Recently, applying the utilization of RGB-D data for robot perception tasks has garnered significant attention in domains like robotics and autonomous driving. However, a prominent challenge in this field lies in the substantial impact of feature robustness on both segmentation and pose estimation tasks. To tackle this challenge, we proposed a pioneering two-stage hybrid Convolutional Neural Network (CNN) architecture, which connects segmentation and pose estimation in tandem. Specifically, we developed Cross-Modal (CM) and Cross-Layer (CL) modules to exploit the complementary information from RGB and depth modalities, as well as the hierarchical features from diverse layers of the network. The CM and CL integration strategy significantly enhanced the segmentation accuracy by effectively capturing spatial and contextual information. Furthermore, we introduced the Convolutional Block Attention Module (CBAM), which dynamically recalibrated the feature maps, enabling the network to focus on informative regions and channels, thereby enhancing the overall performance of the pose estimation task. We conducted extensive experiments on benchmark datasets to evaluate the proposed method and achieved exceptional target pose estimation results, with an average accuracy of 94.5% using the ADD-S AUC metric and 97.6% of ADD-S smaller than 2 cm. These results demonstrate the superior performance of our proposed method.

The mechanism of spatio-temporal gene expression is significantly related to the interaction between the two regulatory elements on the DNA, enhancer and promoter. Identifying enhancer-promoter interactions that disrupt cell-specific gene expression and cause different human diseases remains to be a great challenge. To figure this out, we construct a sequence-based deep learning model, Enhancer-Promoter interactions prediction network, briefly called the EPnet which accurately predicts the interaction between enhancer and promoter with given DNA sequences. The method we proposed requires no genomic data which makes it convenient to make predictions. Comparison with other existing methods and application on predicting interactions show that our method is of superior performance in multiple cell lines which proves that our model is trustworthy and robust.

In dense maritime environments with complex shipping routes, small-scale vessel radar imaging, frequent occlusions, and severe sea clutter present catastrophic challenges for maritime radar-based multi-vessel tracking. Precisely modeling vessel targets under such interference and achieving robust tracking through occluded routes is therefore critical for maritime surveillance tasks. To this end, we propose RadarTracker, a novel end-to-end temporal Transformer framework for X-band marine radar multi-vessel tracking. RadarTracker consists of a CNN-based feature extractor, a density enhancement encoder, and an occlusion-aware dynamic decoder. Specifically, the encoder is built upon a Deformable Transformer backbone, integrated with a density-aware module and density feature enhancement to address the low signal-to-noise ratio and densely distributed vessel targets in radar imagery. Subsequently, to achieve accurate trajectory regression under varying traffic densities and complex occlusions along shipping routes, we design an occlusion-aware decoder with a dynamic query allocation mechanism that enables seamless target initialization, identity preservation, and short-term re-identification, effectively balancing tracking accuracy and efficiency under diverse traffic conditions. Finally, we construct a high-density shipping lane dataset collected along the Yangtze River in China to validate the proposed approach. Experimental results demonstrate that RadarTracker achieves a consistent 5% performance improvement over state-of-the-art baselines across varying traffic densities, while maintaining deployment-friendly computational efficiency. The implementation of our method will be publicly available at https://github.com/wangzihanggg/RadarTracker.

In the civil structural health monitoring fields, monitored data suffer from noise and sensor faults. In practice, redundant sensors are usually deployed to monitor structural condition to obtain more accurate and robust information. This paper proposes a beamforming-based spatial filtering method to improve the data quality by using the information redundancy within sensor networks. Data pre-processing is first implemented, including missing data imputation and thermal response separation. Subsequently, short-term Fourier transform is used to transform the measured time sequences into time–frequency domain to obtain more useful features. Finally, signals in the time and frequency domain are processed using the beamforming algorithm. In the beamformers, a linear filter is applied to suppress noise signals, which is formulated as a constrained optimization problem. Herein, interior point algorithm is used to optimize the allocation of the linear filter, wherein the objective function is to minimize the power of the noise component at the beamformer output. The effectiveness of the proposed method is verified by using signals from strain gauges installed on steel deck plates of the 3^rd Nanjing Yangtze River Bridge. Results through the case study show that signals after spatial filtering have a satisfactory de-noising, which indicates the effectiveness of the proposed beamforming algorithm. We believe that the proposed beamforming algorithm has substantial potential applications, such as providing high quality data source for further investigations.

Any minor fault may spread, accumulate and enlarge through the causal link of fault in a closed-loop complex system of civil aircraft, and eventually result in unplanned downtime. In this paper, the fault propagation path model (FPPM) is proposed for system-level decomposition and simplifying the process of fault propagation analysis by combining the dual strategy ant colony optimization algorithm (DS-ACO) with the Bayesian network (BN). In FPPM, the modeling of the fault propagation path consists of three models, namely shrinking model (SM), ant colony optimization model (ACOM) and assessment model (AM). Firstly, the state space is shrunk by the most weight supported tree algorithm (MWST) at the initial establishment stage of BN. Next, DS-ACO is designed to improve the structure of BN in order to study the fault propagation path accurately. Finally, the experiment is conducted from two different perspectives for the rationality of the well-trained BN’s structure. Experimental results illustrate the effectiveness of the proposed model.

Place recognition (PR) is a key component of simultaneous localization and mapping (SLAM) in autonomous vehicles and robotics. By efficiently matching descriptors generated from the current scene with a prebuilt reference database, existing PR methods enable accurate vehicle re-localization. However, achieving robust and efficient PR remains challenging, particularly in scenarios with sensor corruptions, such as camera overexposure and LiDAR beam missing, which degrade the performance of traditional PR systems. To address this issue, we propose LiteFT-PR, a fault-tolerant and lightweight LiDAR-camera fusion PR network. To balance accuracy and efficiency under extreme conditions, we adopt a knowledge distillation framework for PR tasks: the teacher network learns robust representations under corrupted sensing, while the student distills and inherits this knowledge for lightweight deployment. In addition, we design efficient multimodal fusion mechanisms for both networks, enabling robust place recognition under various sensor corruption scenarios while significantly reducing computational overhead. To meet real-world deployment needs, we build a dataset covering multiple types of sensor failures and conduct extensive field experiments. Experimental results demonstrate that LiteFT-PR, with only 7.24 M of parameters, outperforms state-of-the-art PR methods in both accuracy and efficiency.