基于多值免疫网络的红外与可见光协同目标检测  

Target detection in thermal-visible surveillance based on multiple-valued immune network

作　　者：陈炳文[1] 王文伟[2] 秦前清[3] 

机构地区：[1]中国电子科技集团公司二十八所,江苏南京210007 [2]武汉大学电子信息学院,湖北武汉430079 [3]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079

出　　处：《红外与毫米波学报》2014年第6期654-659,共6页Journal of Infrared and Millimeter Waves

基　　金：国家863项目(2012AA091701);国家科技支撑计划项目(2011BAK08B02);湖北省自然科学基金项目(2011CDB452);中央高校基本科研业务费专项资金资助项目(212274467)~~

摘　　要：采用模糊自适应共振神经网络建立红外与可见光各自的背景模型;依据多值免疫网络模型,将红外背景模型视为B细胞,可见光背景模型视为T细胞,设计一系列免疫应答策略来协同建立B细胞与T细胞的交互模型,并以此分析各像素点的背景模糊隶属度来检测目标.实验结果表明,该算法的F1指标高达96.4%,能有效协同互补红外与可见光信息,检测出复杂场景下的目标.Two fuzzy adaptive resonance neural networks were utilized to build the background models of thermal and visible components.According to the multiple-valued immune network model,a series of immune response strategies were designed to cooperate B cell with T cell to build the interactive model,which takes the infrared background model as B cell and the visible background model as T cell.With the interactive model,the targets are detected according to the degree of fuzzy match between pixels and models.Experimental results show that the F1 measurement of the proposed approach is up to 96.4％.It is able to complement information between thermal and visible components effectively.The method is capable of detecting targets in complex scenes effectively.

关 键 词：目标检测 红外与可见光监控 多值免疫网络 模糊自适应共振神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]