基于深度学习的金属焊接管道内壁缺陷检测方法研究

针对目前管道内壁缺陷检测方法不足的问题，提出了一种基于管道机器人和深度学习模型算法的管道内壁缺陷检测方法，对管道内壁缺陷图像进行识别与分类。通过对Faster RCNN目标检测算法进行改进，以密集连接卷积网络（DenseNet）作为检测模型的特征识别核心，从而提高了模型的泛化能力和识别精度。试验结果表明，基于深度学习的识别方法实现了金属焊接管道缺陷的检测，运用改进后的Faster RCNN深度学习算法进行管道缺陷识别具有识别精度高、成本低的优点，平均准确率达到93.2%。

Research on Inner Wall Defect Detection Method of Metal Welded Pipe Based on Deep Learning

In order to solve the problem of insufficient detection methods for pipeline inner wall defects, a pipeline inner wall defect detection method based on pipeline robot and deep learning model algorithm is proposed to identify and classify the pipeline inner wall defect image. Through the improvement of Faster RCNN target detection algorithm, dense connected convolution network (DenseNet) is used as the core of feature recognition of detection model, so the generalization ability and recognition accuracy of the model are improved. The experimental results show that the method based on deep learning can detect the defects of welded metal pipes. The improved Faster RCNN algorithm has the advantages of high recognition accuracy and low cost, with an average accuracy of 93.2%.