International Journal of Pattern Recognition and Artificial Intelligence

Every second, a huge volume of multi-dimensional data is generated in fields such as Social Networking, Industrial Internet of Things, Stock market and E-commerce applications. Knowledge and pattern extraction are a challenging task in the evolving nature of data stream. Major issues are (i) ‘concept drift’ occurs as a result of pattern changes in the data distribution and (ii) ‘concept evolution’ occurs when a new class evolves in the data stream. These issues degrade the performance of learning models. In this paper, we focus on detection of concept evolution and enhance the performance of classifiers. For this, we propose a new model to identify novel classes, namely, Detection of Novel Classes (DNC). The proposed method adopts long short term memory to continuously observe the streaming data in order to detect emerging classes. The continuous monitoring allows the model to distinguish between existing classes and the novel classes which save time and memory. Also, the proposed method is demonstrated for identifying more than one novel class. The experiments are performed over seven different datasets. The results confirm the efficiency is increased ranging from 6% to 34% by the proposed method in identifying new concepts in the evolving data stream than the existing methods available in the literature.

Unmanned surface vessels (USVs) are generally subjected to wind force, wave and undercurrent action. Some unidentified objects such as underwater obstacles and submerged reefs should be monitored and evaded with converged communication, autonomous control and network system so as to achieve autonomous and automatic avoidance of danger and real-time control of flight path. Next, based on performing interference mechanics analysis on the USVs under the disturbance of wind force and wave on water surface, the relationship parameters between the bow angle, the deviation angle, safety evaluation function and the external force were obtained, and the comprehensive information of water surface environment was evaluated in combination with the dynamic window approach (DWA) algorithm. Further, the related model was established for simulation by using the simulation software MATLAB, and both reliability and stability of the modified DWA algorithm were designed and examined. Accordingly, the modified DWA algorithm can enhance the safety of the USVs in autonomous path planning and achieve rapid obstacle avoidance and autonomous path planning of the USVs. Our simulation revealed that the modified DWA algorithm can safely and rapidly correct the bow angle and the deviation angle, optimize the path trajectory and reduce the risk of roll-over of the USVs under sudden external force.

Hyperspectral Image (HSI) with its high resolution spatial and spectral information, has important applications in military, aerospace and civil applications. The classification methods have become the focus of the field as a significant research aspect of hyperspectral remote monitor engineering for earth reflexion. Because of its high dimensional nature, high relation between bands and spectral variety, traditional classification methods are difficult to achieve high precision and accuracy which limits the development of HSI classification technology. In the past years, with the fast recrudesce of deep learning engineering, its powerful feature extraction ability can remarkably ameliorate the accuracy of HSI classification, HSI classification on account of deep learning has become a feasibility study hotspot. In this paper, the methods of HSI classification on account of deep learning are reviewed. First, the research background of HSI classification is introduced and the deep neural network models which are expensively used in the field of HSI classification are summarized. On this basis, some HSI classification methods on account of deep learning are introduced in detail. Finally, the breakthrough aspects of deep learning in the map of HSI classification are summarized at the current stage and the future research direction is prospected.

Advanced scientific and technological measurement methods are the basis of sports research and the progress of sports research methods. It promotes the continuous improvement of sports technology to meet athletes’ pursuit of excellent sports achievements. Therefore, the method of sports scientific research must be changed from the traditional empirical training mode to the programmed training method, from the qualitative analysis of the training effect to the nuanced analysis of the training process. Based on this, this paper uses in-depth learning to study sports motion capture. First, two-dimensional human joints are extracted based on Mask R-CNN. Then, the 3D human motion skeleton is constructed by using the binocular vision system, and the Mask R-CNN human pose estimation algorithm is optimized. On this basis, a sports motion capture system is designed, and the system’s accuracy is verified. The error of the depth information obtained by the sports capture system constructed in this paper is less than 3% in the experiment of about 2m. It has strong practicability.

To address the problem of missed diagnosis in rib fracture detection from CT scans, this study introduces an enhanced model, called Faster-RCNN-SE-FA, which is built upon the traditional Faster-RCNN architecture. The proposed model integrates a novel filter anchor method and thoroughly considers the specific imaging characteristics of ribs in CT images. The preprocessing of the image is followed by applying the Squeeze-and-Excitation (SE) module, which enhances the discrimination of features in the channel dimension while preserving the location Sensitivity (Sen) important for target detection tasks. Consequently, this modification leads to a significant improvement in model performance. Empirical experiments, conducted on CT sequences of 130 rib feature cases provided by the First Affiliated Hospital of Ningbo University, demonstrate that the Faster-RCNN-SE-FA model achieves better Sen and accuracy compared to traditional methods, including the baseline Faster-RCNN.

With the continuous development of autonomous driving technology, realizing high-precision road obstacle detection is crucial to ensure traffic safety and driving experience. However, traditional obstacle detection methods often perform poorly in complex driving scenarios, such as obstacle movement and occlusion. To cope with this problem, this study proposes a road obstacle detection method based on a two-stream convolutional neural network model, aiming to overcome the limitations of traditional methods in capturing spatiotemporal features and handling complex situations. Our research approach is based on the following innovations: First, we introduce a dual-stream convolutional neural network structure, where one stream is used to extract spatial features from the contour information of the obstacle frames, and the other stream extracts temporal features from the temporal stream information. This dual-stream structure can fully capture the appearance and dynamic information of the obstacles, thus improving detection accuracy. Second, we design a feature fusion module to fuse the two features to obtain richer obstacle features. In addition, we propose a new loss function, i.e. clustering loss, for better optimizing the model training process, reducing intra-class variation, and increasing inter-class differences, thus improving the generalization performance of the model. In the experimental section, we conducted extensive experimental analysis using Citycapes and BDD100K datasets. The experimental results show that our model achieves significant performance improvement compared to both the traditional convolutional neural network method and the YOLOv5 method in a variety of scenarios such as obstacle stationary, obstacle moving, and obstacle with occlusion. Specifically, our method improves the recognition rate up to 4.8% to 14.5% on the Citycapes dataset and 6.5% to 12.8% on the BDD100K dataset, respectively, under different scenarios. In addition, our model also exhibits more advantages on small datasets, showing higher generalization ability and robustness.

Few-Shot Class Incremental Learning (FSCIL) is a trending topic in deep learning, addressing the need for models to incrementally learn novel classes, particularly in real-world scenarios where continuously emerging classes come with limited labeled samples. However, the majority of FSCIL research has been dedicated to image classification and object recognition tasks, with limited attention given to video action classification. In this paper, we present a new Cluster Compression and Generative Separation (CCGS) method for Incremental Few-Shot Video Action Recognition (iFSVAR), which introduces contrastive learning to boost the degree of class separation in the base session. Simultaneously, it creates numerous fine-grained classes with diverse semantics, effectively filling the unallocated representation space. Experimental results on UCF101, Kinetics, and Something-Something-V2demonstrate the effectiveness of the framework.

The fusion of millimeter-wave radar and camera for three-dimensional (3D) object detection represents a pivotal technology for autonomous driving, yet it is not without its inherent challenges. First, the radar point cloud contains clutter, which can result in the generation of impure radar features. Second, the radar point cloud is sparse, which presents a challenge in fully extracting the radar features. This can result in the loss of object information, leading to object misdetection, omission, and a reduction in the robustness. To address these issues, a 3D object detection method based on the semantic information of radar features and camera fusion (Semantics-Fusion) is proposed. Initially, the image features are extracted through the centroid detection network, resulting in the generation of a preliminary 3D bounding box for the objects. Subsequently, the radar point cloud is clustered based on the objects’ position and velocity, thereby eliminating irrelevant point cloud and clutter. The clustered radar point cloud is projected onto the image plane, thereby forming a radar 2D pseudo-image. This is then input to the designed 2D convolution module, which enables the full extraction of the semantic information of the radar features. Ultimately, the radar features are fused with the image features, and secondary regression is employed to achieve robust 3D object detection. The performance of our method was evaluated on the nuScenes dataset, achieving a mean average precision (mAP) of 0.325 and a nuScenes detection score (NDS) of 0.462.

Person re-identification methods currently encounter challenges in feature learning, primarily due to difficulties in expressing the correlation between local features and integrating global and local features effectively. To address these issues, a pose-guided person re-identification method with Two-Level Channel–Spatial Feature Integration (TLCSFI) is proposed. In TLCSFI, a two-level integration mechanism is implemented. At the first level, TLCSFI integrates the spatial information from local features to generate fine-grained spatial features. At the second level, the fine-grained spatial feature and the coarse-grained channel feature are integrated together to complete channel–spatial feature integration. In the method, a Pose-based Spatial Feature Integration (PSFI) module is introduced to generate the pose union feature, which calculates intra-body affinity to guide the integration of spatial information among local pose feature maps. Then, a Channel and Spatial Union Feature Integration (CSUFI) module is proposed to efficiently integrate the channel information of the global feature and the spatial information of the pose union feature. Two individual networks are designed to extract channel and spatial information, respectively, in CSUFI, which are then weighted and integrated. Experiments are conducted on three publicly available datasets to evaluate TLCSFI, and the experimental results demonstrate its competitive performance.

Aiming to address the challenges of low object detection precision in remote sensing images due to high background complexity and significant target scale variations, a novel model for large-scale disparate object detection in remote sensing images is proposed based on the modified YOLOv8. The model incorporates a Context Aggregation Module (CAM) with an attention mechanism in the backbone network to exploit contextual information, enabling multi-scale feature fusion for effective small object detection. The Neck network utilizes GSConv modules and employs the Slim-Neck design paradigm to enhance model robustness, making it better suited for detecting objects in high-complexity backgrounds. Furthermore, the model adopts the Wise-IoU as the loss function, incorporating a dynamic nonmonotonic focusing mechanism and a gradient gain allocation strategy to enhance the overall performance of disparate object detection. The experimental results indicate that promising performance improvements in the face of large-scale variations in remote sensing image targets. Specifically, the model achieves mAP values of 71.4% and 90.3% on the DOTA and NWPU VHR-10 datasets, respectively, representing increases of 4.4% and 3.7% compared to the original model. Compared to other typical algorithms, it also has considerable advantages in both comprehensive detection accuracy and detection speed.

The precise quantification of cerebral hemorrhage volume assumes paramount significance, given its consequential impact on subsequent medical interventions. While antecedent scholarly endeavors have undertaken cerebral hemorrhage volume computations, the attendant accuracy remains amenable to refinement. The main contribution to this research is the development of a rectangle algorithm using the Otsu method, which can be used to calculate the total volume of brain hemorrhage in CT-Scan images of the human brain. The present investigation seeks to augment the segmentation paradigm inherent to the Otsu method, thereby advancing the precision of cerebral hemorrhage volume assessments. The Otsu method is conceived with the purpose of automating the segmentation of the hemorrhagic domain through the determination of threshold values. The refinement of the Otsu method’s accuracy is effectuated by the formulation of an algorithm aimed at maximizing the mean discrepancy between object and background intensities within the object image, predicated upon the ascertained threshold values. These determined thresholds subsequently govern the transformation of grayscale images into binary format, achieved through the quotient of the threshold value and the mean variance, thereby amplifying segmentation precision. The incorporation of variance and mean values in the binarization of grayscale images affords adaptability in threshold modulation commensurate with the intensity spectrum of the image. The research corpus leverages CT-Scan images sourced from RSUP M. Djamil Padang, West Sumatra, encompassing a compendium of 100 image slices culled from a quintet of patients. Empirical findings attest to the fortification of the Otsu method in the segmentation milieu, substantiating the augmentation of threshold value accuracy. This enhancement is conspicuously manifest in the outcomes of segmentation tests, underscoring the superior efficacy of the Otsu method in discerning between objects and backgrounds. The corollary effect is a discernibly more accurate computation of volumes employing the Otsu-T methodology. The import of this research rests in its successful establishment of automated threshold values and the concomitant elevation of segmentation precision.

Machine learning based on genetic algorithms is an important application. The multi-dimensional ordered sample clustering problem is often solved using Fisher’s optimal segmentation method. However, this method has obvious shortcomings when encountering long sample problems due to its high storage requirements during the computation process. Therefore, Fisher’s optimal two-segmentation method is generally used in practical problems instead, which avoids storage problems. But it is prone to local optima. Based on the analysis of the shortcomings of the Fisher optimal segmentation and optimal two-segmentation algorithms, this paper proposes a genetic-based machine learning clustering algorithm, which overcomes the problem of Fisher’s optimal two-segmentation algorithm being prone to local optima and also solves the problem of high storage requirements during the computation process of Fisher’s optimal segmentation method. The application of this algorithm in the optimization system of water environment monitoring points shows that it is effective.

Climate change poses severe risks to the survival of many whale populations, whose habitats and migration patterns are affected by environmental changes. To detect these whales effectively, especially in deep-sea environments, we need to use AI-based techniques to handle the acoustic diversity and variability of different species. However, current methods for whale detection are based on pre- and post-processing steps that reduce their efficiency and generalizability. To address this issue, we present DeepWhaleNet, a novel deep-learning model that automates whale detection in Underwater Passive Acoustic Monitoring datasets. DeepWhaleNet simplifies the detection process by extracting relevant features from raw log-power spectrograms and helps protect these threatened species by supporting conservation efforts. Our model uses a larger short-time Fourier transform as input and a custom ResNet-18 architecture for classification, which enables it to separate whale sounds from noise and capture their temporal and spectral characteristics. We evaluate the performance of DeepWhaleNet and show that it surpasses state-of-the-art methods, achieving an 8.3% improvement in the F-1 score and 21% higher average precision of binary relevance than the baseline method. Moreover, our model demonstrates its versatility and suitability for species-specific retrieval problems through an ablation study on multi-label retrieval problems and a 99.1% recall for Blue Whales.

Diabetic nephropathy (DKD) is a severe diabetes complication and a leading cause of mortality. While urine microalbuminuria and eGFR are common markers, some advanced cases may exhibit normal values. The progression from microalbuminuria to overt proteinuria in DKD is gradual and requires comprehensive physiological tests for diagnosis. This study employs deep reinforcement learning to predict DKD progression based on patients’ physiological data, aiming to assist clinical diagnosis and treatment. A multivariate logistic regression algorithm is used to describe the DKD prediction probability. An optimized DKD progression prediction algorithm (DPDCD) based on deep reinforcement learning determines the best model coefficients, enabling accurate prediction of DKD progression using routine clinical data. Experimental results demonstrate that DPDCD outperforms other algorithms in predicting DKD progression, providing valuable support for clinicians.

The advent of the Internet has led to the emergence of multi-source heterogeneous knowledge graphs, which have become a crucial means of storing and disseminating knowledge. Node importance estimation is a technique that is employed extensively in a number of fields, including recommender systems, intelligent search and resource allocation. This study introduces a Multi-perspective attention mechanism Fusion Algorithm for the mapping of multi-perspective features of knowledge graphs to Node Importance Estimation (MFA-NIE). First, structural embedding features, relational predicate features, and attribute features (both textual and quantitative) are established for the nodes. Subsequently, an enhanced attention mechanism is employed to extract, compress, and fuse these features. The fused hidden layer vector is employed in the design of a key-based attention mechanism, which enables the propagation of messages from neighboring nodes to the source node. This process results in the iterative updating of the source node’s hidden features. Finally, TOPSIS centrality, based on the topology of the source node, is employed to dynamically adjust the mapping between the fused features and node importance. Experiments were conducted on real-world, large-scale, multi-source, heterogeneous knowledge graphs, comparing the MFA-NIE algorithm with traditional and advanced baseline algorithms, including PR, PPR, GENI, RGTN, CLINE, MCRL, and others. The results demonstrate that the MFA-NIE algorithm significantly improves effectiveness and accuracy, showcasing its superiority, versatility, and practical application value.