Ground terrain perception has become the primary visual task for the robust navigation of intelligent systems in unstructured outdoor environments. However, complex ter-rain poses a significant challenge to vision-based perception. This work introduces a novel estimation task using RGB images to facilitate low-cost terrain perception in extracting surface roughness information. The proposed task presents both semantic-aware and edge-aware roughness descriptors at the pixel level instead of a single value for a given image. To promote the research on the proposed novel terrain roughness estimation task, we introduce a multimodal synthetic dataset for terrain perception in outdoor scenes, containing multiple terrain categories, diverse viewpoints, different lighting and weather conditions, as well as semantic and roughness annotations. Additionally, inspired by computer graphics, we introduce TRENet, a roughness estimation architecture to model the intrinsic correlation of depth-normal-roughness. We also perform ablation studies on the effect of each component and diverse types of inputs. Extensive evaluations and comparisons demonstrate that our method can effectively predict pixel-wise terrain surface roughness with high accuracy.

Seismic signal reconstruction in (micro)seismic monitoring presents a persistent challenge, as contemporary research has predominantly focused on temporal correlations while neglecting the spatial distribution of seismic stations. This study formulates the reconstruction task as an optimization problem by modeling seismic signals through time-varying graphs, introducing an auxiliary weight matrix and anL-order temporal difference matrix to effectively capture spatio-temporal dynamics and enhance signal reconstruction. To address this optimization challenge, we propose an iterative solution method and implement Deep Algorithm Unrolling (DAU) techniques for parameter optimization. Empirical validation across four heterogeneous datasets, encompassing both earthquake and landslide-induced microseismic signals, demonstrates substantial improvements in reconstruction accuracy, as evidenced by reduced Root Mean Square Error (RMSE) compared to existing methodologies. The framework exhibits robust performance under varying noise conditions, with comprehensive waveform analyses confirming high-fidelity signal reconstruction across diverse seismic events. This research advances seismic signal reconstruction through the introduction of a spatio-temporal awareness component, thereby enhancing the reliability and precision of geophysical monitoring.

Semi-supervised domain generalization (SSDG) has recently emerged as a potential research topic. Compared to domain generalization, SSDG represents a realistic and challenging goal, which only requires a few labels from source domains. To tackle this problem, this work presents a novel pseudo-labeling method that facilitates incremental learning on a large amount of unlabeled data. With edge weighting optimization, the proposed method utilizes the graph Laplacian regularizer (GLR) in a multi-class setting that relies on the generated similarity graph. The proposed overall SSDG scheme mitigates the overfitting problem by an adaptive threshold module based on a two-stage GLR denoiser. Our experiments on PACS and OfficeHome verify that the proposed method effectively improves the quality of pseudo-labeling and domain generalization, achieving top performance in terms of accuracy.

With the rapid growth of urban air mobility (UAM) and the potential market opportunities it presents, achieving precise and reliable localization for electrical vertical takeoff and landing (eVTOL) aircraft is crucial. Traditional global navigation satellite system (GNSS) based methods suffer from reliability issues, making localization without GNSS a pressing need. This paper presents a scene-matching localization method based on Bag-of-Visual-Words (BOVW) for autonomous eVTOL. The proposed method overcomes challenges of environmental differences between live images and reference maps, as well as the computational limitations of existing solutions. The system utilizes on-cloud training, on-cloud encoding, and onboard localization stages to ensure robust localization performance. Real-world experiments on an eVTOL prototype– ZJ-Copter validate the effectiveness of the approach, with a matching success rate exceeding 94% and localization accuracy within 3 meters. The method holds promise as a GNSS-independent localization solution for autonomous eVTOLs in the emerging field of UAM.

Automatic seismic signal classification methods are extensively investigated to reduce or replace manual interpretation, with great potential in previous research. Discriminative seismic wave propagation physical characteristics, such as velocities and accelerations, are rarely considered for classification. A multitask learning scheme is proposed that utilises the seismic wave equation and three-dimensional (3D) P-wave velocityVpmodel for signal representation learning. The classifier uses the obtained latent feature maps on a convolutional neural network architecture for classification of rockfall, slide quake, earthquake, and natural/anthropogenic noise events, recorded at an ongoing landslide. Our experimental results show that our approach outperforms state-of-the-art methods.

Panoptic segmentation is a challenging scene understanding task that unifies semantic segmentation and instance segmentation. Namely, each pixel of an image is assigned a semantic label and an instance id. Existing works have elaborated end-to-end panoptic segmentation networks and made great progress in non-proposal-based methods. In this work, we adopt a box-free strategy and incorporate a graph-based clustering method to merge repetitive kernel weights for object instances. An alternative graph-based clustering algorithm like Markov clustering performs effective random walks for unsupervised clustering without pre-defined cluster numbers. Our proposed deep Markov clustering scheme provides an efficient alternative to guarantee instance-aware label prediction in both training and inference stages. On the COCO dataset, our method achieves promising accuracy (PQ=42.1), which is comparable with state-of-the-art methods.

Recent advances in image acquisition and analysis have resulted in disruptive innovation in physical rehabilitation systems facilitating cost-effective, portable, video-based gait assessment. While these inexpensive motion capture systems, suitable for home rehabilitation, do not generally provide accurate kinematics measurements on their own, image processing algorithms ensure gait analysis that is accurate enough for rehabilitation programs. This paper proposes high-accuracy classification of gait phases and muscle actions, using readings from low-cost motion capture systems. First, 12 gait parameters, drawn from the medical literature, are defined to characterize gait patterns. These proposed parameters are then used as input to our proposedmulti-channel time-series classificationand gait phase reconstruction methods. Proposed methods fully utilize temporal information of gait parameters, thus improving the final classification accuracy. The validation, conducted using 126 experiments, with 6 healthy volunteers and 9 stroke survivors with manually-labelled gait phases, achieves state-of-art classification accuracy of gait phase with lower computational complexity compared to previous solutions.1

With growing interest in data-driven classification, deep learning is now prevalent due to its ability to learn feature mapping functions solely from data. For very small training sets, however, deep learning, even with traditional regularization techniques, often overfits, resulting in sub-par classification performance. In this paper, we propose a novel binary classifier deep learning method, based on an iterative quadratic programming (QP) formulation with a graph Laplacian regularizer (GLR), combining the merits of model-based and data-driven approaches. Specifically, the proposed network employs a convolutional neural network (CNN) to learn deep features, which are used to define edge weights for a graph to pose a convex QP problem. Further, we design a novel loss function to penalize samples at the class boundary during semi-supervised learning. Results demonstrate that, given a small-size training dataset, our network outperforms several state-of-the-art classifiers, including CNN, model-based GLR and dynamic graph CNN classifiers.

With growing ageing population, acquiring joint measurements with sufficient accuracy for reliable gait assessment is essential. Additionally, the quality of gait analysis relies heavily on accurate feature selection and classification. Sensor-driven and one-camera optical motion capture systems are becoming increasingly popular in the scientific literature due to their portability and cost-efficacy. In this paper, we propose 12 gait parameters to characterise gait patterns and a novel gait-phase classifier, resulting in comparable classification performance with a state-of-the-art multi-sensor optical motion system. Furthermore, a novel multi-channel time series segmentation method is proposed that maximizes the temporal information of gait parameters improving the final classification success rate after gait event reconstruction. The validation, conducted over 126 experiments on 6 healthy volunteers and 9 stroke patients with handlabelled ground truth gait phases, demonstrates high gait classification accuracy.