一种基于视觉显著图的舰船红外图像目标检测方法

Algorithm for Detecting Ship Target in Infrared Image Based on Saliency Map

提出了一种基于视觉显著图的红外舰船目标定位方法,即通过改进的Itti模型生成视觉显著图,并基于视觉显著图分割出目标区域,从而实现目标检测。先用小波变换替代Itti模型中的高斯滤波来生成图像多尺度金字塔,然后用center-surround算子提取出多尺度的视觉差异特征,并对生成的视觉特征图进行合成,生成显著图。最后,利用阈值分割方法分割出目标区域,并对原始图像进行标记,从而实现目标检测。实验结果表明,与传统的Otsu阈值分割方法相比,该方法能够准确检测出目标区域。

An infrared target location method based on a saliency map is proposed. That is, the target detection is implemented by using an improved Itti model to generate a saliency map and segmenting the target area in the saliency map. Firstly, the wavelet transformation other than the Gaussian filter in the Itti model is used to generate a multi-scale pyramid. Then, the center-surround operators are used to extract multi-scale saliency difference signatures and the generated saliency signature maps are combined into a saliency map. Finally, a threshold segmentation method is used to segment the target area. The original images are marked so as to implement target detection. The experimental result shows that compared with the traditional Otsu threshold segmentation method, this method can detect the target area more accurately.