牛畅, 尹奎英, 黄银和. 无人机对地目标自动检测与跟踪技术[J]. 应用光学, 2020, 41(6): 1153-1160. DOI: 10.5768/JAO202041.0601003
引用本文: 牛畅, 尹奎英, 黄银和. 无人机对地目标自动检测与跟踪技术[J]. 应用光学, 2020, 41(6): 1153-1160. DOI: 10.5768/JAO202041.0601003
NIU Chang, YIN Kuiying, HUANG Yinhe. Automatic target detecting and tracking technology based on UAV ground target images[J]. Journal of Applied Optics, 2020, 41(6): 1153-1160. DOI: 10.5768/JAO202041.0601003
Citation: NIU Chang, YIN Kuiying, HUANG Yinhe. Automatic target detecting and tracking technology based on UAV ground target images[J]. Journal of Applied Optics, 2020, 41(6): 1153-1160. DOI: 10.5768/JAO202041.0601003

无人机对地目标自动检测与跟踪技术

Automatic target detecting and tracking technology based on UAV ground target images

  • 摘要: 针对无人机图像帧序列具有平台高速运动,视角旋转强烈,需要实时处理等特点,提出一种基于双级旋转不变特征空间检测(粗匹配-精细匹配)与并行特征提取跟踪的无人机对地目标图像帧序列自动快速目标检测与跟踪算法。采用图像子块的平均灰度值、灰度值方差、灰度值梯度构建特征空间。通过构造图像特征空间的方法来快速筛选待匹配图像的可疑区域,删除大量的背景区域,检测算法使用全局初步匹配加局部精细匹配的方法来规避算法复杂度的缺陷。理论及实验分析表明:该算法实时性强,对图像的旋转畸变具有抵消作用,对异常情况可以恰当处理,且全局初步匹配流程具有可移植性,可以在其他图像匹配跟踪算法中充当预处理器。实验结果表明:该算法在无人机对地的情况下可以保证对地面目标的稳定跟踪,配套检测算法具有较好的实时性,满足无人机图像目标检测跟踪实时处理的需要。

     

    Abstract: Since UAV(unmanned aerial vehicle) image frames has characteristics such as high speed moving platform, intensity view rotation, real-time processing and so on, a UAV ground target image frames detecting and tracking algorithm based on two-level rotation invariant feature space detection and tracking on account of parallel feature extraction was put forward. The detection algorithm used primary matching and meticulous matching to reduce the algorithm complexity. The theoretical analysis and experiments showed that the proposed algorithm has better real-time performance, insensitive to image rotation and have better performance on abnormal conditions and the primary overall-match process in proposed algorithm can act as the pretreatment of other image matching and tracking algorithms. The experiment results indicate that the proposed algorithm can ensure stable target tracking under the situations of UAV images. At the same time, the detecting process has better real-time performance, which can meet the real-time processing demand of target detection and track of UAV image frames.

     

/

返回文章
返回