End Image Defect Detection of Float Glass Based on Faster Region-Based Convolutional Neural Network

Abstract

The float glass contains various defects for reasons of raw materials and production process. These defects can be observed on the end images of the glass. Since the defects are correlated with specific links of the production process, it is possible to discover the process problems by identifying the location and type of defects in end images. Based on faster region-based convolutional neural network (Faster RCNN), this paper proposes a deep learning method that improves the feature extraction network, and adds a Laplacian convolutional layer to preprocess the end images. Considering the defect features in end images, the anchor box size was adjusted to speed up the training. Besides, the lack of generalizability induced by small dataset was solved through data enhancement. With improved VGG16 as the feature extraction layer, a glass defect detection model was established, whose generalizability was improved through transfer learning. The experimental results show that the proposed model achieved a mean detection accuracy of 94% on actual test set, meeting the requirements for actual use in factories.

Using Data From Similar Systems for Data-Driven Condition Diagnosis and Prognosis of Engineering Systems: A Review and an Outline of Future Research Challenges

“…Yang et al [330] and Kim et al [331] also used such pretrained networks to detect the water-binder ratio of concrete and to monitor the quality in laser-assisted micromilling of glass, respectively. Similar approaches were also presented in [332] for the quality assessment of flat glass using images of the glass produced and in [333], [334], and [335] for detecting defects in welded joints.…”

Green Bean Classification: Fully Convolutional Neural Network with Adam Optimization

“…The test results were obtained by utilizing epoch 30 on a data scenario with a 80:20 split. The third test of the optimization technique yielded the following results, as shown in Table 5: The Adam optimizer has exceptional performance in the early epochs (20) for both the 70:30 and 80:20 data scenarios, with an accuracy of 75% in both instances. Nevertheless, as the ratio of testing data climbs to 90:10, there is a notable decline in accuracy.…”

“…Convolution operations involve the manipulation of two functions that take real values as arguments. At this stage, convolutions are applied to the output of the preceding layer using the convolutions operation [20]. This technique utilizes output functions as feature maps for the input image.…”

Artificial Intelligence for Product Quality Inspection in Manufacturing Industry - Online Detection of Edge Defects on Inorganic Solid Material

“…The experiment results showed that this detection method has a high recognition accuracy on both original and enhanced datasets of the curved glass. Another application based on a faster region-based convolutional neural network (Faster RCNN) for end image defect detection of oat glass was reported in (17). The paper proposed a deep learning method that improves the feature extraction network and adds a Laplacian convolutional layer to preprocess the end images.…”