机器视觉技术的红外弱小目标状态感知方法  

Infrared weak and small target state perception method based on machine vision technology

作　　者：韩菊 HAN Ju(Department of Computer Science and Technology,Taiyuan University,Taiyuan 030012,China)

机构地区：[1]太原学院计算机科学与技术系,太原030012

出　　处：《激光杂志》2023年第5期236-241,共6页Laser Journal

基　　金：山西省教育科学“十四五”规划课题(No.GH-21403)。

摘　　要：为在复杂场景下精准感知多个红外弱小目标,提出基于机器视觉技术的红外弱小目标感知方法。依据搜索窗口与筛选条件,从红外图像中筛选出候选感兴趣区域,采用多目标分割算法,应用显著性特征从候选感兴趣区域中提取感兴趣区域,并提取感兴趣区域的加权灰度均值特征,将其作为Soft-FART网络的输入,通过Soft-FART网络模糊匹配知识库与感兴趣区域的加权灰度均值特征,选取大于警戒条件的目标作为感知到的红外弱小真实目标,并将感知结果输出。实验结果表明,该方法可在不同场景下精准感知红外单独弱小目标和多个红外弱小目标,且在高噪声环境下依然具备较好的红外弱小目标感知性能。In order to accurately perceive multiple infrared weak and small targets in complex scenes,an infrared weak target perception method based on machine vision technology is proposed.According to the search window and screening conditions,the candidate regions of interest are selected from the infrared image,the multi-target segmenta-tion algorithm is adopted,the salient feature is applied to extract the regions of interest from the candidate regions of interest,and the weighted gray mean feature of the regions of interest is extracted as the input of soft far network,through the soft far network fuzzy matching knowledge base and the weighted gray mean feature of the region of inter-est,the target greater than the warning condition is selected as the perceived infrared weak and small real target,and the perception result is output.Experimental results show that this method can accurately perceive single infrared weak and small targets and multiple infrared weak and small targets in different scenes,and still has good infrared dim small target perception performance in high noise environment.

关 键 词：机器视觉技术 红外弱小目标 搜索窗口 筛选条件 感兴趣区域 Soft-FART网络 

分 类 号：TN299[电子电信—物理电子学]