基于蜜獾算法的破片序列图像多阈值分割  

Multi-Threshold Segmentation of Fragment Sequence Images Based on Honey Badger Algorithm

作　　者：牛天利 于丽霞[1] 刘吉[1,2] 武锦辉[2] 牛雅昕 NIU Tian-li;YU Li-xia;LIU Ji;WU Jin-hui;NIU Ya-xin(School of Information and Communication Engineering,North University of China,Taiyuan 030051,China;Key Laboratory of Electronic Testing Technology,North University of China,Taiyuan 030051,China)

机构地区：[1]中北大学信息与通信工程学院,山西太原030051 [2]中北大学电子测试技术重点实验室,山西太原030051

出　　处：《测控技术》2023年第1期92-98,共7页Measurement & Control Technology

基　　金：山西省基础研究计划(自然探索类)面上项目(202103021224188);山西省回国留学人员科研资助项目(HGKY2019068)。

摘　　要：针对静爆试验中拍摄的图像中破片目标小、背景复杂等情况,基于经典阈值分割法在破片图像分割的应用中存在不能将目标所在像素准确分离的问题,提出一种基于蜜獾算法(HBA)的多阈值图像分割方法,该方法引入HBA求解Tsallis相对熵的最小值作为目标函数值来计算最佳阈值,在分析经典阈值分割方法处理破片图像的不足后,选择合适的阈值数,将HBA与遗传算法(GA)、蝗虫优化算法(GOA)、麻雀搜索算法(SSA)三种优化算法进行性能对比,利用分离出的目标绘制破片轨迹图并确定有效破片。分析结果表明,阈值数为2时分割效果满足需求,HBA运行时间1.32 s,进行100次重复实验后其结果的标准偏差约为0,分割出的目标中有效破片达83.8%,说明该算法的实时性和稳定性强,分割效果可满足对破片群运动参数测试的需求。Aiming at the small target and complex background in the image of static explosion test and the classical threshold-based segmentation method has the problem of failing to accurately separate the target pixels in the application of fragment image segmentation, a multi-threshold image segmentation method based on the honey badger algorithm(HBA) is proposed, which introduces the HBA to solve the minimum value of Tsallis relative entropy as the objective function value to calculate the optimal threshold.After analyzing the shortcomings of classical threshold segmentation methods for processing fragment images, the appropriate number of thresholds is selected, the performance of HBA is compared with genetic algorithm(GA),grasshopper optimization algorithm(GOA) and sparrow search algorithm(SSA),the fragment trajectories are mapped and the effective fragment is determined by using the separated targets.The analysis results show that the segmentation effect meets the demand when the threshold number is 2,the HBA running time is 1.32 s, the standard deviation is about 0 after conducting 100 repetitions of the experiment, and the effective fragment reaches 83.8% in the segmented target, which indicates that the algorithm is real-time and stable, and the segmentation effect can meet the demand for testing the motion parameters of the fragment group.

关 键 词：蜜獾算法 多阈值图像分割 Tsallis相对熵 破片序列图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]