Controlling seamless steel pipe thickness in production process using computer vision and deep-learning techniques

In this paper a computer vision-based method for detecting thickness distortion of seamless steel pipes produced by Iran National Steel Industrial Group (INSIG) is presented. The main idea of the proposed method is that, thickness distortion can be detected by measuring the pipe length since the mass of steel block for pipe production is identical for all pipes. So, in this paper a method for real-time measuring of the pipe length is presented using the images captured by an IP camera. In the object detection step of the proposed method, we calculate SSD score of each image to find a bounding box enclosing the pipe. By having the coordinates of start and end point of the pipe length calculation is done by comparing with a predefined index. We compared our object detection results with state-of-the-art deep learning-based object detectors. The experimental results show that the proposed method is superior in sense of precision and computation speed.

Controlling seamless steel pipe thickness in production process using computer vision and deep-learning techniques Keywords:

Computer vision , Object detection , Image-based measuring , Seamless pipe , Deep learning