摘要
可见光低慢小飞行目标检测技术在军用民用领域有着特殊的意义,当视频背景中包含动态干扰、复杂云像等复杂情况时,检测诸如民用无人机等低慢小飞行目标十分困难,为此本文提出了一种基于视觉显著性的飞行目标智能检测算法,该算法首先通过帧间差分法提取运动信息,再利用改进SR算法对运动目标周边进行检测,检测时,首先通过局部复杂度分类模块对运动信息进行分类,排除地面的动态干扰信息,再提取目标邻域LAB空间中B通道图像,再对该图像进行云、天边缘部分提取,随后将其与SR算法的输出进行归一化做差获取最终检测结果。实验结果表明该算法在地空背景、复杂云像背景、过曝光背景中可以良好工作,并能达到实时处理需求。
The detection technology of small flying object with low speed and low altitude has special meaning in the military and civil fields.When the background of the video includes complicated situations like dynamic interference on the ground,complex cloud image and overexposure area,the detection of small flying object with low speed and low altitude such as civil unmanned aerial vehicle will become very difficult.To solve this problem,this paper proposes an intelligent flying object detection algorithm based on visual saliency.Firstly,the algorithm extracted motion information by inter-frame difference method,and then used the improved spectral residual(SR)algorithm to detect the margin of the moving target.When detecting,firstly,the motion information was classified by the local complexity classification module to eliminate the dynamic interference information of the ground,and then the image in B channel in the neighborhood Lab color space(LAB)space was extracted,and further,the image was extracted from the cloud and the sky edge portion,and next,the image was normalized with the output of the SR algorithm to obtain the final detection result.The experimental results show that the algorithm can work well in the ground-space background,complex cloud image background,over-exposure background,achieving real-time processing requirements.
作者
刘鲁涛
王晓
李欣雨
LIU Lutao;WANG Xiao;LI Xinyu(College of Information and Communication Engineering,Harbin Engineering University,Harbin 150001,China)
出处
《应用科技》
CAS
2019年第1期88-93,共6页
Applied Science and Technology
基金
国家自然科学基金项目(61201238)
关键词
空中目标
运动目标检测
局部复杂度分类
显著性检测
无人机
动态干扰
复杂背景
边缘提取
aerial target
moving target detection
local complexity classification
saliency detection
UAV
dynamic interference
complex background
edge extraction
