基于机器视觉的ROV水下管线自动跟踪方法

针对混浊水下管线检测过程中由于光学图像成像不清而难以实现自动跟踪的问题,提出一种基于机器视觉和声纳图像处理的水下管线自动跟踪方法。鉴于声纳图像噪声大的问题,该方法采用Gabor滤波器增强管线特征,然后通过二值化,以及Canny边缘提取管线轮廓,最后通过霍夫变换得到管线的几何特征。为了提高管线提取的效率,采用卡尔曼滤波对管线走向和位置进行预测。通过ROV实验,结果表明,该方法能够减少图像搜索的面积,减少处理时间,成功地提取出管线特征。该方法应用于混浊水下管线跟踪是可行的、有效的。

A Detection Method Based on Sonar Image for Underwater Pipeline Tracker

For the problem of unclearness of pipeline detection optical image in turbid underwater, a detection method based on sonar image for underwater pipeline tracter is proposed. In view of the problem of sonar image with large noise, the method uses Gabor filter to enhance pipeline feature, then pipeline profile is extracted through binarization and Canny edge, finally geometrical characteristics of the pipeline is obtained through the Hough transform. To improve the efficiency of the pipeline extraction, direction and location of the pipeline are forecasted using Kalman filter. Through ROV experiments, The results show that the method can reduce the image search area and the processing time, Pipeline characteristics are successfully extracted, which used in turbid underwater pipeline detection is feasible.