论文部分内容阅读
传统的CFAR检测应用到光学卫星遥感图像舰船目标检测中时不能对黑极性目标进行判断,针对此提出改进的基于广义似然比检验(Generalized Likelihood Ratio Test,GLRT)的舰船目标检测算法。该算法采用滑动窗口检测形式,在假设背景和目标灰度均服从高斯分布的前提下,通过GLRT判断背景窗口与目标窗口是否同分布来检测目标,兼顾了目标黑白两种极性的情况。算法实现中对图像进行了分块检测,并通过形态学处理对检测结果进行了目标聚类。采用SPOT5与CBERS实测数据进行实验,验证了海背景服从高斯分布的假设。典型数据检测结果表明,该算法可以检测黑极性目标,且相比CFAR虚警率更低,大量数据计算ROC曲线的结果以及比CFAR检测少约40%的耗时进一步表明该算法性能更优。
The traditional CFAR detection can not judge the black polar target when applied to the satellite target detection of optical satellite remote sensing images. In view of this, an improved GLRT (Generalized Likelihood Ratio Test) based ship target detection algorithm . The algorithm uses the sliding window detection form. Under the premise of assuming Gaussian distribution of the background and the target grayscale, GLRT judges whether the background window and the target window are the same distribution to detect the target, taking into account the target black and white two polarities. In the algorithm implementation, the image was divided into blocks, and the morphological processing was used to cluster the detection results. The experimental data of SPOT5 and CBERS were used to verify the hypothesis that the background of the sea obeys Gaussian distribution. The results of typical data detection show that this algorithm can detect black-polar targets and has a lower false alarm rate than CFAR. The results of large-scale data calculation of ROC curves and the time-consuming of about 40% less than CFAR detection further show that the proposed algorithm has better performance .