Magnetic Tile Surface Defect Detection Algorithm Based On Densely Generative Adversarial Network

With the development of intelligent manufacturing,automated defect detection technology based on digital image processing and computer vision is widely used in various manufacturing industries.As an important component of a permanent magnet motor,the magnetic tile plays a direct and important role in the performance and lifetime of the motor.At present,most magnetic tile manufacturers use manual visual inspection of magnetic tile defect,which leads to high labor costs and slow detection speed.In addition,the magnetic tile surface defect is complex and small in size,which is difficult to be distinguished from the detection background.All of these make it difficult for the traditional digital image processing method to extract the appropriate feature vector to detect magnetic tile defect.Thus,the detection method based on computer vision and convolutional neural networks technique will be a better choice.Convolutional neural networks have excellent learning and feature extraction abilities for image data,and they have been applied to the surface defect detection of magnetic tile,and have achieved better detection performance.To address the problem of the complex surface defect of magnetic tile,this paper proposes a magnetic tile surface defect detection algorithm based on densely generative adversarial network to realize the intelligent detection of magnetic tile defect.The algorithm combines deep learning semantic segmentation technology with generative adversarial networks one,so that the semantic segmentation map of magnetic tile surface defect can be successfully obtained.Subsequently,the detection and discrimination technology can be used to judge whether magnetic tile has defect or not.Compared with other convolutional neural networks,it is verified that densely generative adversarial network algorithm has best defect detection capability,and it becomes the core algorithm of magnetic tile surface defect detection task,which is introduced in this paper.