基于改进RGHS和Canny算子的水下图像边缘检测方法

Underwater Image Edge Detection Method Based on Improved RGHS and Canny Operator

无人船航行过程中容易遇到漂浮物、礁石、冰山等障碍物,需获取障碍物水上、水下的完整轮廓图像,以进行有效避障。由于水下环境和水对光波的折射、反射、吸收等原因,直接对相机获取的障碍物水下图像进行边缘探测难以达到理想效果。为此,首先,采用相对全局直方图拉伸方法实现障碍物水下图像的去雾处理,获取障碍物水下图像色彩恢复后的图像信息;然后,利用改进Canny算子的边缘检测算法进行障碍物水下图像外轮廓和内轮廓检测;最后,根据图像中水平面的成像特征进行拟合寻找水平线,去除倒影。实验结果表明:该方法能够有效获得完整的障碍物水下图像轮廓。

In the course of navigation,unmanned ship is prone to encounter obstacles such as floating objects,reefs,icebergs,etc.the complete contour images of obstacles above and below water are needed to avoid obstacles effectively.However,because of underwater environment and refraction,reflection and absorption of light waves,it is difficult to detect the underwater image of obstacles directly.Therefore,the relative global histogram stretching method is used to remove fog from underwater image of obstacles to obtain the image information after the color recovery of the underwater image of obstacles.Then,the edge detection algorithm of improved Canny operator is used to detect the outer contour and inner contour of underwater image of obstacles;finally,the image is fitted to find water according to the imaging characteristics of the horizontal plane in the image Flat line,remove reflection.The experiment shows that the method can obtain the complete underwater image contour of obstacles effectively.