基于视觉显著性的水面垃圾目标检测

针对实际水面复杂环境提出了一种基于视觉显著性的水面垃圾目标检测算法。首先对输入图像进行超像素分割，在CIELab、RGB和HSV颜色空间中提取超像素级的显著性特征，然后使用随机森林回归器将显著性特征进行融合得到疑似显著性图，并使用自适应阈值分割得到疑似二值显著性图，最后使用MLP分类器对原始图像中的疑似垃圾目标区域进行判别，去除水波、倒影和反光的干扰，最终检测出水面的垃圾目标。实验结果表明所提基于视觉显著性的水面垃圾目标检测算法的性能优于其他水面目标检测算法。

Water Surface Garbage Object Detection Based on Visual Saliency

A water surface garbage object detection algorithm based on visual saliency is proposed for the complex water surface environment.Firstly,the input image was superpixel segmented,and the superpiexl-level saliency feature were extracted in CIELab,RGB and HSV color spaces.Then,the random forest regression was used to fuse the saliency features to obtain the suspected saliency map,and the adaptive threshold segmentation was used to obtain the suspected binary saliency map.Finally,the MLP classifier was used to discriminate the suspected garbage target area in the original image,the interferences of waves,reflections and reflective were removed,and the garbage target of the water surface was finally detected.The experimental results show that the performance of the proposed water surface garbage object detection algorithm based on visual salience is better than other water surface object detection algorithms.