基于偏振特征的水下目标检测

Underwater target detection based on the polarization features

近年来机器视觉技术在目标检测领域得到广泛的应用,其中最常用的是基于Treisman理论的Itti模型。传统基于Itti模型的检测方法往往过分依赖于光谱图像,然而水下环境完全迥异于陆地环境,光谱信息几乎消失殆尽,简单地将Itti模型直接应用于水下并不能取得很好地效果。基于此本文提出了一种将偏振信息与Itti模型相结合的水下目标检测模型。主要思想是:与传统Itti模型提取颜色、亮度、方向等特征不同,针对水下环境的特殊性,分析并提取了更加适合水下环境目标检测的偏振度特征、边缘特征以及直线特征进而利用Itti模型生成显著图,实现水下目标检测。实验结果表明该方法针对水下目标检测有很好的效果。

The traditional detection methods based on the Itti model always depend on the spectral images, however, underwater environment is completely different from terrestrial environment and spectrum information almost disappeared. So the traditional methods would not get defective results when directly used in underwater environment. This paper proposes a novel underwater target detection method combining polarization information and Itti model, the main idea is that aiming at the specificity of underwater environment, this paper extract the polarization feature, edge feature and linear feature which are more suitable for target detection than the color, brightness, and direction features. Then we use the Itti model generate saliency map to detect the underwater target. The experimental results show that our method has a better effect than others.