摘要
现有的伪装效果评估主要针对静止的单幅图像,不能很好地模拟侦察人员对目标的判读过程。结合Mean shift目标跟踪技术,提出一种基于特征统计的动态伪装效果评估方法。该方法通过统计目标与背景8联通域的相关性特征数据,建立归一化联合高斯分布,利用概率密度的分布范围评估目标伪装效果。计算联合分布的概率密度时,提出对数放大概率,解决了高维联合分布概率密度数值敏感度低、不便于阈值设定的问题。引入样本更新策略,使样本库按照一定的概率随机更新,从而较好地适应了由于季节交替等因素引起的背景大范围变化。实验过程分别对某一指挥车实施1级伪装、2级伪装和3级伪装。采集数据后计算其对数放大概率并对曲线作出统计,结果表明:实际中划分的3种伪装状态与依据3σ准则预先设定的3种伪装状态完全对应;该模型能够有效反映出目标的伪装效果。
The existing camouflage effect evaluation is mainly for a single still image,which can not simulate the process of the reconnaissance personnel’s interpretation of a target.A feature statistics-based dynamic camouflage effect evaluation method is proposed based on mean shift target tracking algorithm.The proposed method is to establish a normalized joint Gaussian distribution by using the data of correlation features between the target and the background of eight-way domain,and use the distribution range of probability density to evaluate the camouflage effect of target.A logarithmic amplification probability is proposed for calculating the probability density of joint distribution,which solves the problem that the high-dimensional joint distribution probability density has low numerical sensitivity and is inconvenient to set a threshold.At the same time,a sample update strategy is introduced to make the sample library update randomly according to a certain probability,so as to better adapt to the change in large-scale background caused by the turn of seasons and other factors.In the experimental process,the first-level,second-level and third-level camouflages are applied to a certain command vehicle.After collecting the data,the logarithmic amplification probability is calculated,and the statistics on the curves is made.The results show that 3 camouflages divided in reality completely correspond to the first-level,second-level and third-level camouflages pre-set by the 3σcriterion.The experimental results show that the model can effectively reflect the camouflage effect of target.
作者
杨鑫
许卫东
贾其
YANG Xin;XU Weidong;JIA Qi(Field Engineering College,Army Engineering University of PLA,Nanjing 210007,Jiangsu,China)
出处
《兵工学报》
EI
CAS
CSCD
北大核心
2019年第8期1693-1699,共7页
Acta Armamentarii
基金
江苏省自然科学基金项目(BK20180579)
关键词
伪装效果
效果评估
特征统计
概率密度
高斯分布
目标跟踪
camouflage effect
effect evaluation
feature statistics
probability density
Gaussian distribution
target tracking
