Defect Detection Of Sanitary Ceramics Based On Deep Learning

Sanitary ceramic defect detection is an important link in the production process of Sanitary Ware,and it is very important for the control of product quality.At present,the defect detection of finished sanitary ceramic parts mainly relies on manual detection.It takes lots of time to manually complete the defect detection of a finished product.The detection results are closely related to the experience of the worker and the accuracy is difficult to guarantee.In order to improve the automation of sanitary ceramics production line,reduce production costs,and improve the accuracy and speed of defect detection,this paper studies the sanitary ceramics defect detection algorithm based on deep learning for the surface defects of sanitary ceramics.The main contents of this paper are as follows:Firstly,this paper introduces a variety of object detection models,and compare the performance of these models on open resource data sets.Aiming at the variety of sanitary ceramics surface defects and the complex background in this paper,the YOLOv3 object detection model was finally selected as the benchmark algorithm for this paper.Secondly,I go to the production base in Jiangmen,Guangdong to collect the original images.In view of the possible noise in the original images and the actual deployment of the model,the factory production environment is noisy,which may also affect the image quality.Image preprocessing is performed on the original images.In detail,the median filter is used to denoising.Aiming at the problem of imbalance data in the number of various defect samples,a data augmentation method is used to solve the problem.The data is expanded by image processing methods such as rotation,flip,zoom,etc.,and a sufficient data set is obtained.The open resource image auxiliary software Label Img is used for labeling all the images.These images are made into a data set,and divided into training set,validation set and test set according to the ratio of 8:1:1.Finally,the original YOLOv3 network model was constructed,and the sanitary ceramic defect images are feed into the network model in batches according to the computing power of the hardware equipment to train the YOLOv3 model.Then,the experimental results and the characteristics of the surface defects of sanitary ceramics are in-depth analyzed.According to that,corresponding improvement strategies are proposed.The backbone network,the multi-scale feature map,the anchor box and activation function of YOLOv3 are changed.Then the performance of the new model is verified through experiments.On the test set,it achieved 94.90% mAP,and reached 25 FPS.In summary,this paper proposes a deep learning-based detection method for sanitary ceramic surface defects,which greatly improves the detection efficiency compared to manual detection.It can meet the real-time detection requirements of sanitary ceramic surface defects under complex backgrounds,which is useful for improving the efficiency of sanitary ceramics defect detection.This new method is great significance for economic of sanitary ceramics industry.