摘要
为研究机器视觉的战场适应性,分析了战场环境中不确定性因素对军事目标机器视觉探测的影响。研究针对Faster R-CNN、YOLOv4和CenterNet 3种目标检测模型,利用样本数据充分训练,确保检测模型有足够高的检测效率。随后,在检测数据中引入不确定性因素,包括目标特性和背景的不确定性。分析表明:不确定性因素的引入确实能显著降低目标检测模型的检测效率。这意味着基于军事运用的机器视觉研究应该充分考虑战场环境的影响,也意味着隐身技术措施必须提高对抗目标自动识别技术方面的潜力。
In order to study the battlefield adaptability of machine vision,the influence of uncertain factors in battlefield environment on military target detection by machine vision was analyzed.Aiming at three target detection models of Faster R-CNN,YOLOv4 and CenterNet,this paper used the sample data to fully train to ensure that the detection model has enough high detection efficiency.Then,the uncertainty factors were introduced into the detection data,including the uncertainty of the target characteristics and the uncertainty of the background.Comparative analysis shows that the introduction of uncertainty can significantly reduce the detection efficiency of the target detection model.This means that the influence of battlefield environment should be fully considered in the research of machine vision based on military application,and it also means the potential of stealth technology measures in the aspect of automatic target recognition technology.
作者
吴晓强
曾朝阳
WU Xiaoqiang;ZENG Zhaoyang(Field Engineering College, Army Engineering University of PLA, Nanjing 210007, China)
出处
《兵器装备工程学报》
CSCD
北大核心
2022年第2期181-185,共5页
Journal of Ordnance Equipment Engineering
基金
国防科技重点实验室项目(6142206190204)。
关键词
军事目标
适应性
目标特性
目标检测模型
military target
adaptability
target characteristic
target detection model
