Rod insulators are vital parts of the catenary of high speed railways (HSRs). There are many different catenary insulators, and the background of the insul

As an important equipment for high-speed railway (HSR) to obtain electric power from outside, the state of the pantograph will directly affect the operation safety of HSR. In order to solve the problems that the current pantograph detection method is easily affected by the environment, cannot effectively deal with the interference of external scenes, has a low accuracy rate and can hardly meet the actual operation requirements of HSR, this study proposes a pantograph detection algorithm. The algorithm mainly includes three parts: the first is to use you only look once (YOLO) V4 to detect and locate the pantograph region in real-time; the second is the blur and dirt detection algorithm for the external interference directly affecting the high-speed camera (HSC), which leads to the pantograph not being detected; the last is the complex background detection algorithm for the external complex scene “overlapping” with the pantograph when imaging, which leads to the pantograph not being recognized effectively. The dirt and blur detection algorithm combined with blob detection and improved Brenner method can accurately evaluate the dirt or blur of HSC, and the complex background detection algorithm based on grayscale and vertical projection can greatly reduce the external scene interference during HSR operation. The algorithm proposed in this study was analyzed and studied on a large number of video samples of HSR operation, and the precision on three different test samples reached 99.92%, 99.90% and 99.98%, respectively. Experimental results show that the algorithm proposed in this study has strong environmental adaptability and can effectively overcome the effects of complex background and external interference on pantograph detection, and has high practical application value.

Deep convolutional neural networks (DCNNs) are vulnerable to examples with small perturbations. Improving DCNNs’ robustness is of great significance to the safety-critical applications, such as autonomous driving and industry automation. Inspired by the principal way that human eyes recognize objects, i.e., largely relying on the shape features, this paper first designs four learnable edge detectors as layer kernels and proposes a binary edge feature branch (BEFB) to learn the binary edge features, which can be easily integrated into any popular backbone. The four edge detectors can learn the horizontal, vertical, positive diagonal, and negative diagonal edge features, respectively, and the branch is stacked by multiple Sobel layers (using learnable edge detectors as kernels) and one threshold layer. The binary edge features learned by the branch, concatenated with the texture features learned by the backbone, are fed into the fully connected layers for classification. We integrate the proposed branch into VGG16 and ResNet34, respectively, and conduct experiments on multiple datasets. Experimental results demonstrate the BEFB is lightweight and has no side effects on training. And the accuracy of the BEFB-integrated models is better than the original ones when facing white-box attacks and black-box attack. Besides, BEFB-integrated models equipped with the robustness enhancement techniques can achieve better classification accuracy compared to the original models. The work in this paper for the first time shows it is feasible to enhance the robustness of DCNNs through combining both shape-like features and texture features.

Gene expression programming (GEP) is one of the most prominent algorithms in function mining. In order to obtain a more accurate function model in configuration parameters-execution efficiency (CP-EE) of map-reduce job in the high-speed railway catenary monitoring system, this paper proposes a novel algorithm, called GEP based on multi-strategy (MS-GEP). Compared to traditional GEP, the proposed algorithm can escape premature convergence and jump out of local optimum. First, an adaptive mutation rate is designed according to the evolutionary generations, population diversity, and individual fitness values. A manual intervention strategy is then proposed to determine whether the algorithm enters the dilemma of local optimum based on the generations of population evolutionary stagnation. Finally, the average quality of the population is changed by randomly replacing individuals, and the ancestral population is traced to change the evolutionary direction. The experimental results on the benchmarks of function mining show that the proposed MS-GEP has better solution quality and higher population diversity than other GEP algorithms. Furthermore, the proposed MS-GEP has higher accuracy on the function model of CP-EE of high-speed railway catenary monitoring system than other commonly used algorithms in the field of function mining.

Adversarial training has become a primary method for enhancing the robustness of deep learning models. In recent years, fast adversarial training methods have gained widespread attention due to their lower computational cost. However, since fast adversarial training uses single-step adversarial attacks instead of multi-step attacks, the generated adversarial examples lack diversity, making models prone to catastrophic overfitting and loss of robustness. Existing methods to prevent catastrophic overfitting have certain shortcomings, such as poor robustness due to insufficient strength of generated adversarial examples, and low accuracy caused by excessive total perturbation. To address these issues, this paper proposes a fast adversarial training method—fast adversarial training with adaptive similarity step size (ATSS). In this method, random noise is first added to the input clean samples, and the model then calculates the gradient for each input sample. The perturbation step size for each sample is determined based on the similarity between the input noise and the gradient direction. Finally, adversarial examples are generated based on the step size and gradient for adversarial training. We conduct various adversarial attack tests on ResNet18 and VGG19 models using the CIFAR-10, CIFAR-100 and Tiny ImageNet datasets. The experimental results demonstrate that our method effectively avoids catastrophic overfitting. And compared to other fast adversarial training methods, ATSS achieves higher robustness accuracy and clean accuracy, with almost no additional training cost.