基于视觉注意机制的遥感图像船只检测

摘要

针对视觉注意机制中Itti模型在显著性区域提取中提取颜色、方向和亮度3种特征，应用到遥感图像中容易造成错检的问题，在Itti模型基础上提出了改进的显著性模型，用纹理特征代替Itti模型中的颜色特征和亮度特征，同时考虑了图像在频域中的显著性。对改进后的算法进行仿真，仿真结果表明，改进算法检测精度高、实时性好，显著区域与目标基本一致，能够有效地检测遥感图像中船只目标。%The Itti model only extracts color,orientation,and intensity features in saliency area in visual attention mechanism, which has high false detection ratio in remote sensing.We proposed an improved Itti model,which puts texture feature into model and considers the saliency in frequency domain.Results show that this improved algorithm provides high detection ratio,pretty good realtime capability.The saliency area is basically identical to the object,which can effectively detect ship object in remote sensing.