Fig 4 - uploaded by Xi Chen
Content may be subject to copyright.
Cross-similarity of all 50 tasks constructed from 5 scenes in the local railway scene dataset (left panel) and 5 types of binary classification tasks in the MiniImagenet dataset (right panel).
Source publication
Video surveillance is gaining increasing popularity to assist in railway intrusion detection in recent years. However, efficient and accurate intrusion detection remains a challenging issue due to: (a) limited sample number: only small sample size (or portion) of intrusive video frames is available; (b) high inter-scene dissimilarity: various railw...
Context in source publication
Context 1
... verify that the MiniImagenet dataset is better constructed, we randomly select 5 types of binary classification tasks from the MiniImagenet dataset and 5 scenes from the railway dataset, we then draw 10 tasks with K = 10 from each type/scene. The cross-similarity of all 50 tasks is shown in Figure 4, where one sees that the railway scene tasks has low inter-scene dissimilarity and higher intra-scene similarity when compared to those of the MiniImagenet dataset. ...
Similar publications
![Figure 1. Map of deserts in Saudi Arabia [4].](publication/362487848/figure/fig1/AS:11431281078769404@1660257077231/Map-of-deserts-in-Saudi-Arabia-4_Q320.jpg)
![Figure 2. Different methods of installing fences [21].](publication/362487848/figure/fig2/AS:11431281078769405@1660257077316/Different-methods-of-installing-fences-21_Q320.jpg)
![Figure 3. Hejaz railway station in the Arabian Desert [22].](publication/362487848/figure/fig3/AS:11431281078796511@1660257077433/Hejaz-railway-station-in-the-Arabian-Desert-22_Q320.jpg)
Sandstorms that cause shifting dunes are a huge technical challenge in the development of the railway network in Saudi Arabia, and are known as one of the most concerning atmospheric aspects. In this case, the weather or climate change makes railways more problematic and costly in Saudi Arabia compared to other countries. The purpose of this articl...
Citations
... Explainability is crucial for establishing trust in few-shot learning models, ensuring their safe deployment in real-world applications. For example, we can consider the context of a few-shot learning intrusion detection system for railway video surveillance, as examined in Gong et al. (2021). While the authors did not develop it using SNs, such an approach remains entirely feasible. ...
Machine learning models often struggle to generalize accurately when tested on new class distributions that were not present in their training data. This is a significant challenge for real-world applications that require quick adaptation without the need for retraining. To address this issue, few-shot learning frameworks, which includes models such as Siamese Networks, have been proposed. Siamese Networks learn similarity between pairs of records through a metric that can be easily extended to new, unseen classes. However, these systems lack interpretability, which can hinder their use in certain applications. To address this, we propose a data-agnostic method to explain the outcomes of Siamese Networks in the context of few-shot learning. Our explanation method is based on a post-hoc perturbation-based procedure that evaluates the contribution of individual input features to the final outcome. As such, it falls under the category of post-hoc explanation methods. We present two variants, one that considers each input feature independently, and another that evaluates the interplay between features. Additionally, we propose two perturbation procedures to evaluate feature contributions. Qualitative and quantitative results demonstrate that our method is able to identify highly discriminant intra-class and inter-class characteristics, as well as predictive behaviors that lead to misclassification by relying on incorrect features.
... The authors in [17] implement a Multi-Scale Residual Network for a dehazing model aimed at railway monitoring images; the solution is based on the fusing multiscale convolutional information; this solution improves the monitoring of railway infrastructure under harsh weather conditions. Finally, the authors in [18] present an enhanced few-shot learning video surveillance system for intrusion detection that addresses high and low similarity between surveillance scenes and low number of video frames. ...
... Authors in [35] implement Efficient Single Shot MultiBox Detector (SSD) algorithm for a real-time intrusion detection surveillance model. While the authors in [18] present an enhanced fewshot learning video surveillance system that addresses high and low similarity between surveillance scenes and low number of video frames. The study presented in [36] implements an improved YOLOv4 algorithm for detecting the intrusion of pedestrians in the railway tracks. ...
... [49]. Also, the combination of intelligent models with traditional video surveillance systems [18] [39], IoT devices [45] [47] [48] and smart sensing [41]. Some studies focus on railway scene analysis and prediction of pedestrian behaviour in accident-prone areas [54] [55], others in the safety in crossing areas through smart signalling and warnings [50] [57] [58]. ...
Since railways are considered among the most critical transportation infrastructures, significant research exists on the use of modern technology for their safety and security. In order to enable full autonomy of railway operations, it is not only important to guarantee safety from the control system perspective, but also to ensure that external threats of both intentional and natural origin are detected automatically. This literature review highlights the potential of Artificial Intelligence (AI) in physical infrastructure security and surveillance of autonomous railway operations. We survey AI applications to identify unauthorized access to restricted areas, as well as intelligent monitoring against natural disasters. We also discuss the use of AI to improve situational awareness, in order to analyze data from various sources to prevent incidents, provide real-time information to decision-makers, and support emergency response. Finally, we highlight the challenges to overcome, including data quality, privacy concerns, regulations, and other technical limitations.
... However, it is unavoidable for inducing network security issues in the smart grid due to different connections provided over power and communication resources [9]. Intrusion detection systems are considered mitigation and detection models and they involve random uncertainty-based communication noise [10]. General issues based on network security in the smart grid framework are summarized, and different classifications of smart grid attacks are offered [11]. ...
The IoT offered an enormous number of services with the help of multiple applications so it faces various security-related problems and also heavy malicious attacks. Initially, the IoT data are gathered from the standard dataset as Message Queuing Telemetry Transport (MQTT) set. Further, the collected data are undergone the pre-processing stage, which is accomplished by using data cleaning and data transformation. The resultant processed data is given into two models named (i) Autoencoder with Deep Belief Network (DBN), in which the optimal features are selected from Autoencoder with the aid of Modified Archimedes Optimization Algorithm (MAOA). Further, the optimal features are subjected to the AL-DBN model, where the first classified outcomes are obtained with the parameter optimization of MAOA. Similarly, (ii) Long Short-Term Memory (LSTM) with DBN, in this model, the optimal features are chosen from LSTM with the aid of MAOA. Consequently, the optimal features are subjected into the AL-DBN model, where the second classified outcomes are acquired. Finally, the average score is estimated by two outcomes to provide the final classified result. Thus, the findings reveal that the suggested system achieves outstanding results to detect the attack significantly.
... However, since the conventional means such as manual inspections are extremely low-efficient and labor-intensive, they have been gradually replaced by automatic monitoring devices with the development of monitoring sensors. In railway, the common automatic detection means include contact detection and non-contact detection (Gong et al., 2021;Cao et al., 2021). Among them, the contact methods identify the occurrence of intrusion by detecting the inductive signals of the monitoring devices that are caused by the intrusive objects. ...
In recent years, video surveillance has become increasingly popular in railway intrusion detection. However, it is still quite challenging to detect the intruded objects efficiently and accurately because: (a) The backgrounds of video frames generated by the fixed cameras are similar and only few intrusive frames are available, resulting in a lack of diversity among video frames, and further leading to over fitting of the detection models during training; (b) The intrusion of small targets or targets far from the location of camera exhibits sparsity relative to the wide monitoring view of the camera, which challenges the detection of such targets in a complex background; (c) The extreme imbalance between non-intrusive frames and intrusive frames, as well as a large number of unlabeled frames, hinder the effective training of the detection model and weaken its capacity of generalization. To tackle with the above issues, this article develops an effective intrusion detection method by combining low-rank and sparse decomposition (LRSD) and Semi-supervised Support Vector Domain Description (Semi-SVDD). Firstly, LRSD is used to decompose the monitored video into a background and a foreground. Then, based on the semantic segmentation method, we extract the mask of the track region in the decomposed background, which is used to mask the foreground. Next, by using both the labeled and unlabeled frames of the masked foreground, Semi-SVDD is established for the intrusion detection. Numerical results show that the removal of background interference and the combination of the labeled and unlabeled information help to improve the performance of the proposed method, and thus superior to benchmark methods.
... Tracking human movements is a challenging task applicable to a variety of fields, such as real-time exercise tracking [1][2][3][4][5], fall detection for the elderly [6][7][8][9][10][11], intrusion detection-based surveillance [12,13] in restricted areas such as railroad tracks and railway stations [14], etc. Wearable sensors [15][16][17][18] are popular for monitoring the health status in certain patients; however, these sensors only track certain vital signs and are insufficient for monitoring movement. Closed-Circuit Television (CCTV) and Internet Protocol (IP) cameras, on the other hand, can fully track movement, but are stationary and can seldom monitor remote areas, such as those where disasters have occurred. ...
Human movement tracking is useful in a variety of areas, such as search-and-rescue activities. CCTV and IP cameras are popular as front-end sensors for tracking human motion; however, they are stationary and have limited applicability in hard-to-reach places, such as those where disasters have occurred. Using a drone to discover a person is challenging and requires an innovative approach. In this paper, we aim to present the design and implementation of a human motion tracking method using a Tello EDU drone. The design methodology is carried out in four steps: (1) control panel design; (2) human motion tracking algorithm; (3) notification systems; and (4) communication and distance extension. Intensive experimental results show that the drone implemented by the proposed algorithm performs well in tracking a human at a distance of 2–10 m moving at a speed of 2 m/s. In an experimental field of the size 95×35m2, the drone tracked human motion throughout a whole day, with the best tracking results observed in the morning. The drone was controlled from a laptop using a Wi-Fi router with a maximum horizontal tracking distance of 84.30 m and maximum vertical distance of 13.40 m. The experiment showed an accuracy rate for human movement detection between 96.67 and 100%.
... The most important AI methods have been implemented in a number of the models, and those models also demonstrate intriguing expectations. These models include hereditary and fluffy calculation based models [10], grouping and k-closest neighbour based models [11], IDS utilising Backing Vector Machines (SVM) for preparation [12,13], and a lot of other similar models. These models additionally rely on signaturebased detection of intrusions in order to function well. ...
With the benefits of a wide monitoring range, easy installation and maintenance, as well as good remarkable observability, vision sensors are gradually being applied to engineering measurement and environmental perception research in rail transportation. A sliding pane-like track limit detection algorithm based on vision sensor is proposed for the track limit detection in the field of obstacle detection for fully automatic driverless trains in rail transit. The algorithm uses the two-stage inter-frame differential thresholding method to detect the motion of video frames, and uses grayscale distribution feature extraction and adaptive Sobel operator thresholding edge detection methods to improve the accuracy and integrity of the image scene recognition and edge detection. The track limit was extracted by the tack limit search module based on the sliding pane, and Kalman filter is used to improve the accuracy and robustness of the detection results. The experimental validation results show that the algorithm has good detection performance on track limit.
