In relation extraction tasks, distant supervision is a common method for automatic data labeling. However, this method will introduce a large amount of noisy data, which affects the performance of the model. In order to solve the problem of noisy data, a relation extraction method based on negative training and transfer learning was proposed. Firstly, a noisy data recognition model was trained through negative training method. Then, the noisy data were filtered and relabeled according to the predicted probability value of the sample, Finally, a transfer learning method was used to solve the domain shift problem existing in distant supervision tasks, and the precision and recall of the model were further improved. Based on Thangka culture, a relation extraction dataset with national characteristics was constructed. Experimental results show that the F1 score of the proposed method reaches 91.67%, which is 3.95 percentage points higher than that of SENT (Sentence level distant relation Extraction via Negative Training) method, and is much higher than those of the relation extraction methods based on BERT (Bidirectional Encoder Representations from Transformers), BiLSTM+ATT(Bi-directional Long Short-Term Memory and Attention), and PCNN (Piecewise Convolutional Neural Network).
The existing endoscopic image highlight removal algorithms often have some problems such as unreasonable removal structure and color distortion, which leads to the wrong results of the focus recognition algorithms and image enhancement algorithms. In order to solve the above problems, in the aspect of highlight localization, a method based on the combination of growth in dark region and Scharr filtering was proposed to locate relative highlight; in the aspect of highlight filling, an improved Crinminisi algorithm was proposed. Firstly, through the statistics on a huge amount of data, the search scope was limited and the filling efficiency was increased. Secondly, the statistical scope of priority was improved to avoid repeated meaningless calculations. Finally, the reasonable reconstruction of texture was performed according to the adaptive templates of different regions. Experiments were carried out on endoscopic image dataset of different human tissues, compared with the dichromatic reflection model based method, the Robust Principle Component Analysis (RPCA) method, the thermal diffusion method and the original Criminisi algorithm, the Natural Image Quality Evaluator (NIQE) value of the proposed algorithm was the lowest. Compared with the RPCA method, the thermal diffusion method and the original Crimnisi algorithm, the running time of the proposed algorithm was the lowest. Experimental results show that the proposed algorithm not only has better objective image indicators than other algorithms, but also has a nearly 100-fold improvement in efficiency compared to the original Criminisi algorithm.
The complexity of pedestrian interaction is a challenge for pedestrian trajectory prediction, and the existing algorithms are difficult to capture meaningful interaction information between pedestrians, which cannot intuitively model the interaction between pedestrians. To address this problem, a multi-head soft attention graph convolutional network was proposed. Firstly, a Multi-head Soft ATTention (MS ATT) combined with involution network was used to extract sparse spatial adjacency matrix and sparse temporal adjacency matrix from spatial and temporal graph inputs respectively to generate sparse spatial directed graph and sparse temporal directed graph. Then, a Graph Convolutional Network (GCN) was used to learn interaction and motion trend features from sparse spatial and sparse temporal directed graphs. Finally, the learned trajectory features were input into a Temporal Convolutional Network (TCN) to predict double Gaussian distribution parameters, thereby generating the predicted pedestrian trajectories. Experiments on Eidgenossische Technische Hochschule (ETH) and University of CYprus (UCY) datasets show that, compared with Space-time sOcial relationship pooling pedestrian trajectory Prediction Model (SOPM), the proposed algorithm reduces the Average Displacement Error (ADE) by 2.78%, and compared to Sparse Graph Convolution Network (SGCN), the proposed algorithm reduces the Final Displacement Error (FDE) by 16.92%.
With the rapid development of e-commerce and the popularity of the Internet, it is more convenient to exchange and return goods. Therefore, the customers’ demands for goods show the characteristics of timeliness, variety, small batch, exchanging and returning. Aiming at Location-Routing Problem with Simultaneous Pickup and Delivery (LRPSPD) with capacity and considering the characteristics of customers’ diversified demands, a mathematical model of LRPSPD & Time Window (LRPSPDTW) was established. Improved FireWorks Algorithm (IFWA) was used to solve the model, and the corresponding neighborhood operations were carried out for the fireworks explosion and mutation. The performance of the fireworks algorithm was evaluated with some benchmark LRPSPD examples. The correctness and effectiveness of the proposed model and algorithm were verified by a large number of numerical experiments. Experimental results show that compared with Branch and Cut algorithm (B&C), the average error between the result of IFWA and the standard solution is reduced by 0.33 percentage points. The proposed algorithm shortens the time to find the optimal solution, and provides a new way of thinking for solving location-routing problems.
The virtual try-on technologies based on image synthesis mask strategy can better retain details of the clothing when the warped clothing is fused with the human body. However, because the position and structure of the human body and the clothing are difficult to align during the try-on process, the try-on result is likely to produce severe occlusion, affecting visual effect. In order to solve the occlusion in the try-on process, a U-Net based generator was proposed. In the generator, a cascaded spatial attention module and a channel attention module were added to the U-Net decoder, thereby achieving the cross-domain fusion between local features of warped clothes and global features of the human body. Formally, first, by predicting the Thin Plate Spline (TPS) conversion using the convolutional network, the clothing was distorted according to the target human body pose. Then, the dressed-on person representation information and the warped clothing were input into the proposed generator, and the mask image of the corresponding clothing area was obtained to render the intermediate result. Finally, the strategy of mask synthesis was used to synthesize the warped clothing with the intermediate result through mask processing to obtain the final try-on result. Experimental results show that the proposed method can not only reduce occlusion, but also enhance image details. Compared with Characteristic-Preserving Virtual Try-On Network (CP-VTON) method, the proposed method has the generated image with the average Peak Signal-to-Noise Ratio (PSNR) increased by 10.47%, the average Fréchet Inception Distance (FID) decreased by 47.28%, and the average Structural SIMilarity (SSIM) increased by 4.16%.
Aiming at the problem of poor visual effects of block-based compressed sensing reconstructed images at low sampling rates, a compressed sensing image reconstruction method that fused Spatial Location and Structure Information (SLSI) was proposed. Firstly, observations were linearly mapped to obtain initial estimated values of image blocks. Then, based on block grouping reconstruction branch and whole image reconstruction branch, the spatial location information and structure information of the image were extracted, enhanced and fused. Finally, weighted strategy was used to fuse the outputs of the two branches to obtain final reconstructed whole image. In the block grouping reconstruction branch, reconstruction resources were allocated according to the data characteristics of the image blocks. In the whole image reconstruction branch, information exchange between adjacent image block pixels was mainly carried out through bilateral filtering and structural feature interaction module. Experimental results show that compared with compressed sensing reconstruction methods based on non-iterative Reconstruction Network (ReconNet) and Multi-scale Reconstruction neural Network with Non-Local constraint (NL-MRN), due to the combination of the image prior with strong autocorrelation between pixels, when sampling rate is 0.05, the average Peak Signal-to-Noise Ratio (PSNR) and Structural SIMilarity index (SSIM) of the proposed method on the test image data commonly used in the compressed sensing field increase 2.617 5 dB and 0.105 3 respectively, and the visual effects of reconstructed images are better.
Considering different blockchains being isolated and the data interaction and sharing difficulties in the current rapid development process of blockchain technology, a cross-chain mechanism based on Spark blockchain was proposed. Firstly, common cross-chain technologies and current mainstream cross-chain projects were analyzed, the implementation principles of different technologies and projects were studied, and their differences, advantages and disadvantages were summarized. Then, using the blockchain architecture maned main-sub blockchain mode, the key core components such as smart contract component, transaction verification component, transaction timeout component were designed, and the four stages of cross-chain process were elaborated in detail, including transaction initiation, transaction routing, transaction verification and transaction confirmation. Finally, the feasible experiments were designed for performance test and security test, and the security was analyzed. Experimental results show that Spark blockchain has significant advantages compared to other blockchains in terms of transaction delay, throughput and spike testing. Besides, when the proportion of malicious nodes is low, the success rate of cross-chain transactions is 100%, and different sub chains can conduct cross-chain transactions safely and stably. This mechanism solves the problem of data interaction and sharing between blockchains, and provides technical reference for the design of Spark blockchain application scenarios in the next step.
Focused on the issue that different blockchains are independent from and difficult to communicate with each other, a new type of permissioned public blockchain architecture of "main chain + sub chain" was proposed. Firstly, based on the existing algorithms such as Delegated Proof Of Stake (DPOS), Verifiable Random Function (VRF) and Practical Byzantine Fault Tolerance (PBFT), an innovative two-layer consensus algorithm was designed. And a trusted permission mechanism was added to make the blockchain have both permission and public characteristics. Secondly, the design process of the main and sub chains was described in detail. The management of the chain group and public services was provided by the main chain, while the sub chains were designed independently for different business scenarios, and cross-chain data communication was realized by connecting the main chain relay, thereby realizing the data secure isolation. Finally, an experimental environment was built for testing to verify the feasibility of the permissioned public blockchain design. Experimental results show that compared with some existing blockchains such as the Hyperledger Fabric, the proposed permissioned public blockchain has significant advantages, including a throughput of up to 25 000 times per second and an average delay time of about 8 s. It can be seen that this permissioned public blockchain provides technical support for further research on cross-chain data interconnection of different types of blockchains.
The main task of text segmentation is to divide the text into several relatively independent text blocks according to the topic relevance. Aiming at the shortcomings of the existing text segmentation models in extracting fine-grained features such as text paragraph structural information, semantic correlation and context interaction, a text segmentation model TS-GCN (Text Segmentation-Graph Convolutional Network) based on Graph Convolutional Network (GCN) was proposed. Firstly, a text graph based on the structural information and semantic logic of text paragraphs was constructed. Then, the semantic similarity attention was introduced to capture the fine-grained correlation between text paragraph nodes, and the information transmission between high-order neighborhoods of text paragraph nodes was realized with the help of GCN, so that the model ability of multi-granularity extraction of text paragraph topic feature representations was enhanced. The proposed model was compared with the representative model CATS (Coherence-Aware Text Segmentation), and its basic model TLT-TS (Two-Level Transformer model for Text Segmentation), which were commonly used as benchmarks for text segmentation task. Experimental results show that TS-GCN’s evaluation index Pk is 0.08 percentage points lower than that of TLT-TS without any auxiliary module on Wikicities dataset. And the proposed model has the Pk value decreased by 0.38 percentage points and 2.30 percentage points respectively on Wikielements dataset compared with CATS and TLT-TS. It can be seen that TS-GCN achieves good segmentation effect.
Aiming at the low embedding capacity of Reversible Data Hiding (RDH) in encrypted videos, a high-capacity RDH scheme in encrypted videos based on histogram shifting was proposed. Firstly, 4×4 luminance intra-prediction mode and the sign bits of Motion Vector Difference (MVD) were encrypted by stream cipher, and then a two-dimensional histogram of MVD was constructed, and (0,0) symmetric histogram shifting algorithm was designed. Finally, (0,0) symmetric histogram shifting algorithm was carried out in the encrypted MVD domain to realize separable RDH in encrypted videos. Experimental results show that the embedding capacity of the proposed scheme is increased by 263.3% on average compared with the comparison schemes, the average Peak Signal-to-Noise Ratio (PSNR) of encrypted video is less than 15.956 dB, and the average PSNR of decrypted video with secret can reach more than 30 dB. The proposed scheme effectively improves the embedding capacity and is suitable for more types of video sequences.
Existing product and service quality analysis is often based on questionnaire survey or product reviews, but there are problems such as difficulty in questionnaire collection and invalid data in product reviews. As a bridge between customers and businesses, the customer service dialogue contains rich customer opinions from product to service perspective, however, there are still few studies using customer service dialogues to analyze product and service quality. A product and service quality analysis method based on customer service dialogues was proposed, which firstly combined the product features and service blueprint to determine product and service quality evaluation factors, and used the Important?Performance Analysis (IPA) method to define the importance and performance index of evaluation factors. Then, quantitative analysis of the importance and satisfaction of products and services was performed by using the dialogue topic extraction and sentiment analysis. The method was applied on the real customer service dialogues of a Taobao flagship store which sells disinfection and sterilization products, and 18 evaluation factors were established, whose importance and performance were quantified based on more than 900 thousand real historical customer service dialogues, thereby analyzing the quality of products and services of the flagship store. Finally, a questionnaire on the professional customer service employees was carried out to verify the effectiveness of the proposed method.
Aiming at the problem that most existing directed network link prediction methods only focus on the directed local and reciprocal link information and ignore the directed global structure information, a High-order Self-included Collaborative Filtering (HSCF) link prediction framework was proposed. Firstly, random walk method was used to calculate the high-order similarity matrix to preserve the high-order path information of the directed network. Secondly, an HSCF framework was constructed by combining the high-order similarity matrix with collaborative filtering method. Finally, the proposed framework was integrated with four typical directed structure similarity indices including Directed Common Neighbor (DCN), Directed Adamic-Adar (DAA), Directed Resource Allocation (DRA) and potential theory (Bifan), and four directed network prediction indices HSCF-DCN, HSCF-DAA, HSCF-DRA and HSCF-Bifan were proposed on this basis. Compared with the baseline indices on ten real directed networks, the experimental results show that the AUC (Area Under Curve of Receiver Operating Characteristic (ROC)) values of HSCF-DCN, HSCF-DAA, HSCF-DRA and HSCF-Bifan are increased by an average of 8.16%, 8.85%, 9.64% and 10.33% respectively and the F-score values of them are increased by an average of 66.62%, 68.32%, 68.95% and 76.18% respectively.
In view of the small target of cigarette butts in surveillance videos of public places and the easy divergence of smoke generated by smoking, it is difficult to determine the smoking behavior only by target detection algorithm. Considering that the algorithm of posture estimation using skeleton key points is becoming more and more mature, a smoking behavior detection algorithm was proposed by using the relationship between human skeleton key points and smoking behavior. Firstly, AlphaPose and RetinaFace were used to detect the key points of human skeleton and face respectively. According to the ratio of distance between wrist and middle point of two corners of mouth and between wrist and the eye on the same side, a method for calculating whether the Smoking Action Ratio (SAR) in humans falls within the Golden Ratio of Smoking Actions (GRSA) to distinguish smoking from non-smoking behaviors was proposed. Then, YOLOv4 was used to detect whether cigarette butts existed in the video. The results of GRSA determination and YOLOv4 were combined to determine the possibility of smoking behavior in the video and make a determination of whether smoking behavior was present. The self-recorded dataset test shows that the proposed algorithm can accurately detect smoking behavior with the accuracy reached 92%.
Concerning that the traditional saliency detection algorithm has low segmentation accuracy and the deep learning-based saliency detection algorithm has strong dependence on pixel-level manual annotation data, an unsupervised saliency object detection algorithm based on graph cut refinement and differentiable clustering was proposed. In the algorithm, the idea of “coarse” to “fine” was adopted to achieve accurate salient object detection by only using the characteristics of a single image. Firstly, Frequency-tuned algorithm was used to obtain the salient coarse image according to the color and brightness of the image itself. Then, the candidate regions of the salient object were obtained by binarization according to the image’s statistical characteristics and combination of the central priority hypothesis. After that, GrabCut algorithm based on single image for graph cut was used for segmenting the salient object finely. Finally, in order to overcome the difficulty of imprecise detection when the background was very similar to the object, the unsupervised differentiable clustering algorithm with good boundary segmentation effect was introduced to further optimize the saliency map. Experimental results show that compared with the existing seven algorithms, the optimized saliency map obtained by the proposed algorithm is closer to the ground truth, achieving an Mean Absolute Error (MAE) of 14.3% and 23.4% on ECSSD and SOD datasets, respectively.
Service description texts for Internet of Things (IoT) are short in length and sparse in text features, and direct modeling the IoT service by using traditional topic model has poor clustering effect, so that the best service cannot be discovered. To solve this problem, an IoT service discovery method based on Biterm Topic Model (BTM) was proposed. Firstly, BTM was employed to mine the latent topic of the existing IoT services, and the service document-topic probability distribution was calculated and deduced through global topic distribution and theme-word distribution. Then, K-means algorithm was used to cluster the services and return the best matching results of service requests. Experimental results show that the proposed method can improve the clustering effect of services for IoT and thus obtain the matched best service. Compared with the methods of HDP (Hierarchical Dirichlet Process) and LDA-K (Latent Dirichlet Allocation based on K-means), the proposed method achieves better performance in terms of Precision and Normalized Discounted Cumulative Gain (NDCG) for best service discovery.
In view of maintenance difficulties and high cost in large-scale development of Wireless Local Access Network (WLAN), the Control and Provisioning of Wireless Access Points (CAPWAP) protocol that applied to communication between Access Controller (AC) and Wireless Terminator Point (WTP) was researched and implemented. In Linux environment, main features were realized, such as state machine management, and WTP centralized configuration. A platform of WLAN centralized management system based on local Medium Access Control (MAC) framework was built up. Wireshark capture tool, Chariot and Iperf were used to test the platform. The capture test results verify the feasibility of the framework, and the results of throughput and User Datagram Protocol (UDP) test also show that network performance is efficient and stable.
The existing methods of constructing Voronoi diagram have low efficiency and high complexity, to remedy the disadvantages, a new method of constructing and updating Voronoi diagram based on the hybrid methods was given to query the nearest neighbor of the given spatial data effectively, and a new method of searching the nearest neighbor based on Voronoi diagram and the minimum inscribed circle was presented. To deal with the frequent, changes of the query point position, the method based on Voronoi diagram and the minimum bounding rectangle was proposed. To improve the efficiency of the dual nearest neighbor pair and closest pair query, a new method was given based on Voronoi polygons and their minimum inscribed circles. The experimental results show that the proposed methods reduce the additional computation caused by the uneven distribution of data and have a large advantage for the big dataset and the frequent query.