摘要
为提高复杂背景下红外小目标的检测能力,提出了一种联合梯度判别与自适应匹配的红外小目标检测算法。通过多方向梯度特征在图像中首次筛选出疑似目标区域,利用区域内灰度信息生成自适应模型,进行再次判断。针对梯度判断与自适应模型匹配分别建立量化评价,通过加权方式引入置信度函数对不同疑似目标区域进行评估,筛除疑似目标。为使算法能在无人机等动载平台应用,搭建嵌入式系统,通过红外相机取景,实现探测系统对真实场景下小目标的检测。通过对不同的公共数据集测试,不同复杂场景下与加权增强局部对比度算法(weighted strengthened local contrast measure, WSLCM)、三层滤波窗口局部对比度算法(tri-layer local contrast measure, TLLCM)算法进行比对,所提算法有较好的适应能力,不同样本下的识别率在92%以上。该算法通过在嵌入式平台IP核的定制和软硬协同设计,进行了硬件加速处理,实时视频帧率>30 frame/s,验证了其有效性。
In order to improve the detection ability of infrared small targets in complex backgrounds,a small infrared target detection algorithm based on joint gradient discrimination and adaptive matching was proposed.The suspected target area is screened out for the first time in the image through multi-directional gradient features,and the adaptive model is generated by using the grayscale information in the region for re-judgment.Quantitative evaluation was established for gradient judgment and adaptive model matching,and confidence functions were introduced to evaluate different suspected target areas and screen out suspected targets.In order to enable the algorithm to be applied in dynamic platforms such as UAVs,an embedded system was built to realize the detection of small targets in real scenes by the detection system through infrared camera framing.By testing different public datasets and comparing with WSLCM(weighted strengthened local contrast measure)and TLLCM(tri-layer local contrast measure)algorithms in different complex scenarios,the proposed algorithm has good adaptability,and the recognition rate under different samples is more than 92%.The algorithm is processed by hardware acceleration through the customization of the IP core of the embedded platform and the co-design of software and hardware,and the real-time video frame rate is greater than 30 frame/s,which verifies its effectiveness.
作者
王磊
孙立业
徐昕阳
冯凯
WANG Lei;SUN Liye;XU Xinyang;FENG Kai(College of Mechanical Engineering,Shenyang Ligong University,Shenyang 110159,Liaoning,China;National Key Laboratory of Electromagnetic Space Security,Tianjin 300308,China)
出处
《弹箭与制导学报》
北大核心
2025年第3期273-280,共8页
Journal of Projectiles,Rockets,Missiles and Guidance
基金
国家自然科学基金(U2341245)
辽宁省属本科高校基本科研专项资金(LJ232410144074)。
关键词
红外小目标
目标检测
梯度判断
自适应
模型匹配
infrared small target
target detection
gradient judgment
adaptive
model matching
