Abstract In order to improve the robustness of the image watermarking and reduce the influence of the overall image quality of the carrier when watermark is embedded on the carrier, an edge detection method combined with holographic watermarking was proposed to realize the hiding of watermark information. The carrier image is processed by edge detection, the watermark embedding area is segmented in the carrier image, and the suitable embedding area is filtered according to the size of the watermark; the original watermark is processed by a four-step phase shift algorithm to generate a holographic watermark, and the holographic watermark is embedded in the selection The embedding area is completed to embed the watermark, and the carrier image which the watermark is embedded in is carrier image with watermark. The watermark extraction method is to perform the same edge detection segmentation and region screening on the watermark carrier image, find the embedded area of the watermark, extract the holographic watermark, perform holographic inverse transformation to obtain the original watermark image, and complete the watermark recovery. The algorithm which is based on edge detection and holographic watermarking reduces the influence of embedded watermark on the carrier image, increases the flexibility and concealment of embedded position, and improves the robustness of the image when it is under geometric attacks.
In order to improve the robustness and invisibility of watermark, a grayscale computing holographic image watermarking algorithm based on discrete cosine transform has been proposed. First, the watermark image is calculated to generate a hologram to improve the security of the watermark. The carrier image is divided into 4×4 blocks, and each block is separately subjected to discrete cosine transform, and the low-frequency coefficients of each block are extracted to form a new matrix.Then, the watermark image is subjected to discrete cosine transform, and embed it into the carrier image according to prescript embedding rules, thereby completing the embedding process of the watermark.The extraction process is the inverse of the embedding process. The obtained image is decomposed into 4×4 blocks, and each block is subjected to discrete cosine transform, and extracted according to the inverse law of the embedding.Through test, the experimental results of the method show that the proposed algorithm has strong robustness in image clipping, rotation, translation and Gaussian noise attack. It ensures that the watermark is hidden while resisting a certain degree of attack to meet the needs of copyright identification.
Abstract With the rapid development of power systems, the switchgear, which is widely used, plays an increasingly important role. Partial discharge is one of the main causes of insulation failure of the switchgear, which is a serious threat to personal and equipment safety. In this paper, the partial discharge of the switchgear is detected based on the ultrasonic method. According to the characteristics of ultrasonic waves generated by partial discharge, a partial discharge detection method for the switchgear based on the characteristics of ultrasonic wave signal is designed. The wavelet threshold denoising algorithm is used to denoise the ultrasonic signal and draw the partial discharge phase distribution map by using the characteristics of low amplitude and wide frequency distribution of the interference signal. Finally, the improved support vector machine algorithm is used to identify and diagnose four types of discharge based on 280 groups of experimental data. By adjusting the kernel function and optimizing the support vector machine algorithm, the final recognition accuracy is more than 90%.
Abstract In order to improve the robustness and invisibility of watermarking, a digital holographic watermarking algorithm based on DWT-DCT is proposed. First, the watermark image is generated into a digital hologram to improve the security and non-tearability of the watermark, and the I1 image is selected as the watermark image to be embedded. Perform discrete wavelet transform on the original carrier image to extract its low-frequency coefficients; secondly, the low-frequency coefficients are subjected to DCT transformation, and finally the watermark information is embedded in the high-frequency coefficient matrix after DCT, thereby completing the watermark information embedding. This algorithm combines the transformation characteristics of DWT and DCT, makes up for each other’s shortcomings, and makes digital watermarking more robust and invisible. In order to verify the anti-attack ability of this algorithm, different attack tests were performed on the watermark image. The results show that this algorithm is robust against simple linear attacks and noise attacks.
Vibration fatigue failure of small branch pipes poses a great threat to the safe operation of nuclear power plants. However, the transient and wide-band vibration problems are not adequately considered in ASME and RCC-M code, resulting in repeated fatigue failures. In addition, the current research mainly focuses on vibration test methods and fatigue analysis methods, neglecting the study of pipeline vibration characteristics. Therefore, innovative approaches were essential for effectively managing complex dynamic loads. In this study, an innovative approach combining backpropagation artificial neural networks (BP-ANN) and non-dominated sorting Genetic Algorithm II (NSGA-II) was proposed to optimize the vibration of these pipes. The goal was mitigating vibration-induced failures by enhancing operational stability. The methodology progressed through several key stages. Firstly, BP-ANN was utilized for regression analysis, correlating pipe characteristics to vibration effects. Through regression analysis, the complex interrelationships governing the pipes' dynamic behavior was revealed. Subsequently, based on the regression model, NSGA-II was used to derive an optimal combination of design parameters to minimize the vibration response. The proposed technique was validated on an L-shaped cantilevered pipe via finite element simulations and physical experiments. The analysis case shows that the BP-ANN model demonstrated excellent accuracy in predicting vibration responses. Meanwhile, NSGA-II successfully revealed the trade-offs between conflicting objectives, generating a Pareto-optimal set balancing stability under different excitation directions. This study highlights the potential of machine learning methods for dynamic optimization of small branch pipes in nuclear power plants, and verifies the accuracy and effectiveness of the method through experiments. The research results provide a new idea for small branch pipe design and vibration control of nuclear power plant, which contributes to enhancing the safety and reliability of small branch pipes.
The "pre-train, prompt-tuning'' paradigm has demonstrated impressive performance for tuning pre-trained heterogeneous graph neural networks (HGNNs) by mitigating the gap between pre-trained models and downstream tasks. However, most prompt-tuning-based works may face at least two limitations: (i) the model may be insufficient to fit the graph structures well as they are generally ignored in the prompt-tuning stage, increasing the training error to decrease the generalization ability; and (ii) the model may suffer from the limited labeled data during the prompt-tuning stage, leading to a large generalization gap between the training error and the test error to further affect the model generalization. To alleviate the above limitations, we first derive the generalization error bound for existing prompt-tuning-based methods, and then propose a unified framework that combines two new adapters with potential labeled data extension to improve the generalization of pre-trained HGNN models. Specifically, we design dual structure-aware adapters to adaptively fit task-related homogeneous and heterogeneous structural information. We further design a label-propagated contrastive loss and two self-supervised losses to optimize dual adapters and incorporate unlabeled nodes as potential labeled data. Theoretical analysis indicates that the proposed method achieves a lower generalization error bound than existing methods, thus obtaining superior generalization ability. Comprehensive experiments demonstrate the effectiveness and generalization of the proposed method on different downstream tasks.
Although people are impressed by the content generation skills of large language models, the use of LLMs, such as ChatGPT, is limited by the domain grounding of the content. The correctness and groundedness of the generated content need to be based on a verified context, such as results from Retrieval-Augmented Generation (RAG). One important issue when adapting LLMs to a customized domain is that the generated responses are often incomplete, or the additions are not verified and may even be hallucinated. Prior studies on hallucination detection have focused on evaluation metrics, which are not easily adaptable to dynamic domains and can be vulnerable to attacks like jail-breaking. In this work, we propose 1) a post-processing algorithm that leverages knowledge triplets in RAG context to correct hallucinations and 2) a dual-decoder model that fuses RAG context to guide the generation process.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)
'/%3e%3cuse xlink:href='%23a' transform='rotate(72 0 0)'/%3e%3cuse xlink:href='%23a' transform='rotate(-72 0 0)'/%3e%3cuse xlink:href='%23a' transform='scale(-1 1) rotate(72)'/%3e%3c/g%3e%3c/defs%3e%3cpath fill='%23012169' d='M0 0h1200v600H0z'/%3e%3cpath stroke='%23FFF' stroke-width='60' d='m0 0 600 300M0 300 600 0' clip-path='url(%23b)'/%3e%3cpath stroke='%23C8102E' stroke-width='40' d='m0 0 600 300M0 300 600 0' clip-path='url(%23c)'/%3e%3cpath stroke='%23FFF' stroke-width='100' d='M300 0v300M0 150h600' clip-path='url(%23b)'/%3e%3cpath stroke='%23C8102E' stroke-width='60' d='M300 0v300M0 150h600' clip-path='url(%23b)'/%3e%3cuse xlink:href='%23d' fill='%23FFF' transform='matrix(45.4 0 0 45.4 900 120)'/%3e%3cuse xlink:href='%23d' fill='%23C8102E' transform='matrix(30 0 0 30 900 120)'/%3e%3cg transform='rotate(82 900 240)'%3e%3cuse xlink:href='%23d' fill='%23FFF' transform='rotate(-82 519.022 -457.666) scale(40.4)'/%3e%3cuse xlink:href='%23d' fill='%23C8102E' transform='rotate(-82 519.022 -457.666) scale(25)'/%3e%3c/g%3e%3cg transform='rotate(82 900 240)'%3e%3cuse xlink:href='%23d' fill='%23FFF' transform='rotate(-82 668.57 -327.666) scale(45.4)'/%3e%3cuse xlink:href='%23d' fill='%23C8102E' transform='rotate(-82 668.57 -327.666) scale(30)'/%3e%3c/g%3e%3cuse xlink:href='%23d' fill='%23FFF' transform='matrix(50.4 0 0 50.4 900 480)'/%3e%3cuse xlink:href='%23d' fill='%23C8102E' transform='matrix(35 0 0 35 900 480)'/%3e%3c/svg%3e)