In this paper, a Hankel matrix-based fully distributed algorithm is proposed to address a minimal-time deadbeat consensus prediction problem for discrete-time high-order multi-agent systems (MASs). Therein, each agent...
详细信息
Graph structured data is a fundamental form of data in fields such as urban planning, molecular biology, social network mining, and recommendation systems, and has important research value. However, due to the frequen...
详细信息
Graph structured data is a fundamental form of data in fields such as urban planning, molecular biology, social network mining, and recommendation systems, and has important research value. However, due to the frequent lack of labels in graph structure data, especially in the field of chemical biology, the cost of manual annotation is very expensive. Therefore, studying the use of unsupervised methods to learn the representation of graph structure data has certain practical significance. In recent years, unsupervised representation learning has made a series of breakthroughs in the fields of natural language processing and computer vision. Naturally, moving the most advanced unsupervised contrastive learning methods in these fields to graph structure data has become a hotspot in the direction of graph unsupervised learning. However, graph structured data is non Euclidean and does not exhibit shift invariance, resulting in structural differences from image data. However, when transferring unsupervised algorithms suitable for image data to graph structured data, there is still a problem where data augmentationt methods only focus on local information, which seriously affects the representation accuracy of graph contrastive learning methods. Therefore, this article proposes a line graph augmentation method that perturbs global information, mapping the points of the original graph to edges and edges to points, and generating augmented views by perturbing the entire graph through point edge linkage exchange, rather than simply modifying the local *** this basis, Line Graph Augmentation Contrastive Learning (LgaCL)was designed, LgaCL used the augmented views of the original and line graphs as a pair of positive examples to maximize the consistency of node representations between the two views. The experimental results show that LgaCL outperforms multiple baseline methods on multiple graph classification datasets, especially on the NCI1 dataset, with a classificat
The solid rocket engine is the main power unit of various missile weapons today, and it also has a wide range of applications in the aerospace field. However, in the process of welding and longterm operation, micro-in...
详细信息
The solid rocket engine is the main power unit of various missile weapons today, and it also has a wide range of applications in the aerospace field. However, in the process of welding and longterm operation, micro-invisible defects such as cracks and pores will inevitably occur inside the shell, which directly affects the performance of the solid rocket motor. This paper takes the solid engine shell X-ray film as the object. Most of the existing false-color methods are to quantize the 256-level grayscale in enhancing the image's details. In high-level grayscale image enhancement, the grayscale loss is serious and the adaptive adjustment is not satisfactory. An adaptive false color enhancement algorithm for high-level gray-scale (super 8-bit) images is proposed. In addition, for the adaptive enhancement problem, a new method of weld locating based on the Difference of Gaussian statistics is proposed. Moreover, a new method based on the fitness function of image quality evaluation is integrated with weighted information entropy, in which PSO is improved for fast optimization. To verify the effectiveness of the method in this paper, it is applied to the visual enhancement of the low-contrast and fuzzy defect information of the solid engine shell X-ray film. The experimental results show that our design obtains the best performance on the high-level X-ray film of a solid rocket engine shell, the algorithm in this paper has greatly improved various indicators compared with the existing algorithms. The high-level false-color enhancement method is used to process the Steel Pipe Weld X-ray in small sample dataset (3408), and apply the YOLOV4 mixed training to obtain the best detection results. The method in this paper can adaptively realize the false-color enhancement of high grayscale X-ray film, with a better visual effect, which can enhance the definition of the image detail of the welding defects of the solid engine shell, and display the high grayscale image on the or
The MO-YOLOv8 algorithm represents a significant enhancement of the YOLOv8 framework, integrating the Multi-Head Self-Attention (MHSA) mechanism to improve object detection performance, especially in complex and clutt...
详细信息
ISBN:
(数字)9798350374315
ISBN:
(纸本)9798350374322
The MO-YOLOv8 algorithm represents a significant enhancement of the YOLOv8 framework, integrating the Multi-Head Self-Attention (MHSA) mechanism to improve object detection performance, especially in complex and cluttered environments. By leveraging MHSA, MO-YOLOv8 can focus on multiple relevant parts of the input data simultaneously, capturing intricate details and contextual information often missed by traditional convolutional approaches. This integration allows the model to dynamically prioritize important regions in an image, enhancing its capability to detect and localize objects accurately. The MO-YOLOv8 algorithm processes input images through convolutional layers to extract preliminary feature maps, which are then fed into the MHSA module. Here, the data is split into multiple attention heads, each performing scaled dot-product attention. The resulting attention-weighted value matrices are concatenated and linearly projected to form the final feature representation. Experimental results show a significant improvement in mean Average Precision (mAP), along with a reduction in false positives and negatives compared to the baseline YOLOv8. The algorithm's robustness in handling occlusions and overlapping objects, coupled with its real-time processing capability, makes MO-YOLOv8 suitable for applications such as power transmission line monitoring, autonomous driving, and security surveillance. Overall, MO-YOLOv8 demonstrates superior detection performance through its advanced attention mechanism, offering a promising solution for various object detection challenges.
image Quality Assessment (IQA) has received unprecedented attention due to the extensive applications in benchmarking imageprocessingalgorithms and systems. Despite great progress in IQA, most previous frameworks ei...
详细信息
The video logging technology is applied to detect the general condition of the well. Due to the influence of water fog, the contrast of the video logging image is decreased and the color contrast is distorted. The app...
详细信息
ISBN:
(纸本)9781665447300
The video logging technology is applied to detect the general condition of the well. Due to the influence of water fog, the contrast of the video logging image is decreased and the color contrast is distorted. The application and research of the dark channel dehazing algorithm in the video logging image enhancement system is proposed. In this paper, the dark channel defogging algorithm and Retinex defogging algorithm are compared and analyzed respectively, the realization process and application of the two algorithms are described briefly, and the advantages and disadvantages of the two algorithms applied in video logging image enhancement are compared.
image encryption is an important tool for encrypted communication. Pixel dots can carry a lot of information and are more confusing than text encryption, which cannot be recognized with the naked eye and cannot be eas...
详细信息
ISBN:
(纸本)9781665464697
image encryption is an important tool for encrypted communication. Pixel dots can carry a lot of information and are more confusing than text encryption, which cannot be recognized with the naked eye and cannot be easily analyzed for encryption laws, so it has an important position in modern encryption technology. We have improved the traditional image encryption technology by using symmetric and asymmetric encryption techniques and adding the clock parameter as a factor to participate in the synthesis to achieve the automatic destruction function. This algorithm has a much higher encryption and decryption speed than conventional encryption algorithms due to the operation at the array level and the use of xor encryption on bit operations. After an experiment of encryption and decryption of images, we proved that this method has a promising application for commercial and military encrypted communications.
Multiple object tracking is one of the critical directions in computer vision research. In the application of vision-based tracking methods, cameras are sometimes installed far from the targets to obtain a global view...
详细信息
ISBN:
(纸本)9781665464697
Multiple object tracking is one of the critical directions in computer vision research. In the application of vision-based tracking methods, cameras are sometimes installed far from the targets to obtain a global view. There would be a large number of targets in the videos with relatively low resolution, which increases the difficulty of visual tracking. Applying existing tracking methods directly in such low-resolution scenarios will result in low recall and a large number of discontinued trajectory fragments, due to the instability of the target detection results. To alleviate the tracking performance degradation in low-resolution scenarios, a multiple object tracking method based on tracking compensation (MOT-TC) is proposed in this paper. A detector is applied to produce the candidate bounding boxes of the targets in the current frame. Then trajectories from previous frames are used to predict their states in the current frame. An assignment method is adopted to match the candidate bounding boxes to the predicted states. For the unmatched trajectories in the current frame, a single object tracking method for compensation is used to provide the target positions, which can increase the recall and reduce trajectory fragments. Meanwhile, a strategy based on the response map of single object tracking is designed to evaluate the tracking performance. Extensive experiments on low-resolution videos have shown that the proposed method outperforms the baseline and other state-of-the-art methods by a large margin.
暂无评论