图像分割场景中定义区间二型模糊熵。然后,利用量子比特将其模糊参数集编码为量子染色体,设置若干量子染色体构成初始种群,并以定义的区间二型模糊熵作为适应度评价函数,对种群中的个体进行适应度评价,保留和记录最优个体。在提出的进化策略中,利用量子旋转门的动态旋转角机制使种群以更好的适应性和效率自动确定模糊参数的最优组合,据此,以最大模糊性原则得到多级阈值,实现图像最优多级阈值分割。在实验中选取基于最大熵和模糊熵的多级阈值分割方法作为对比算法,对具有不同地物的高分辨率全色遥感图像进行了分割实验。实验平均评价结果表明本文方法能在减少计算时间的同时获得更好的分割结果,面积加权方差降低了39.7%,Jeffries-Matusita距离降低了14.7%,运行时间为6.403 s。可满足高分辨全色遥感图像分割结果对空间连续且光谱均匀的要求且具有高实时性。To address the problems of fuzzy entropy-based multilevel threshold segmentation methods,such as insufficient fuzzy characteristics,high computational complexity,and poor automaticity,a multilevel threshold segmentation method for high-resolution panchromatic remote sensing imagery is proposed based on interval type-2 fuzzy entropy.First,a ridge-type fuzzy membership function is applied to construct an interval type-2 fuzzy set,and interval type-2 fuzzy entropy is defined in the multilevel image segmentation scene based on the constructed fuzzy set and the number of thresholds.Then,qubits encode a fuzzy parameter set as quantum chromosomes,and several quantum chromosomes are set to form the initial population.In addition,the defined interval type-2 fuzzy entropy is adopted as the fitness evaluation function to evaluate the fitness of individuals in the population,retaining and recording the best individuals.In the proposed evolutionary strategy,the dynamic rotation angle mechanism of quantum rotation gates is applied such that the population can automatically determine the optimal combination of fuzzy parameters with better adaptability and efficiency.Based on this,the multilevel threshold is obtained by the principle of maximum fuzziness,and the optimal multilevel threshold segmentation of the image is realized.In an experiment,a multilevel threshold segmentation method based on maximum entropy and fuzzy entropy was employed as the comparison algorithm to segment high-resolution panchromatic remote sensing images with different ground objects.The averages of the experimental evaluation results show that the proposed method can obtain better segmentation results while reducing the computation time.The area weighted variance is reduced by 39.7%,the Jeffries-Matusita distance is reduced by 14.7%,and the running time is 6.403 s.The method can meet the requirements of high-resolution panchromatic remote sensing image segmentation for spatial continuity and spectral uniformity,resulting in high real-time performance.

关 键 词：高分辨率 全色遥感图像 图像分割 区间二型模糊熵 优化 多级阈值 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置] TP79[自动化与计算机技术—控制科学与工程]