基于Rail-GPSAR实测数据的地雷检测新方法

本文借鉴恒虚警率(CFAR)思想估计图像数据中杂波的统计量,实现图像的预白化处理。在训练图像中选取地雷图像切片的距离向中心切线作为本征向量构造训练数据阵,获取检测器的参数估计,结合主分量分析设计高效的偏差最优线性二次(DOLQ)检测器。验证了DOLQ检测器输出的灰度图像服从Gamma分布,利用其分布参数和实测数据特点确定整体阈值。对得到的二元图像利用形态学操作获取可疑目标块,然后通过整体阈值和局部门限确定的自适应阈值对其进行二次检测,以确保无漏检的情况下尽可能降低虚警。基于图像中目标能量的空间范围,提出了利用能量环对预白化处理后的图像进行目标检测。最后,提出了一种将能量环与DOLQ检测器的检测结果进行决策级融合的新方法,完成对不同土壤环境、埋设深度地雷的检测。对实际图像的检测结果表明,新方法能够显著降低虚警,且具有较好的稳健性。

A novel method of landmines detection based on the real data of Rail-GPSAR

The statistic of the clutters in images is estimated by exploiting the idea of the constant false alarm rate (CFAR) in this paper and applied to carry out pre-whitening the images.A one-dimensional range cut through the image chip of land mine in the pre-whitened training image is selected as a feature vector to construct the training data matrix.The parameters of the detector can be es- timated with the training set.Combining the principal component analysis (PCA),a computationally efficient deflection-optimal linear- quadratic (DOLQ) detector can be designed.The output of DOLQ detector can be validated obeying Gamma distribution,whose parame- ters can be used to determine the global threshold combining the real data.To ensure all targets are detected and the false alarm is low enough,we exploit morphologic operation on the binary image obtained by the global threshold,then using the adaptive thresholds deter- mined by global and local thresholds to carry out secondary detection on the suspect block images.Based on the spatial extent of energy in the image,a processing technique called energy ring is employed to the detection of buried mines in the pre-whitened images.In the end,a decision-level fusion method between the detection results of DOLQ detector and energy ring is presented for detection of land mines with different soil environments and buried depth.The detection results of the real data show that the new method can significantly reduce the false alarms and possesses a good robustness.