利用图像区域自然尺度特征的海面舰船目标检测

Ship target detection on the sea surface based on natural measure feature of image block

为了提高动态海面背景中目标的检测性能,提出了一种新的基于图像区域自然尺度特征的海面舰船目标检测算法。算法将一维信号的自然尺度特征提取方法拓展到二维图像信号,通过相空间重构和分类,提取了图像灰度时间序列的自然尺度特征。利用BP神经网络,得到背景与目标自然尺度特征的辨识模型,然后对图像序列的自然尺度进行分类,检测得到舰船目标。实验结果表明,新算法检测海面背景下的舰船目标时获得了较高的检测率,花费了较少的检测时间,具有较好的类间可分性和较小的计算量。

In order to improve the detection capability of target on dynamic sea background, a novel algorithm for ship target detection on sea surface based on natural measure feature of image block was proposed. The proposed algorithm extended the extraction of the natural measure feature from single dimensional signals to two dimensional image signals and extracted natural measure feature of image gray level time series through reconstruction and classification of phase space. Using the BP neural networks, the proposed algorithm got the identification model of background and target, classifed the natural measure feature of image sequence and detected ship targets. The experimental results indicate that the novel algorithm not only has highly detection rate and spend less time, but also has good divisibility between classes and small calculated amount for ship target detection on sea surface.