Surface Defect Detection Algorithm Of Domestic Ceramics Based On Convolutional Neural Network

The strategic idea of "Made in China 2025" has been advanced in light of the advancement of artificial intelligence technology.Chinese manufacturing companies have implemented technological innovation in response to state policies to increase production efficiency,and Chinese manufacturing facilities are becoming more sophisticated and intelligent.Only the quality inspection link still uses the manual quality inspection method,despite the fact that certain everyday ceramic businesses currently exist that have essentially automated manufacturing.The typical machine vision method,which designs a specific feature operator for each imperfection,is ineffective and unable to handle the complex variations of everyday ceramic surface flaws.Manual quality inspection is not only ineffective but also has a negative influence on the health of the workers.This research investigates a convolutional neural network-based method for domestic ceramics surface defect identification in order to address the aforementioned issues.The traditional target detection technique based on convolutional neural networks is often built for large and medium target detection,but ignores the processing of small target feature information,making it unsuitable for the surface defect detection of domestic ceramics.To solve this problem,this paper proposes a defect detection algorithm based on recursive gated convolution and mixed attention.In this algorithm,the full-dimensional dynamic convolution is introduced to replace a portion of the ordinary convolution,the recursive gated convolution C3 module is used to strengthen the high-order feature interaction ability of the YOLOv5 s model,and the mixed attention module is designed to be added at the end of the Neck part to improve the feature fusion network’s ability to focus on the feature information of small targets.Finally,CIoU takes the role of SIoU to speed up the network’s training.Experiments show that the technique can improve the precision of surface fault detection in domestic ceramics.Additionally,this research suggests a defect identification method integrating location attention and improved texture features in light of the small and similar defect kinds.In order to save computation,this approach eliminates the superfluous network detection layer and C3 module of YOLOv5,and creates the GBC3 lightweight module to replace the other C3 modules in the backbone network.The primary information processing positions of the YOLOv5 s model incorporate a fusion module that enhances texture and position features to increase detection performance.Studies reveal that the proposed method is more accurate and uses a lighter model than the one that was previously used.This paper also built and constructed an application system for surface defect identification of domestic ceramics in accordance with the algorithm.The system has the statistical capability to provide defect information and unqualified product information,which can assist business managers in enhancing production measures and production quality.It can not only detect surface flaws in domestic ceramics.In this paper,a surface defect detection algorithm for domestic ceramics using convolutional neural networks is studied.This is an attempt to apply convolutional neural networks to the detection of domestic ceramics and also an expansion of the target detection algorithm’s application scenario.This method not only helps domestic ceramic businesses save time and labor costs,but it also gives defect classification and identification a more objective definition,enhances the ability to detect surface flaws in domestic ceramic products,and thus raises the overall quality of domestic ceramic products.