As samples of steel defects are industrially limited, it is challenging for most deep learning methods that rely on ample labeled data to identify steel surface defects. Recently, contrastive learning has achieved good performance in natural image classification tasks with few labeled samples, yet two obstacles prevent its effective application to steel surface defect images. One is that due to the presence of inter-class and intra-class similar samples in steel surface defect, the fixed contrast strength in contrastive learning will destroy the potential semantic information of defect samples. Another is that contrastive learning requires a large amount of unlabeled data, whereas steel surface defect samples are insufficient. To overcome the above-mentioned problems, a novel framework named flexible and diverse contrastive learning (FDCL) is proposed. This framework consists of two parts, flexible contrast (FiCo) and diverse generative adversarial networks (DGANs). Diverse images generated by the DGAN and real images are fed into FiCo for representation learning. In the FiCo, the contrast strength among samples is flexibly adjusted by the proposed variable temperature discrimination and feature reconstruction (FR). In addition, the output features (OF) of FiCo will be used as input to the DGAN generator to improve image quality, thus further facilitating representation learning. The proposed FDCL is implemented on four standard steel surface defect data sets, and the experimental results demonstrated that it achieves superior performance over state-of-the-art methods. Our code is available at: https://github.com-/jiacongc/FDCL .

For the multi-rope structure of the container crane system and its large mass payload anti-swing positioning problem, an equivalent double-swing model based on radial spring-dampers is established and a time-varying PD sliding mode controller (TVPD-SMC) with improved transient performance is proposed. In particular, the dynamics of the container crane system are first analyzed using the Lagrange method, and an equivalent double pendulum dynamics model of the crane is established. Compared with the traditional double pendulum model, this model ensures the accuracy of modeling without measuring the second pendulum angle. On this basis, an enhanced coupled time-varying sliding mode control method is designed to eliminate the sliding mode control method’s reaching phase and improves the robustness of the controller. Finally, the convergence and stability of the closed-loop system are proved using the Lyapunov technique and the Lasalle invariance theorem, and the simulation results demonstrate the effectiveness of the method.

Container marking anomaly detection (CMAD) aims to identify markings that deviate from a standard reference image. Currently, the manual execution of CMAD is inefficient and susceptible to errors. Existing machine vision methods are limited to detecting text-based or specific types of markings, rendering type-agnostic CMAD unattainable. To address this limitation, we propose an innovative framework for CMAD, named Bidirectional Feature Pyramid Siamese Anomaly Detection Network (BiSiNet). BiSiNet consists of the Siamese Encoder (SE), the Bidirectional Feature Pyramid Decoder (BiD), and the Triplet Classifier Prediction Head (TCPH). SE extracts the difference feature pyramid between the target and reference images. BiD decodes the difference feature pyramid in both top-down and bottom-up paths, effectively solving the problem of global information loss during decoding in traditional single-path decoder. TCPH utilizes both local and global features to predict outcomes at both the pixel and patch levels, integrating these results to resolve inconsistencies in predictions. BiSiNet not only possesses strong resistance to disturbances such as reflections and shadows but also has accurate localization capabilities. Due to the limited diversity and the difficulty in obtaining anomalous samples, we designed an innovative Cellular Anomaly Generation Strategy (CAGS). CAGS employs Cellular Noise to generate marking masks, creating realistic pseudo marking images. Through innovative design, it produces four types of samples, thereby enhancing sample diversity and improving training generalization. Extensive experiments using our established ContainerMAD dataset have demonstrated that BiSiNet and CAGS outperform other methods in both container marking anomaly detection and localization. Additionally, favorable outcomes have been achieved on the MvtecAD dataset.

With the widespread use of containers, the demand for Container Marking Detection and Recognition (CMDR) is gradually increasing. The use of deep learning algorithms can greatly improve the efficiency of marking detection and recognition. However, there is still a lack of research on CMDR in both academia and industry, resulting in the current task being completed manually and inefficiently. In this paper, we probe into the importance of data-driven and task paradigms for CMDR tasks. Firstly, we constructed an open large scale container surface marking text dataset called ContainerText. This dataset consists of 12 k high-resolution images and provides two types of annotation information: bounding box used for detection and text for recognition tasks. In addition, we also propose an efficient semi-automatic annotation method based on deep learning, which reduces the cost of manual annotation. Subsequently, we have innovatively proposed a CMDR method combining Scene Text Recognition (STR) with CMDR tasks. The method based on STR can locate and recognize container marking from a fine-grained level. We conducted a comprehensive series of experiments on the ContainerText dataset using state-of-the-art (SOTA) scene text detection and scene text recognition models. The experimental results demonstrate that the CMDR method, based on STR, exhibits exceptional adaptability and feasibility. All experimental results obtained from the ContainerText dataset will act as performance benchmarks for future researchers. Finally, an automated Container Marking Image Acquisition Mechanism (CMIAM) are construucted, which can effectively avoid complex lighting in the workshop environment and achieve high-quality and automated image acquisition. We have conducted extensive experiments to measure the distance, resolution, and field of view required for clearly capturing container markings. Our research providing reference for future CMDR research from task solution and hardware selection.