Roller bearings are some of the most critical and widely used components in rotating machinery. Appearance defect inspection plays a key role in bearing quality control. However, in real industries, bearing defects are usually extremely subtle and have a low probability of occurrence. This leads to distribution discrepancies between the number of positive and negative samples, which makes intelligent data-driven inspection methods difficult to develop and deploy. This paper presents a small data-driven convolution neural network (SDD-CNN) for roller subtle defect inspection via an ensemble method for small data preprocessing. First, label dilation (LD) is applied to solve the problem of an imbalance in class distribution. Second, a semi-supervised data augmentation (SSDA) method is proposed to extend the dataset in a more efficient and controlled way. In this method, a coarse CNN model is trained to generate ground truth class activation and guide the random cropping of images. Third, four variants of the CNN model, namely, SqueezeNet v1.1, Inception v3, VGG-16, and ResNet-18, are introduced and employed to inspect and classify the surface defects of rollers. Finally, a rich set of experiments and assessments is conducted, indicating that these SDD-CNN models, particularly the SDD-Inception v3 model, perform exceedingly well in the roller defect classification task with a top-1 accuracy reaching 99.56%. In addition, the convergence time and classification accuracy for an SDD-CNN model achieve significant improvement compared to that for the original CNN. Overall, using an SDD-CNN architecture, this paper provides a clear path toward a higher precision and efficiency for roller defect inspection in smart manufacturing.

In the field of surface defect detection, the scale difference of product surface defects is often huge. The existing defect detection methods based on Convolutional Neural Networks (CNNs) are more inclined to express macro and abstract features, and the ability to express local and small defects is insufficient, resulting in an imbalance of feature expression capabilities. In this paper, a Multi-Scale Feature Learning Network (MSF-Net) based on Dual Module Feature (DMF) extractor is proposed. DMF extractor is mainly composed of optimized Concatenated Rectified Linear Units (CReLUs) and optimized Inception feature extraction modules, which increases the diversity of feature receptive fields while reducing the amount of calculation; the feature maps of the middle layer with different sizes of receptive fields are merged to increase the richness of the receptive fields of the last layer of feature maps; the residual shortcut connections, batch normalization layer and average pooling layer are used to replace the fully connected layer to improve training efficiency, and make the multi-scale feature learning ability more balanced at the same time. Two representative multi-scale defect data sets are used for experiments, and the experimental results verify the advancement and effectiveness of the proposed MSF-Net in the detection of surface defects with multi-scale features.

With the rapid development of information technology and the widespread use of the Internet, QR codes are widely used in all walks of life and have a profound impact on people’s work and life. However, the QR code itself is likely to be printed and forged, which will cause serious economic losses and criminal offenses. Therefore, it is of great significance to identify the printer source of QR code. A method of printer source identification for scanned QR Code image blocks based on convolutional neural network (PSINet) is proposed, which innovatively introduces a bottleneck residual block (BRB). We give a detailed theoretical discussion and experimental analysis of PSINet in terms of network input, the first convolution layer design based on residual structure, and the overall architecture of the proposed convolution neural network (CNN). Experimental results show that the proposed PSINet in this paper can obtain extremely excellent printer source identification performance, the accuracy of printer source identification of QR code on eight printers can reach 99.82%, which is not only better than LeNet and AlexNet widely used in the field of digital image forensics, but also exceeds state-of-the-art deep learning methods in the field of printer source identification.

In production, due to natural conditions or process peculiarities, a single product often may exhibit more than one type of defect. The accurate identification of all defects has an important guiding significance and practical value to improve the planting and production processes. Concerning the surface defect classification task, convolutional neural networks can be implemented as a powerful instrument. However, a typical convolutional neural network tends to consider an image as an inseparable entity and a single instance when extracting features; moreover, it may overlook semantic correlations between different labels. To address these limitations, in the present paper, we proposed a feature-wise attention-based relation network (FAR-Net) for multilabel jujube defect classification. The network included four different modules designed for (1) image feature extraction, (2) label-wise feature aggregation, (3) feature activation and deactivation, and (4) correlation learning among labels. To evaluate the proposed method, a unique multilabel jujube defect dataset was constructed as a benchmark for the multilabel classification task of the jujube defect images. The results of experiments show that owing to the relation learning mechanism, the average precision of the three main composite defects in the dataset increases by 5.77%, 4.07%, and 3.50%, respectively, compared to the backbone of our network, namely Inception v3, which indicated that the proposed FAR-Net effectively facilitated the learning of correlation between labels and eventually, improved the multilabel classification accuracy.

Estimating the crowd density of public territories, such as scenic spots, is of great importance for ensuring population safety and social stability. Due to problems in scenic spots such as illumination change, camera angle change, and pedestrian occlusion, current methods are unable to make accurate estimations. To deal with these problems, an ensemble learning (EL) method using support vector regression (SVR) is proposed in this study for crowd density estimation (CDE). The method first uses human head width as a reference to separate the foreground into multiple levels of blocks. Then it adopts the first-level SVR model to roughly predict the three features extracted from image blocks, including D-SIFT, ULBP, and GIST, and the prediction results are used as new features for the second-level SVR model for fine prediction. The prediction results of all image blocks are added for density estimation according to the crowd levels predefined for different scenes of scenic spots. Experimental results demonstrate that the proposed method can achieve a classification rate over 85% for multiple scenes of scenic spots, and it is an effective CDE method with strong adaptability.