| abstract |
The invention belongs to the field of deep learning and sonar image processing, and in particular relates to a deep learning-based sonar image underwater pipeline detection method. The invention includes the following steps: constructing a side scan sonar image sample data set; preprocessing the images in the data set; constructing a deep convolutional neural network for judging the laying mode of the underwater pipeline and detecting the position of the pipeline, and training the weight of the network Obtain the trained network; judge the laying method of the underwater pipeline in the preprocessed side-scan sonar image and give the bounding box set of the location; obtain the center position line of the underwater pipeline according to the center point of the bounding box set, according to the surrounding The box set covers the area to segment out the object. Compared with the existing method, the method of the patent can more accurately determine the laying method of the underwater pipeline, more accurately detect the position of the underwater pipeline and its center position line, and has strong generalization ability, and can be used in the parallel acceleration unit. With hardware support, the detection speed is fast and the efficiency is high. |