High frequency surface wave radar (HFSWR) is well proved to have over the horizon (OTH) detection capability to weak aerial targets, such as concealed airplanes or cruise missiles. The most important problem of detect...High frequency surface wave radar (HFSWR) is well proved to have over the horizon (OTH) detection capability to weak aerial targets, such as concealed airplanes or cruise missiles. The most important problem of detection of fast and small targets using HFSWR is earlier warning, i.e. enlargement of detection range oftargets. Therefore, the detection threshold should be decreased as low as possible, but numerous false alarms are brought about at the same time. On this condition, conventional track initiation techniques, which normally require the probability of false alarm to be at the level of 10-6, will initiate enormous false tracks and lead to abnormal operation of tracking system. An adaptive modified hough transform (AMHT) track initiator is proposed accordingly and the relation of detection range to the performance of track initiator is analyzed in this paper. Simulations are performed to confirm the capability of track initiation to fast and small targets in dense clutter by AMHT track initiator. The tolerable probability of false alarm of detector can reach the level of 10 -3 . And it performs better than track initiator based on modified hough transform (MHT).展开更多
针对现有基于图域变换的雷达海面小目标检测算法仅考虑单个样本之间的相关性,存在性能不佳、复杂度高等问题,提出了一种基于序数图变换的改进算法。首先,将雷达回波信号的自相关函数转换成可有效获取状态向量之间关联性的序数图,进而提...针对现有基于图域变换的雷达海面小目标检测算法仅考虑单个样本之间的相关性,存在性能不佳、复杂度高等问题,提出了一种基于序数图变换的改进算法。首先,将雷达回波信号的自相关函数转换成可有效获取状态向量之间关联性的序数图,进而提取该图的无符号拉氏矩阵最大特征值作为检测统计量,以实现区分海杂波和目标信号的目的。以McMaster大学提供的实测全相参X波段雷达数据集为对象,对算法的性能进行了广泛评估,结果表明所提算法的平均接收机工作特性曲线下面积(Area under Curve,AUC)可达0.907,优于现有算法,且其计算复杂度适中。展开更多
文摘High frequency surface wave radar (HFSWR) is well proved to have over the horizon (OTH) detection capability to weak aerial targets, such as concealed airplanes or cruise missiles. The most important problem of detection of fast and small targets using HFSWR is earlier warning, i.e. enlargement of detection range oftargets. Therefore, the detection threshold should be decreased as low as possible, but numerous false alarms are brought about at the same time. On this condition, conventional track initiation techniques, which normally require the probability of false alarm to be at the level of 10-6, will initiate enormous false tracks and lead to abnormal operation of tracking system. An adaptive modified hough transform (AMHT) track initiator is proposed accordingly and the relation of detection range to the performance of track initiator is analyzed in this paper. Simulations are performed to confirm the capability of track initiation to fast and small targets in dense clutter by AMHT track initiator. The tolerable probability of false alarm of detector can reach the level of 10 -3 . And it performs better than track initiator based on modified hough transform (MHT).
文摘针对无人机(UAV)影像中道路小目标漏检和目标检测精度低、鲁棒性差等问题,设计一种基于全局特征提取的UAV道路病害检测算法GFE-RDD(Global Feature Extraction-Road Disease Detection)。将卷积神经网络(CNN)与Transformer融合的GFE-Transformer模块嵌入主干网络,提升捕获长距离依赖关系的能力以获得全局上下文信息。为了更好地检测出小目标的道路病害,提出一个融合高效双通道注意力机制(EDA)的小目标检测头。另外,采用WIoUv3(Wise-Intersection over Union vision 3)作为网络的损失函数,解决训练数据中锚框质量差异较大的问题,并提高检测的准确性。在自制的道路多病害数据集上的实验结果表明,所提算法在道路病害检测任务中的F1分数达到0.765,mAP50达到0.796,均高于DETR(DEtection TRansformer)等当前主流算法,取得了较高的检测准确率。
文摘针对现有基于图域变换的雷达海面小目标检测算法仅考虑单个样本之间的相关性,存在性能不佳、复杂度高等问题,提出了一种基于序数图变换的改进算法。首先,将雷达回波信号的自相关函数转换成可有效获取状态向量之间关联性的序数图,进而提取该图的无符号拉氏矩阵最大特征值作为检测统计量,以实现区分海杂波和目标信号的目的。以McMaster大学提供的实测全相参X波段雷达数据集为对象,对算法的性能进行了广泛评估,结果表明所提算法的平均接收机工作特性曲线下面积(Area under Curve,AUC)可达0.907,优于现有算法,且其计算复杂度适中。
