Small object detection is a challenging but vital task for applications such as search and rescue, surveillance, and remote sensing, where the goal is to detect small targets such as people, vehicles, or boats in complex environments. The minimal pixel representation of such targets, combined with cluttered backgrounds and environmental variations like lighting changes, makes accurate detection particularly difficult. To address these challenges, we propose Focal Spotter, a novel transformer-based detector that incorporates two key modules:the Energy-Based Focal Module (EFM) and the Focalized Inter-Scale Feature Complementary Module (FICM). The EFM leverages an Energy-Based Model(EBM) to achieve inner-scale feature focalization. The EBM dynamically allocates weights through an energy function, focusing on low-energy target signals (such as small object features), significantly enhancing the model’s sensitivity to sparse or weak signals. This capability allows EFM to outperform conventional attention methods in both robustness and precision.The FICM facilitates inter-scale feature focalization by aligning and integrating high-level semantic and low-level detailed features spatially and channel-wise. By preemptively extracting and merging compensatory information across scales, it boosts feature fusion efficiency, minimizes redundancy, ensures semantic consistency, and achieves precise spatial localization, significantly enhancing small object detection in complex scenes. Extensive experiments on multiple benchmark datasets, including SeaDronesSee and VisDrone, demonstrate that Focal Spotter outperforms state-of-the-art methods in small object detection. Ablation studies highlight the critical contributions of EFM and FICM, with EFM showing superior performance over other attention mechanisms in capturing sparse small targets. The results underscore the robustness and effectiveness of our approach across diverse and challenging scenarios, from maritime to urban environments.

LiDAR point cloud semantic segmentation has emerged as a fundamental technique for enabling intelligent perception in autonomous driving, robotics, and geospatial analysis. Point cloud segmentation methods are typically categorized into three types: point-based, voxel-based, and projection-based techniques. While point-based and voxel-based methods offer robust feature extraction, they face challenges related to computational efficiency and the handling of large-scale point clouds. Projection-based methods, on the other hand, project 3D point clouds into 2D representations, enabling the application of established 2D convolutional neural networks (CNNs) for segmentation tasks. Despite their advantages in efficiency, projection-based methods often suffer from the loss of spatial precision, leading to suboptimal segmentation performance, especially in complex and cluttered environments. In this paper, we propose a novel projection-based approach for semantic segmentation that addresses the limitations of existing methods. Our approach introduces a Multi-Scale Feature Embedding (MSFE) module to enhance feature extraction from the projected range images, combined with a Multi-Feature Fusion Module (MFFM) to integrate features at multiple scales. We further improve segmentation accuracy for challenging objects, such as pedestrians, traffic signs, and occluded structures, by utilizing a Dual Segmentation Head. Our experiments on the SemanticPOSS and SemanticKITTI datasets show significant improvements over existing methods, achieving a mean Intersection over Union (mIoU) of 53.6% on SemanticPOSS and 67.8% on SemanticKITTI. Notably, we achieve high performance on small and occluded objects, like trashcans (55.9% mIoU) and fences (49.5% mIoU), demonstrating the effectiveness of our approach for real-world applications like autonomous driving.

This article presents a generalized method for enhancing visual simultaneous localization and mapping (SLAM) by leveraging the maximum information entropy in texture distribution. Traditionally, visual SLAM methods are based on the assumption of independent and identically distributed normal models for observed errors, uniformly minimizing the combined errors to estimate the states, while failing to account for the information differences between individual observations, which may hinder achieving ideal results. In response to this, we introduce the texture distribution entropy to quantify the information content of each feature, where higher entropy indicates greater importance. Then, the conventional state estimation strategy is improved by assigning weights to points based on their importance, promoting a balance between the contributions of regions with various texture densities, thereby enhancing the reliability of the results under uneven observations. Moreover, based on the maximum distribution entropy of texture features, a novel keyframe decision strategy is proposed that fully evaluates the utilization of current texture information and the richness of the scene texture, ensuring effective and timely keyframe construction. The proposed method is highly generalizable and can be applied to a wide range of visual SLAM systems, provided they are based on texture feature-based state estimation or keyframe decision. Finally, experimental results on public datasets demonstrate that, compared to recent state-of-the-art visual SLAM methods, our method significantly improves the accuracy of trajectory tracking estimation.

With the recent development of the joint classification of hyperspectral image (HSI) and light detection and ranging (LiDAR) data, deep learning methods have achieved promising performance owing to their locally sematic feature extracting ability. Nonetheless, the limited receptive field restricted the convolutional neural networks (CNNs) to represent global contextual and sequential attributes, while visual image transformers (VITs) lose local semantic information. Focusing on these issues, we propose a fractional Fourier image transformer (FrIT) as a backbone network to extract both global and local contexts effectively. In the proposed FrIT framework, HSI and LiDAR data are first fused at the pixel level, and both multisource feature and HSI feature extractors are utilized to capture local contexts. Then, a plug-and-play image transformer FrIT is explored for global contextual and sequential feature extraction. Unlike the attention-based representations in classic VIT, FrIT is capable of speeding up the transformer architectures massively and learning valuable contextual information effectively and efficiently. More significantly, to reduce redundancy and loss of information from shallow to deep layers, FrIT is devised to connect contextual features in multiple fractional domains. Five HSI and LiDAR scenes including one newly labeled benchmark are utilized for extensive experiments, showing improvement over both CNNs and VITs.

Face detection is a fundamental task for numerous face-related applications (e.g., face recognition and age estimation), which directly affects the performance of the subsequent processing. Recent anchor-based face detectors have demonstrated the great potential by matching anchors and target boxes during training, which is crucial for high performance and training efficiency. However, existing anchor matching strategies still suffer from: 1) ignoring the inherent relationship between the targets and the anchors, which may cause unsuitable matched pairs, 2) adopting a fixed matching threshold, which cannot meet the varying demands of matched pairs for quality and quantity in different feature levels and training processes, and 3) the heuristic anchor setting, whose matching range is too narrow to capture about 20% target faces in the training dataset. This paper proposes an Adaptive Anchor Matching Strategy (AAMS) to address these issues, which selectively assigns proper targets to anchors in different feature levels by using adaptive matching thresholds and a robust anchor setting determined by the statistical characteristic of the training samples. Extensive experiments on popular benchmarks reveal that the proposed approach has significant improvements on anchor-based models and outperforms the recent state-of-the-arts methods in terms of both accuracy and generalization.

Abstract Skeleton‐based action recognition algorithms have been widely applied to human action recognition. Graph convolutional networks (GCNs) generalize convolutional neural networks (CNNs) to non‐Euclidean graphs and achieve significant performance in skeleton‐based action recognition. However, existing GCN‐based models have several issues, such as the topology of the graph is defined based on the natural skeleton of the human body, which is fixed during training, and it may not be applied to different layers of the GCN model and diverse datasets. Besides, the higher‐order information of the joint data, for example, skeleton and dynamic information is not fully utilised. This work proposes a novel multi‐stream adaptive spatial‐temporal attention GCN model that overcomes the aforementioned issues. The method designs a learnable topology graph to adaptively adjust the connection relationship and strength, which is updated with training along with other network parameters. Simultaneously, the adaptive connection parameters are utilised to optimise the connection of the natural skeleton graph and the adaptive topology graph. The spatial‐temporal attention module is embedded in each graph convolution layer to ensure that the network focuses on the more critical joints and frames. A multi‐stream framework is built to integrate multiple inputs, which further improves the performance of the network. The final network achieves state‐of‐the‐art performance on both the NTU‐RGBD and Kinetics‐Skeleton action recognition datasets. The simulation results prove that the proposed method reveals better results than existing methods in all perspectives and that shows the superiority of the proposed method.

Abstract Automatic segmentation of breast cancer lesions in dynamic contrast-enhanced magnetic resonance imaging is challenged by low accuracy of delineation of the infiltration area, variable structure and shapes, large intensity heterogeneity changes, and low boundary contrast. This study constructed a two-stage breast cancer image segmentation framework and proposes a novel breast cancer lesion segmentation model (TR-IMUnet). The benchmark U-Net network model enables a rough delineation of the breast area in the acquired images and eliminates the influence of unrelated tissues (chest muscle, fat, and heart) on breast tumor segmentation. Based on the extracted results of the region of interest, the rectified linear unit (ReLU) function of the encoding–decoding structure in the model was replaced by an improved ReLU function to reserve and adjust the data dynamically according to input information. The segmentation accuracy of breast cancer lesions was improved by embedding a multi-scale fusion block and a transformer module in the coding path of the model, thereby obtaining multi-scale and global attention information. The experimental results showed that the breast tumor segmentation indexes Dice coefficient (Dice), Intersection over Union (IoU), Sensitivity (SEN), and Positive Predictive Value (PPV) increased by 4.27, 5.21, 3.37, and 3.68%, respectively, relative to the U-Net reference model. The proposed model improves the segmentation results of breast cancer lesions and reduces small area mis-segmentation and calcification segmentation.

Vision-based Marine ship detection has been explored for many years, effectively improving maritime transportation management. However, many ship detection methods get troubles on the movable platform, which are summarized as follows: 1) high time efficiency is needed to handle the rapid changing of maritime scenes so that ships can be detected successfully on movable platform; 2) different appearances caused by sizes and viewpoints enlarge the intra-class distance of ships, which exceed the representation capability of some features like histogram of oriented gradients (HOG) with fixed size. For addressing these issues, we propose a rapid ship detection method based on multi-size gradient features and multi-branch support vector machine (SVM) in a “coarse-to-fine” manner that can be applied on a movable platform. The proposed multi-size gradient features are used to represent the ships with different sizes, including the coarse and fine gradient features. To speed up the detection process in a sliding way, the coarse ship locations are firstly generated only based on the coarse gradient features, which highly reduces the computational cost. Then, the multi-size gradient features extracted from these locations are determined by a multi-branch SVM model which is designed to deal with features of different dimensions and improve the precision of ship position with fine gradient features. The proposed method is tested on the changing background where ships have different sizes. Experimental results show that the proposed method obtains average precision (AP) of 0.795 and its detection speed achieves 60.6 frame per second, which achieve real-time performance and satisfactory detection precision.

Insufficient illumination and illumination variation in image sequences make it challenging for algorithms to obtain clear outlines for objects in motion. This study proposes a high‐performance adaptive dual fractional‐order variational optical flow model which could be used to resolve these issues. The proposed method revitalises the original dual fractional‐order optical flow model and adopts a fractional differential mask in both the data and smoothness terms of the traditional Horn–Schunck model. The main innovation of this work is to fit a flow field regional to a variety of fractional‐order differential masks. The domain of each region is determined adaptively. The order and size of the fractional‐order differential masks for each region are adjusted by image signal to noise ratio while the shape of the fractional‐order differential mask is regulated to prevent interference from surrounding regions. Adjusting the fractional‐order differential mask adaptively enables the proposed method to accurately segment motion objects in poor and variable illumination regions as well. The experimental results show that our algorithm outperforms the current state‐of‐the‐art algorithms on low‐light real scene videos and also achieves competitive results on the Middlebury, KITTI and MPI Sintel public benchmarks.

Classification of benign and malignant pulmonary nodules can provide useful indicators for estimating the risk of lung cancer. In this study, an improved random forest (RF) algorithm is proposed for classification of benign and malignant pulmonary nodules in thoracic computed tomography images. First, an improved random walk algorithm is proposed to automatically segment pulmonary nodules. Then, intensity, geometric and texture features based on the grey‐level co‐occurrence matrix, rotation invariant uniform local binary pattern and Gabor filter methods are combined to generate an effective and discriminative feature vector. Mutual information is employed to reduce the dimensionality. Finally, an improved RF classifier is trained to classify benign and malignant nodules. An appropriate feature subset is selected by the bootstrap method and an effective combination method is introduced to predict a class label. The proposed classification method on the lung images dataset consortium dataset achieves a sensitivity of 0.92 and the area under the receiver‐operating‐characteristic curve of 0.95. An additional evaluation is performed on another dataset coming from General Hospital of Guangzhou Military Command. A mean sensitivity and a mean specificity of the proposed method are 0.85 and 0.82, respectively. Experimental results demonstrate that the proposed method achieves the satisfactory classification performance.

Pulmonary nodules are potential manifestation of lung cancer. Accurate segmentation of juxta-vascular nodules and ground glass opacity (GGO) nodules are an important and active area of research in medical image processing. At present, the classical segmentation algorithm of pulmonary nodules can not accurately obtain the boundary information of pulmonary nodules. In order to solve the problem, a new segmentation algorithm of pulmonary nodules using Fuzzy clustering based on coefficient of curvature is proposed in this paper. Because the coefficient of curvature can effectively distinguish between pulmonary nodules and surrounding structures, and Fuzzy clustering algorithm is adopted, the boundary information of pulmonary nodules can be accurately obtained. In order to obtain initial segmentation results of pulmonary nodules, the expectation maximization algorithm is adopted. Experimental results show that the proposed algorithm can accurately obtain the boundary information of juxta-vascular nodules and GGO nodules.

In this paper, a neural network approach is presented for the motion control of constrained flexible manipulators, where both the contact force everted by the flexible manipulator and the position of the end-effector contacting with a surface are controlled. The dynamic equations for vibration of flexible link and constrained force are derived. The developed control, scheme can adaptively estimate the underlying dynamics of the manipulator using recurrent neural networks (RNNs). Based on the error dynamics of a feedback controller, a learning rule for updating the connection weights of the adaptive RNN model is obtained. Local stability properties of the control system are discussed. Simulation results are elaborated on for both position and force trajectory tracking tasks in the presence of varying parameters and unknown dynamics, which show that the designed controller performs remarkably well.