A New Method for Generating Simulated Anomaly Images Based on Perlin Noise

Fast display of realistic dynamic cloud scene is a challenging task for researchers in computer graphics. This paper describes a novel method of simulating animations of cloud. We first model the motion of cloud based on the physical theory of cloud formation. Then, the physical model of cloud is solved in discrete sparse grids. Thus, we can get approximate motion of cloud. Next, we advect texture image using the calculated velocity field to add the local detail of dynamic cloud scene. Here the texture image we used is Perlin noise map. The color of arbitrary point in the simulation space is got by calculating the product of the density texture and the regenerated Perlin noise texture. Finally, by changing different noise maps, realistic dynamic cloud scenes in different light environment are generated at high rendering rates.

Software breast phantoms are central to the optimization of breast imaging, where in many cases the use of real images would be inefficient - or impossible. Establishing the realism of such phantoms is critical. For this study, patches of simulated breast tissue with different composition - fatty, scattered, heterogenous and dense tissue - were generated using a method based on Perlin noise. The composition of the patches is controlled by numerical parameters derived from input by radiologists and medical physicists with experience of breast imaging. Separate Perlin noise-based methods were used to simulate skin pores, high-frequency noise (representing quantum and electronic noise) and ligaments and vascular structures. In a forced choice reading study, the realism of the simulated tissue patches compared to patches from real mammograms was determined. Patches of 200-500 pixels were extracted from radiolucent, linear, nodular or homogenous (10 per category) mammograms randomly selected from a previously acquired dataset. Eighteen simulated patches in the same size range were added. Four readers, two radiologists and two medical physicists were shown the images in random order and asked to rate them as real or simulated. All readers accepted a substantial fraction of simulated images as real, ranging from 22% to 72%. Only two readers showed a significant difference in the number of images rated real in the real and simulated groups, 22% vs 73% (P=.0003) and 33% vs 63% (P=.04), respectively. These results suggest that the method employed can create images that are almost indistinguishable from patches of real mammograms. (Less)

With the rapid development of autonomous driving technology, traditional object detection algorithms often face challenges such as the trade-off between model parameters and recognition accuracy, insufficient detection accuracy, and slow detection speed when dealing with vehicle and pedestrian detection tasks in complex scenarios. To address these challenges, we propose a novel lightweight real-time driving scene detection network, FEEP-YOLO. Based on the YOLOv8 algorithm, our network not only maintains high detection accuracy but also achieves faster training speed and fewer model parameters. FEEP-YOLO employs a lightweight network, Faster Net, to significantly reduce model parameters, thereby accelerating detection speed and better meeting real-time requirements. To further improve detection accuracy, we introduce an EMA attention module into the network and optimize the bounding box loss function, adopting the EIOU loss function to enhance target features and reduce background interference. Additionally, to detect small objects more accurately, we add feature layers rich in semantic information to the detection end. Experimental results demonstrate that FEEP-YOLO outperforms YOLOv8 in terms of both detection speed and accuracy on the KITTI and SODA datasets, while significantly reducing the number of model parameters and model size. FEEP-YOLO achieves high detection speed while improving detection accuracy and reducing computational resource consumption, meeting the practical application requirements under limited computing power, memory space and power consumption. The key contribution of FEEP-YOLO lies in its ability to balance accuracy speed, and efficiency, making it a robust solution for real-time autonomous driving applications in complex environments.

Efficient monitoring of concrete pouring operations is critical for ensuring compliance with construction regulations and maintaining structural quality. However, traditional monitoring methods face limitations such as overlapping objects, environmental similarities, and detection errors caused by ambiguous boundaries. This study proposes an Ambient Detection-based Monitoring Framework that enhances object detection by incorporating contextual relationships between objects in complex construction environments. The framework employs the You Only Look Once version 11 (YOLOv11) algorithm, addressing issues of boundary ambiguity and misrecognition through relational analysis. Key components including Distance Relationship (DR), Attribute Relationship (AR), and Spatial Relationship (SR) allow the system to quantitatively evaluate contextual associations and improve detection accuracy. Experimental validation using 232 test images demonstrated a 12.07% improvement in detection accuracy and a 71% reduction in false positives compared with baseline YOLOv11. By automating the monitoring process, the proposed framework not only improves efficiency but also enhances construction quality, demonstrating its adaptability to diverse construction scenarios.

Images captured using unmanned aerial vehicles (UAVs) often exhibit dense target distribution and indistinct features, which leads to the issues of missed detection and false detection in target detection tasks. To address these problems, an improved method for small target detection called YOLOv5s is proposed to enhance the detection accuracy for small targets such as apple fruits. By applying improvements to the RFA module, DFP module, and Soft-NMS algorithm, as well as integrating these three modules together, accurate detection of small targets in images can be achieved. Experimental results demonstrate that the integrated, improved model achieved a significant improvement in detection accuracy, with precision, recall, and mAP increasing by 3.6%, 6.8%, and 6.1%, respectively. Furthermore, the improved method shows a faster convergence speed and lower loss value during the training process, resulting in higher recognition accuracy. The results of this study indicate that the proposed improved method exhibits a good performance in apple fruit detection tasks involving UAV imagery, which is of great significance for fruit yield estimation. The research findings demonstrate the effectiveness and feasibility of the improved method in addressing small target detection tasks, such as apple fruit detection.

To develop and validate algorithms for extraction and insertion of three-dimensional (3D) profiles of tumor masses in cone beam breast CT (CBBCT) images to create simulated abnormality for evaluation with observer performance study. A bench-top experimental CBBCT scanner was constructed and used to image mastectomy breast specimens with IRB approval. 5 sets of CBBCT images with confirmed tumor masses and 2 sets of normal CBBCT images were selected and used for this study. All CBBCT images were first corrected for cupping artifacts. The corrected images were then processed to reduce their noise levels and form the denoised images. The corrected and denoised CBBCT images for normal breasts were then segmented into adipose and glandular voxels. The images for abnormal breasts were reviewed by mammographers with the help of clinical images and reports to delineate the tumor masses and form 3D tumor maps. Using these maps, the 3D tumor profiles were rescaled to average glandular signals around the insertion locations in normal images. The boundaries of the 3D rescaled tumor profiles were smoothed to avoid sharp edge. The resultant 3D tumor profiles were inserted onto cupping artifact corrected normal CBBCT images by replacing the signals in adipose voxels near the insertion locations. The cupping artifacts were then added back to generate the simulated abnormal CBBCT images. These images were then visually compared with actual abnormal images for their degree of realism. Based on visual comparison, the simulated abnormal CBBCT images showed no significant difference in realism from actual abnormal CBBCT images. We have successfully demonstrated the technique to extract 3D tumor profiles from abnormal CBBCT images and insert them onto normal CBBCT images to form simulated abnormal images for use in observer performance study. This work was supported in part by grants CA104759, CA13852 and CA124585 from the NIH-NCI, a research grant EB00117 from the NIH-NIBIB, and a subcontract from NIST-ATP.

The increasing complexity of enterprise accounting systems has created significant challenges in detecting fraudulent and anomalous transaction patterns. Traditional rule-based detection methods often fail to identify sophisticated fraud schemes that exploit network structures and temporal dependencies within transaction data. This paper proposes a graph neural network-based framework for detecting anomalous transaction patterns in enterprise accounting systems by modeling financial transactions as heterogeneous graphs. We demonstrate that graph-based representations can capture complex relationships between accounts, merchants, and transactions that are invisible to conventional point-based anomaly detection methods. Our approach leverages multi-level graph analysis to detect anomalies at transaction, account, and network levels, enabling comprehensive fraud detection across different organizational scales. The methodology incorporates temporal dynamics, heterogeneous node types, and structural features to identify suspicious patterns indicative of fraudulent activities. Experimental results demonstrate the superiority of graph neural networks over traditional machine learning approaches in detecting complex fraud schemes including collusion networks, money laundering chains, and coordinated attack patterns. The proposed framework achieves significant improvements in detection accuracy while maintaining interpretability through graph visualization and attention mechanisms that highlight suspicious substructures.

Underwater Wireless Sensor Networks (UWSNs) are a pivotal technology for ocean monitoring, underwater surveillance, and environmental sensing. However, these networks are vulnerable to numerous faults caused by adverse aquatic environments, scarce node resources, and unreliable acoustic communication. This paper presents an adaptive fault detection and tolerance mechanism specially designed for UWSNs. The proposed system integrates a hybrid fault detection approach, combining explicit and implicit approaches with a light-weight tolerance strategy leveraging multi-path rerouting and load-aware decision-making. A dynamic thresholding mechanism is used to enhance fault detection sensitivity and accuracy while ensuring energy efficiency. The mechanism is implemented and evaluated using MATLAB simulations with different fault densities and node behaviors. The findings show significant improvement in detection accuracy, less false alarm, lower latency, and better network lifetime when compared with traditional fault management schemes. This research is a step toward constructing fault-tolerant UWSNs that can maintain operations under the rough and uncertain nature of underwater environments.

Malicious social bots pose a serious threat to social network security by spreading false information and guiding bad opinions in social networks. The singularity and scarcity of single organization data and the high cost of labeling social bots have given rise to the construction of federated models that combine federated learning with social bot detection. In this paper, we first combine the federated learning framework with the Relational Graph Convolutional Neural Network (RGCN) model to achieve federated social bot detection. A class-level cross entropy loss function is applied in the local model training to mitigate the effects of the class imbalance problem in local data. To address the data heterogeneity issue from multiple participants, we optimize the classical federated learning algorithm by applying knowledge distillation methods. Specifically, we adjust the client-side and server-side models separately: training a global generator to generate pseudo-samples based on the local data distribution knowledge to correct the optimization direction of client-side classification models, and integrating client-side classification models' knowledge on the server side to guide the training of the global classification model. We conduct extensive experiments on widely used datasets, and the results demonstrate the effectiveness of our approach in social bot detection in heterogeneous data scenarios. Compared to baseline methods, our approach achieves a nearly 3-10% improvement in detection accuracy when the data heterogeneity is larger. Additionally, our method achieves the specified accuracy with minimal communication rounds.

This paper demonstrates the improved performance of metamaterial based SPR biosensor in terms of sensitivity and detection accuracy over conventional SPR and graphene based SPR biosensor. The performance of the proposed SPR biosensor has been evaluated for three metals Au, Ag and Al. Angular interrogation mechanism has been utilized for plotting the reflectance curves. The sensitivity for the proposed metamaterial based SPR biosensor for Au metal without using silicon is 1.88 and 1.86 times the sensitivity of the conventional SPR and graphene based SPR biosensor with little decrement in detection accuracy. Silicon has been included in the proposed SPR biosensor for improvement in detection accuracy with little compromise in sensitivity. Thus proposed metamaterial based SPR biosensor can be used for detection of biomolecules with improved sensitivity and greater accuracy.

Software was developed in Python for the digital processing of object images before their detection using the You Only Look Once (YOLO) artificial neural network. Detection was carried out on object images, specifically vehicles and other road users. The preprocessing of images was performed through histogram equalization and local contrast enhancement. The original images were read from video cameras or graphic files. Image detection was performed using the medium-sized YOLO convolutional neural network model, YOLOv8m. As a result of detection, the studied objects, particularly cars, were highlighted with rectangular bounding boxes. To evaluate the accuracy of object detection, several metrics (parameters) were used: Precision, Recall, Intersection over Union (IoU), F1, and Average Precision (AP). The YOLO network was pre-trained on the Common Objects in Context dataset. The YOLO neural network returns the object recognition confidence, the coordinates of its center (x, y) on the image, and its width and height (w, h). The detected objects are marked with rectangular bounding boxes. The positioning of objects in the images was manually performed using the Roboflow tool, which allows for accurate spatial positioning of each object. The study of the impact of image preprocessing on the accuracy of object detection by the YOLO network was conducted using a typical traffic scene image. After histogram equalization, the histogram becomes more symmetrical and uniform, improving the visual quality of both dark and bright areas. This led to improved object detection accuracy across almost all metrics. The increase in detection accuracy by the IoU parameter is 0.169. The summary improvement in detection accuracy by the IoU, F1, AP parameters is 0.502. It was shown that after image preprocessing, detection accuracy increased, and all vehicles were correctly detected. The developed software can be used in various computer vision systems as well as in Internet of Things systems for remote monitoring and control.

YOLOv5s-road: Road surface defect detection under engineering environments based on CNN-transformer and adaptively spatial feature fusion

A rich model-based motion vector (MV) steganalysis benefiting from both temporal and spatial correlations of MVs is proposed in this paper. The proposed steganalysis method has a substantially superior detection accuracy than the previous methods, even the targeted ones. The improvement in detection accuracy lies in several novel approaches introduced in this paper. First, it is shown that there is a strong correlation, not only spatially but also temporally, among neighbouring MVs for longer distances. Therefore, temporal MV dependency alongside the spatial dependency is utilized for rigorous MV steganalysis. Second, unlike the filters previously used, which were heuristically designed against a specific MV steganography, a diverse set of many filters, which can capture aberrations introduced by various MV steganography methods is used. The variety and also the number of the filter kernels are substantially more than that of used in the previous ones. Besides that, filters up to fifth order are employed whereas the previous methods use at most second order filters. As a result of these, the proposed system captures various decorrelations in a wide spatio-temporal range and provides a better cover model. The proposed method is tested against the most prominent MV steganalysis and steganography methods. To the best knowledge of the authors, the experiments section has the most comprehensive tests in MV steganalysis field, including five stego and seven steganalysis methods. Test results show that the proposed method yields around 20% detection accuracy increase in low payloads and 5% in higher payloads.

Human-observer receiver-operating characteristic (ROC) results with clinical patient studies indicate that ordered-subset expectation-maximization (OSEM) reconstructions with a combination of corrections for attenuation, scatter, and distance-dependent resolution (DDR) significantly improves perfusion defect detection accuracy over filtered back-projection (FBP) images with no compensation. A recent Monte Carlo study has also shown that for scatter correction (SC) in particular, selection of an appropriate strategy can impact cardiac uniformity. The spatial domain based effective source scatter estimation (ESSE) technique was shown to provide more robust performance in improving cardiac uniformity than the triple-energy window (TEW) method. In this work, we investigate if further improvements in observer performance could be obtained by using the ESSE method of SC instead of the currently employed TEW SC, in combination with attenuation correction (AC) and resolution compensation (RC). We evaluated the effectiveness of the ESSE method to provide additional improvements in image quality objectively by using human-observer ROC studies on clinically acquired patient acquisitions. Results indicate that both ESSE and TEW SC in combination with AC and RC provide significantly higher detection accuracy than FBP with no compensation for the overall detection of coronary artery disease (CAD) as well as in localizing perfusion defects in the left anterior descending (LAD) and left circumflex (LCx) territories. Comparing, the two implementations of SC evaluated in this study, we note that the ESSE method resulted in larger aggregate areas under the ROC curve (A/sub z/) in each case. However, a statistically significant improvement over TEW correction was only observed in the LAD territory. This indicates that SC implemented with the ESSE and TEW methods were close in terms of their improvement in detection accuracy for perfusion defects in the clinical images of this investigation, with the ESSE method arguably being slightly better. However, the clinical implementation of ESSE will be hampered by its longer computing time.

The rapid advancement of cyber threats necessitates the development of more sophisticated intrusion detection systems (IDS). Traditional IDS methods of- ten fail to keep up with the evolving nature of these threats. This paper explores the use of adaptive ma- chine learning algorithms to enhance IDS performance in cybersecurity. By utilizing adaptive learning techniques, we aim to create a system that dynamically adjusts to new and emerging threats in real-time. We evaluate several machine learning algorithms, including decision trees, support vector machines, neural networks, and ensemble methods, to determine their effectiveness in intrusion detection. Our pro- posed framework integrates these algorithms into an adaptive system that continuously improves its detection accuracy and reduces false positives by learning from ongoing threat patterns. Experiments conducted on benchmark datasets re- veal that our adaptive IDS framework outperforms traditional methods, demonstrating significant improvements in detection accuracy and response time. This research highlights the potential of adaptive ma- chine learning algorithms to provide a more robust and intelligent defence against cyber threats, paving the way for next-generation intrusion detection systems.