Abstract:
Hemodialysis filters are widely used in the treatment of kidney diseases. In order to reduce the occurrence of medical accidents, they need to go through a strict inspect...Show MoreMetadata
Abstract:
Hemodialysis filters are widely used in the treatment of kidney diseases. In order to reduce the occurrence of medical accidents, they need to go through a strict inspection process before being put into use to avoid defective products from entering the market, so it is crucial to locate and classify hemodialysis filters defects. In this paper, a deep learning method based on machine vision is proposed to solve the problem that the background of the inspected products is very similar to the defects and difficult to distinguish them. First, image denoising is performed, then the center of the product is located and the circle is expanded into a rectangle, the edge is extracted using a firstorder difference operator to distinguish the background from the defects, and finally the image after the extended channel is provided to the lightweight U-Net for training by online sample sampling. The test results show that the proposed method have achieved 0.78 mean IoU, 4.67% false detection rate and 0.67% missing rate on the hemodialysis filter dataset, which demonstrates the effectiveness of the proposed method.
Published in: 2023 26th International Conference on Computer Supported Cooperative Work in Design (CSCWD)
Date of Conference: 24-26 May 2023
Date Added to IEEE Xplore: 22 June 2023
ISBN Information: