最新刊期
卷 7 ， 期 2 ， 2025
- 摘要：In response to the governance challenges caused by the development of artificial intelligence, a new concept or belief is proposed that "intelligent technology must be governable and will be governed". Building a future intelligent ecosystem through three major systems: achieving LaSEE→CiSEE in economic and social ecology, reducing cost of logistics and mobility and improving effectiveness and efficiency of spatiotemporal utilization; Promoting the transformation of AIR→IST paradigm shift at science and technical levels, making Control and Autonomous Science and Engineering (CASE) a new and independent interdisciplinary field for K21 education. Advocating SEE→H3O for humanistic lifestyle and human-centered technology for our well-being. Call for the use of governed intelligent technology to ensure a health and sustainable social development.关键词：intelligent ecology;intelligent technology;intelligent society;intelligent economy;intelligent life187|81|0更新时间：2025-06-15
Review Intelligence
- 摘要：The intelligent system of a single vehicle has certain perception limitations, such as the inability to perceive beyond visual range, the lack of global perception capabilities, and restricted perception capabilities in extreme situations. To enhance vehicle perception capabilities, vehicle-road collaborative intelligent perception technology has emerged. This technology enables multiple intelligent agents (such as vehicles, infrastructure, etc.) to share perceived information, integrate and analyze it, and thereby gain a more comprehensive and accurate understanding of the environment. This paper focuses on the latest developments in vehicle-road collaborative intelligent perception methods. Firstly, it introduces the fundamental technological support for vehicle-road collaborative intelligent perception methods. Secondly, it categorizes collaborative modes into four types based on data processing and interaction methods. Then, it delves into three primary collaborative perception methods and commonly used large-scale collaborative perception datasets. Finally, it discusses the challenges faced by vehicle-road collaborative intelligent perception technology and the outlook for future research. Vehicle-road collaborative intelligent perception methods help to compensate for the deficiencies of single-agent perception and achieve comprehensive environmental perception.关键词：vehicle-road collaborative;collaborative perception;information fusion;communication technology;autonomous driving821|5550|0更新时间：2025-06-15
- 摘要：In recent years, blockchain-enabled cryptocurrency market has been rapidly evolving to a trillion-dollar economic system with various kinds of emerging business models and application scenarios. Cryptocurrencies, characterized by decentralization, anonymity, and ease of cross-border circulation, have significantly enriched and evolved the mainstream financial system centered around fiat currencies. However, they are also extensively used in illegal financial activities including money laundering. As such, research on anti-money laundering (AML) regulation for cryptocurrencies has become a burgeoning field of interest, with machine learning and artificial intelligence (AI) emerging as primary methodologies in areas such as cryptocurrency transaction monitoring, transaction network analysis, and security risk assessment. First, the recent research progress on intelligent AML regulation for cryptocurrencies was systematically reviewed. The main models, methods, and application patterns in AML regulatory research were summarized from multiple dimensions, including traditional machine learning, deep learning, ensemble learning, graph analysis, and heuristic approaches. Meanwhile, commonly used datasets in cryptocurrency AML research were compiled, and a performance comparison of existing models and algorithms was provided based on the Elliptic dataset. Finally, the research challenges and future directions of cryptocurrency AML were discussed, aiming to offer valuable references and insights for the prosperity and healthy development of the cryptocurrency and decentralized finance industry.关键词：cryptocurrency;anti-money laundering;blockchain;machine learning;artificial intelligence550|3705|0更新时间：2025-06-15
Surveys and Prospectives
- 摘要：The coastline raid task requires an unmanned surface ship to carry out precise, fixed-point raids in a complex coastal environment. The terrain of the coastal area is highly varied, with static obstacles such as reefs and shoals, as well as moving obstacles like floating objects at sea. Moreover, the task must be completed within a strict time frame. Therefore, real-time, safe, and accurate path planning for the unmanned surface ship is crucial. To address the challenge of balancing global optimization, efficiency, and safety in path planning, a method that integrates global search and local optimization was proposed, based on the A* algorithm and the dynamic window approach (DWA). The A* algorithm computed the global shortest path in a static environment, while the improved DWA optimized local path obstacle avoidance between path nodes. The A* algorithm was enhanced by incorporating a heuristic function with dynamic exponential decay weighting, which reduced the running time and exploration space for global path planning. Additionally, the DWA was improved by introducing an evaluation function based on the angle between the heading of dynamic obstacles and the predicted trajectory of the unmanned surface ship, as well as a distance evaluation function based on a logistic curve, which enhanced the real-time responsiveness of local obstacle avoidance. Simulation results demonstrate that the proposed hybrid algorithm can search for a near-optimal global solution in real-time, effectively handle both static and dynamic obstacles, and significantly improve the task completion efficiency and stability of the unmanned surface ship in complex coastal environments.关键词：coastline raid;A* algorithm;dynamic window approach;path planning652|757|0更新时间：2025-06-15
- 摘要：In machine learning, it typically assumes that training data and testing data of models are drawn from the same distribution. However, in real-world applications, data distributions often differ, resulting in domain shift problems that adversely affect model generalization. Existing domain generalization methods primarily focus on extracting domain-invariant features while overlooking the potential impact of domain-specific features on model predictions. To address this issue, a domain discriminator based on cross-space multi-scale information aggregation was proposed. By capturing multi-scale information, domain-specific features were effectively removed and the extraction of domain-invariant features was enhanced. Additionally, the momentum update inference consistency loss function was employed to leverage the inference consistency of sample category centers, further improving model robustness. Comparative experiments and analysis conducted on multiple public datasets demonstrate that the proposed method exhibits superior performance in domain generalization, effectively mitigating the impact of domain-specific features on model performance and providing a technical reference for addressing domain shift problems.关键词：domain generalization;transfer learning;domain drift;attention mechanism;inference consistency460|183|0更新时间：2025-06-15
- 摘要：Lip reading technology establishes the mapping relationship between lip movements and specific language characters by processing a series of consecutive lip images, thereby enabling semantic information recognition. Existing methods mainly use recurrent networks for spatiotemporal modeling of sequential video frames. However, they suffer from significant information loss, especially when the video information is incomplete or contains noise. In such cases, the model often struggles to distinguish between lip movements at different time points, leading to a significant decline in recognition performance. To address this issue, a lip reading method based on adaptive pooling attention transformer (APAT-LR) was proposed. This method introduced an adaptive pooling module before the multi-head self-attention (MHSA) mechanism in the standard Transformer, using a concatenation strategy of max pooling and average pooling. This module effectively suppressed irrelevant information and enhances the representation of key features. Experiments on the CMLR and GRID datasets showed that the proposed APAT-LR method could reduce the recognition error rate, thus verifying the effectiveness of the proposed method.关键词：attention mechanism;Transformer;Convolutional Pooling;Adaptive394|275|0更新时间：2025-06-15
- 摘要：The task of skin cancer image segmentation is a key task in the field of medical image processing. The commonly used segmentation algorithms can't well balance the computational resource requirements of local information and global context information when performing diagnosis. In addition, the problem of fuzzy tumor boundaries and difficulty in correctly identifying segmentation is also an urgent problem to be solved. Aiming at the above problems, a skin cancer image segmentation model Axial-FNet based on fuzzy convolution combined with gated axial self-attention was proposed. The model was composed of a gated axial self-attention branch and a fuzzy convolutional neural network branch. At the end of the gated axial self-attention branch, a gated weight controller was set to control the proportion and degree of capturing local information and global context information. The fuzzy learning module was fused into the convolutional neural network (CNN) to form a fuzzy neural network branch to extract the local information of the image. The segmentation accuracy was improved by the model while reducing the amount of calculation. The performance of the Axial-FNet model was evaluated on the ISIC 2017 dataset, achieving scores of 74.23%, 83.05%, and 92.89% for MIoU, F1-score, and accuracy, respectively, as well as 80.91%, 88.13%, and 93.10% for the same metrics on the ISIC 2018 dataset. The experimental results show that Axial-FNet has better segmentation accuracy and reliability than other advanced segmentation models.关键词：image segmentation;fuzzy learning;gated axial self-attention;gating mechanism;convolutional neural network151|144|0更新时间：2025-06-15
- 摘要：Facing the problem of satellite navigation spoofing interference detection, a satellite navigation spoofing interference detection method based on LSTM was proposed. Based on the analysis of the effect of spoofing interference on the receiver loop, the direct and indirect information in the tracking loop was selected as the spoofing detection feature parameters. The feature parameters were preprocessed using the sliding window, and the training and test sets were constructed according to the time series to ensure the temporal relevance of the features. Based on the experiments conducted on the TEXBAT dataset from the university of Texas at Austin, this paper’s method can improve the accuracy of spoofing detection by 2.19% and spoofing scene recognition by 99.98% compared with the existing SVM methods, which can effectively solve the problem of navigation spoofing detection and scene recognition.关键词：satellite navigation;spoofing interference;sliding window processing;long and short-term memory neural network377|267|0更新时间：2025-06-15
- 摘要：To enhance the path planning capabilities of unmanned mining trucks in open-pit mining scenarios, a grey wolf optimization-based adaptive hybrid A* and artificial potential field (GWO-HAPF)method was proposed. The proposed method emploied the grey wolf optimization algorithm to adaptively adjust the key parameters of the hybrid A*(HA*) algorithm, achieving a multi-objective balance among path length, smoothness, and planning time. This effectively overcame the HA* algorithm's limited adaptability to fixed parameter settings, significantly improving the quality and adaptability of global path planning. For local path planning, an improved artificial potential field method was adopted, optimizing the repulsive force function and incorporating an escape force mechanism, which effectively enhanced real-time obstacle avoidance feasibility and path smoothness. Experimental results demonstrate that, compared to the standard HA* algorithms and the improved hybrid A* (IHA*) algorithms, GWO-HAPF improves computational efficiency in global planning by an average of 80% and 14.9%, respectively, reduces path length by over 9.8%, and increases smoothness by over 53%. In local path planning, GWO-HAPF achieves a planning time that is 95.8% shorter than IHA*, while its smoothness improves to 10.19% of IHA*. These findings indicate that the proposed method exhibits outstanding advantages in planning efficiency, path length, smoothness, and real-time obstacle avoidance, showcasing its practical application value in path planning for unmanned mining trucks in open-pit mining scenarios.关键词：path planning;grey wolf optimization;adaptive hybrid A*;artificial potential field;unmanned mining truck280|249|0更新时间：2025-06-15
- 摘要：Multimodal emotion recognition has broad applications in mental health detection and affective computing. However, most existing methods rely on either global or local features, neglecting the joint modeling of both, which limits emotion recognition performance. To address this, a Transformer-based dual-stage gated segmented multimodal emotion recognition method (DGM). DGM adopts a segmented fusion architecture was proposed, consisting of an interaction stage and a dual-stage gating stage. In the interaction stage, the OAGL fusion strategy was employed to model global-local cross-modal interactions, improving the efficiency of feature fusion. The dual-stage gating stage integrates local and global features was designed to fully utilize emotional information. Additionally, to resolve the misalignment of local temporal features across modalities, a scaled dot-product-based sequence alignment method was developed to enhance fusion accuracy. Experimental were conducted on three benchmark datasets (CMU-MOSI, CMU-MOSEI, and CH-SIMS), and the results demonstrate that DGM outperforms representative algorithms on multiple datasets, validating its ability to capture emotional details and its strong generalization capability.关键词：Multimodal Emotion Recognition;Scaled Dot-Product Attention;Dual-Stage Gated Fusion;Transformer75|70|0更新时间：2025-06-15
- 摘要：There are many complex factors in the special environment of coal transportation in power plants, such as uneven light, dust interference, and the different shapes, sizes, and materials of foreign objects on the coal conveyor belt. In this complex environment, many current target detection algorithms are not sensitive enough to the characteristics of foreign objects, and it is difficult to effectively distinguish foreign objects with different characteristics. In order to solve this problem, the network structure of the original YOLOv8 algorithm was optimized and a YOLOv8-CPCA detection method was proposed. The feature extraction ability of the model was significantly improved by introducing the channel prior convolutional attention mechanism (CPCA), and high-precision detection of foreign objects in the harsh environment of coal transportation in power plants was achieved. A unique combination of convolution and pooling operations was used by the CPCA attention mechanism to perform global average pooling and maximum pooling on the input feature map, multi-dimensional feature information was deeply mined, and then attention weights for each channel and spatial position were accurately generated through nonlinear transformation, guiding the model to focus on the key feature areas of foreign objects and enhance feature extraction capabilities. Experimental results show that the improved model not only ensures the real-time detection, but also increases the average detection accuracy mAP@0.5 to 92.9%, providing a more accurate solution for foreign object detection on coal conveyor belts and effectively ensuring the safe operation of coal transportation in power plants.关键词：YOLOv8;attention mechanism;CPCA;coal conveyor belt;foreign object detection141|149|0更新时间：2025-06-15
- 摘要：Facial micro expressions are subtle, involuntary facial movements that reveal true emotions. To enhance recognition accuracy by exploiting spatial correlations among facial landmarks, a hierarchical Transformer architecture hierarchical continuous attention network (HCANet) was proposed to effectively leverage the self-attention mechanism for capturing relationships between landmarks in sequences. HCANet models optical flow differences between onset and apex frames to capture local details and reduce identity interference, thereby avoiding the oversight of local details when directly extracting features from full video frames It consists of a Transformer layer for local temporal feature extraction and an aggregation layer for global facial feature learning. Initially, the face was divided into four regions. Within the Transformer layer, continuous attention block (CAB) was introduced to focus on the local, minute muscular movements within individual regions for extracting local temporal features. Subsequently, the aggregation layer concentrated on learning the inter-region interactions to extract global semantic facial features through a cross-layer attention mechanism. Finally, comparative validations were conducted using leave-one-out-cross-validation on four publicly available micro expression datasets (CASME Ⅱ, CASME Ⅲ, SMIC, SAMM) against six other algorithms. The results demonstrate that HCANet achieves improved classification accuracy on the CASME Ⅲ, SMIC and SAMM datasets, and exhibits stronger robustness in complex scenarios (e.g., low frame rates, background noise).关键词：micro expression recognition;facial feature;deep learning;attention mechanism646|230|0更新时间：2025-06-15
Papers and Reports
- Postal code：100079
- Tel：（010）53879206 Email：cjist@bjxintong.com.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09064830号-19
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
- Total Visits：161982 Visits Today：145
0