Search

The significant differences in data domains between SAR images and the expensive and time-consuming process of data labeling pose significant challenges to terrain classification. Current terrain classification methodologies face challenges in addressing domain disparities and detecting uncommon terrain effectively. Based on Style Transformation and Domain Metrics (STDMs), we propose an unsupervised domain adaptive framework named STDM-UDA for terrain classification in this paper, which consists of two steps: image style transfer and domain adaptive segmentation. As a first step, image style transfer is performed within the image space to mitigate the differences in low-level features between SAR image domains. Subsequently, leveraging this process, the segmentation network extracts image features, employing domain metrics and adversarial training to enhance alignment between domain gaps in the semantic feature space. Finally, experiments conducted on several pairs of SAR images, each exhibiting varying degrees of differences in key imaging parameters such as source, resolution, band, and polarization, demonstrate the robustness of the proposed method. It achieves remarkably competitive classification accuracy, particularly for unlabeled, high-resolution broad scenes, effectively overcoming the domain gaps introduced by the diverse imaging parameters under studies. [ABSTRACT FROM AUTHOR]

Effective features play an important role in synthetic aperture radar (SAR) image interpretation. However, since SAR images contain a variety of terrain types, it is not easy to extract effective features of different terrains from SAR images. Deep learning methods require a large amount of labeled data, but the difficulty of SAR image annotation limits the performance of deep learning models. SAR images have inevitable geometric distortion and coherence speckle noise, which makes it difficult to extract effective features from SAR images. If effective semantic context features cannot be learned for SAR images, the extracted features struggle to distinguish different terrain categories. Some existing terrain classification methods are very limited and can only be applied to some specified SAR images. To solve these problems, a jigsaw puzzle self-supervised learning (JPSSL) framework is proposed. The framework comprises a jigsaw puzzle pretext task and a terrain classification downstream task. In the pretext task, the information in the SAR image is learned by completing the SAR image jigsaw puzzle to extract effective features. The terrain classification downstream task is trained using only a small number of labeled data. Finally, fully connected conditional random field processing is performed to eliminate noise points and obtain a high-quality terrain classification result. Experimental results on three large-scene high-resolution SAR images confirm the effectiveness and generalization of our method. Compared with the supervised methods, the features learned in JPSSL are highly discriminative, and the JPSSL achieves good classification accuracy when using only a small amount of labeled data. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Engineering (17264073) is the property of Republic of Iraq Ministry of Higher Education & Scientific Research (MOHESR) and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The autonomous navigation of Unmanned Ground Vehicles (UGVs) necessitates the precise identification of the surrounding outdoor environment, emphasizing the significance of terrain classification. This study introduces an Intelligent Double Spike Neural Network (IDSNN) that utilizes supervised learning with vision data for the classification of diverse terrains. Specifically, six terrain types are classified: hydrop, gravel, grass, sand, asphalt, and mud. The extraction of texture features, pivotal for the model's input, is conducted using the Local Binary Pattern (LBP) method. The IDSNN model, characterized by its utilization of a multi-spike learning mechanism with temporal coding, demonstrates superior performance in both accuracy and power efficiency. Comparative analyses reveal that the multi-spike learning approach of IDSNN significantly outperforms the single-spike learning employed in the Semi-Recurrent Spike Neural Network (SRSNN). Evaluation metrics, including accuracy, precision, recall, and F1-score, are employed to quantify this advancement. Notably, IDSNN exhibits a 3% improvement in overall accuracy over SRSNN, achieving an impressive accuracy rate of 90.706%. The findings of this study underscore the potential of IDSNN in enhancing the safety and efficiency of Mobile Robot Navigation (MRN) through reliable terrain classification. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Remote Sensing is the property of Editorial Office of Journal of Remote Sensing & Science Publishing Co. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The safety and ride comfort of the emergency rescue vehicle can be improved greatly when the prior knowledge of the terrain in front of the vehicle is used as the active suspension input. In this paper, we present a robust and precise terrain preview detection and classification system based on vision and laser fusion with deep learning. First, a terrain classification method based on ResNet50 and transfer learning fusion is proposed to identify the terrain type in real time and improve the terrain classification accuracy. Second, a robust and precise terrain preview mapping model is developed to integrate the LiDAR with IMU data based on tightly coupled fusion and overcome the accumulation error difficulties caused by multisensor fusion. Third, introducing a decision-level fusion strategy with LiDAR and vision fusion enables the terrain classification results and three-dimensional terrain geometry information to be aligned in space and time, thereby obtaining effective terrain elevation information. As a result, the problem of inaccurate terrain elevation information can be solved. Based on the terrain detection and classification system framework, combined with the improved strategies above, the terrain classification accuracy of the training dataset and test dataset reaches 98% and 83%, respectively. The terrain mapping accuracy achieves approximately 4–5 cm in the large scene. Hence, our method effectively solves poor terrain elevation accuracy caused by the accumulated error of multisensor fusion. The experimental results highlight that the accuracy and robustness of the proposed terrain preview detection and classification algorithm in dealing with rapid vehicle rotation results are better than those of other current methods.

With the increasing demand for mobile robots and autonomous vehicles, several approaches for long-term robot navigation have been proposed. Among these techniques, ground segmentation and traversability estimation play important roles in perception and path planning, respectively. Even though these two techniques appear similar, their objectives are different. Ground segmentation divides data into ground and non-ground elements; thus, it is used as a preprocessing stage to extract objects of interest by rejecting ground points. In contrast, traversability estimation identifies and comprehends areas in which robots can move safely. Nevertheless, some researchers use these terms without clear distinction, leading to misunderstanding the two concepts. Therefore, in this study, we survey related literature and clearly distinguish ground and traversable regions considering four aspects: a) maneuverability of robot platforms, b) position of a robot in the surroundings, c) subset relation of negative obstacles, and d) subset relation of deformable objects. [ABSTRACT FROM AUTHOR]

The advancement of remote sensing has enabled the creation of high-resolution Digital Elevation Models (DEMs). Topographic features such as slope gradient (SG), local convexity (LC), and surface texture (ST), derived from DEMs, are related to subsurface geological conditions. In South Korea, bedrock depth (Dbedrock) and the average shear wave velocity of soil (VSsoil) serve as metrics for determining the site class, which represents the degree of site amplification in seismic design criteria. These metrics, typically measured through geotechnical and geophysical investigations, require predictive methods for preliminary estimation over large areas. Previous studies developed an automatic terrain classification (AC) scheme using SG, LC, and ST, and subsequent research revealed that terrain classification effectively represents subsurface conditions such as Dbedrcok and average shear wave velocity down to 30 m depth. However, AC intrinsically depends on the regional features of DEMs, dividing regions based on nested means of topographic features (SG, LC, and ST). In this study, we developed two terrain classification methods to determine the thresholds of class divisions, aiming to optimize Dbedrock and VSsoil predictions: Sequentially Optimized Classification (SOC) and Non-Sequentially Optimized Classification (NOC). Through the study of the sensitivity of terrain classification methods, smoothing levels, and threshold levels for terrain class generation, we identified the best classification method by comparing it with the geological and mountainous region distribution. Subsequently, we developed DEM-dependent regression models for each class to enhance the accuracy of predicting Dbedrock and VSsoil. The main findings of this study are: (1) the terrain class map suggested in this study represents the distribution of alluvial plane and mountainous regions well, and (2) the DEM calibration for each class provides increased accuracy of Dbedrock and VSsoil predictions in South Korea. We anticipate that the terrain class map, along with Dbedrock and VSsoil maps, will be effectively utilized in geological interpretations and land-use planning for seismic design. [ABSTRACT FROM AUTHOR]

This study aims to enhance the morphometric analysis of terrain and land classification in Romania's coastal basins through the advanced application of Digital Elevation Models (DEMs). While DEMs are increasingly utilized worldwide for environmental mapping and analysis, their application in the Romanian context, especially for detailed geomorphological studies including flow accumulation, sink routes, and landform classification, remains underexplored. Utilizing DEMs, this research integrates Geographic Information System (GIS) technologies with statistical analyses to investigate geomorphological features such as watershed basins, specific catchment areas, the topographic position index, LS factor, plan curvature, and the topographic wetness index. Results demonstrate the effectiveness of DEMs in identifying complex terrain features and in classifying land with high accuracy, highlighting the utility of incorporating advanced morphometric parameters like geomorphons and flow directions. However, the study acknowledges limitations related to the resolution of DEMs and the necessity for ground verification. The findings have significant implications for environmental management, urban planning, and conservation, offering new insights for land use planning and disaster risk reduction in coastal regions. The relevance of this research lies in its targeted application of DEMs to the complex terrain of Romania's coastal basin, contributing significantly to the field of geomorphology by offering nuanced insights and practical methodologies for terrain analysis and land classification. [ABSTRACT FROM AUTHOR]

This paper introduces an implementation of different GIS tools into Streamlit application: FCNN Terrain Classification, Earth Engine Dataset Parsing, and GIS Timelapse Animations. This toolkit is integrated with terrain multi-classification models using Fully Convolutional Neural Networks (FCNNs) for imagery data into Streamlit microservices. The proposed methodology involves labeled and unlabeled data collection from ESA WorldCover and Sentinel-2 MSI on the Google Earth Engine, compressing datasets into TFRecords format with 9 diverse terrain types, and handling Google Cloud training computations. The experimental results demonstrate the effectiveness of the CNN-based approach, achieving a tolerable from 60% up to 80% accuracy of the model and robust classification performance. The simplicity and efficiency of the proposed method make it suitable for real-world tasks requiring reliable and fast GIS analytics. [ABSTRACT FROM AUTHOR]

Identifying terrain parameters is important for high-fidelity simulation and high-performance control of planetary rovers. The wheel–terrain interaction classes (WTICs) are usually different for rovers traversing various types of terrain. Every terramechanics model corresponds to its wheel–terrain interaction class (WTIC). Therefore, for terrain parameter identification of the terramechanics model when rovers traverse various terrains, terramechanics model switching corresponding to the WTIC needs to be solved. This paper proposes a speed-independent vibration-based method for WTIC recognition to switch the terramechanics model and then identify its terrain parameters. In order to switch terramechanics models, wheel–terrain interactions are divided into three classes. Three vibration models of wheels under three WTICs have been built and analyzed. Vibration features in the models are extracted and non-dimensionalized to be independent of wheel speed. A vibration-feature-based recognition method of the WTIC is proposed. Then, the terrain parameters of the terramechanics model corresponding to the recognized WTIC are identified. Experiment results obtained using a Planetary Rover Prototype show that the identification method of terrain parameters is effective for rovers traversing various terrains. The relative errors of estimated wheel–terrain interaction force with identified terrain parameters are less than 16%, 12%, and 9% for rovers traversing hard, gravel, and sandy terrain, respectively. [ABSTRACT FROM AUTHOR]

The choice of an incorrect terrain classification might lead to consequences in construction costs, design speed, or even safety. However, the current design criteria for terrain classification may be highly subjective. In Brazil, design guidelines use textual descriptors for three classes, namely level, rolling, and mountainous. This study proposes a fuzzy rule-based classifier to predict terrain classes based on average slope and slope variation. The classifier uses fuzzy logic, which can account for imprecise and vague definitions of the input variables. The classifier was built using topographic variables, i.e. slope variation and average slope, and experts' knowledge. A survey was considered to extract experts' opinions regarding different terrain classes. The classifier provided an accuracy of at least 75%, which suggests that the expert system captured the experts' perceptions of the highway classes. As a result, the proposed system can assist decision-making by providing a more consistent method for terrain classification. [ABSTRACT FROM AUTHOR]

Copyright of Guangdong Architecture Civil Engineering is the property of Guangdong Architecture Civil Engineering Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Geomorphodiversity refers to the variety of landforms and morphological processes characterizing the landscape. The definition of an index to quantify geomorphodiversity is a relevant step for multiple fields of Earth sciences, since it is widely accepted that the variability of the geosphere deeply influences the diversity of the biosphere. Such an index should describe the number and type of landforms and geomorphological processes. We propose a quantitative land surface diversity index valid for Italy, considering multiple input quantities to describe geological constraints and geomorphological processes. Critical issues were the selection of moving window size for focal statistics operations, to calculate local diversity of slope, lithology, drainage density and terrain forms in individual raster maps. We compared the index with traditional geomorphological maps, in selected locations, in which information was available. Results show that a minimal set of heterogeneous data is a satisfactory approach to investigate the landscape diversity. Relating processing parameters and terrain spatial characteristics to the dataset resolution is a good choice to assess a reproducible land surface diversity index. Inclusion of drainage density allows the improvement of results in flat areas, in which other factors show trivial results. We argue that the index is relevant for land use management, assessment of ecodiversity, and it may help describing the interaction between abiotic and biotic compartments. [ABSTRACT FROM AUTHOR]

In this work, a single Semi-Recurrent Spike Neural Network (SRSNN) supervised learning method based on time coding is proposed to classify visual terrain encountered by the mobile robot. To this end, the features are extracted using the Local Binary Pattern (LBP) method. Then, the SRSNN is trained to classify multi-class of terrains. This training is used to classify six classes of terrain: hydrop, gravel, asphalt, grass, mud, and sand. The proposed training algorithm is based on adaptive synaptic weights that reach the threshold value. The feature extracted method and SRSNN form the proposed Intelligent structure. This structure effectively evaluates the accuracy, precision, recall, and F1 score. The simulation results achieve good performance in minimizing the mean square error in the training phase and maximizing the overall accuracy to 87.22%, especially in dangerous terrain (i.e. hydrop). The effectiveness of the proposed model is proved by the efficiency of the training algorithm which can learn fast with accurate results.

Polarimetric synthetic aperture radar (PolSAR) is widely used in remote sensing applications due to its ability to obtain full-polarization information. Compared to the quad-pol SAR, the dual-pol SAR mode has a wider observation swath and is more common in most SAR systems. The goal of reconstructing quad-pol SAR data from the dual-pol SAR mode is to learn the contextual information of dual-pol SAR images and the relationships among polarimetric channels. This work is dedicated to addressing this issue, and a multiscale feature aggregation network has been established to achieve the reconstruction task. Firstly, multiscale spatial and polarimetric features are extracted from the dual-pol SAR images using the pretrained VGG16 network. Then, a group-attention module (GAM) is designed to progressively fuse the multiscale features extracted by different layers. The fused feature maps are interpolated and aggregated with dual-pol SAR images to form a compact feature representation, which integrates the high- and low-level information of the network. Finally, a three-layer convolutional neural network (CNN) with a 1 × 1 convolutional kernel is employed to establish the mapping relationship between the feature representation and polarimetric covariance matrices. To evaluate the quad-pol SAR data reconstruction performance, both polarimetric target decomposition and terrain classification are adopted. Experimental studies are conducted on the ALOS/PALSAR and UAVSAR datasets. The qualitative and quantitative experimental results demonstrate the superiority of the proposed method. The reconstructed quad-pol SAR data can better sense buildings' double-bounce scattering changes before and after a disaster. Furthermore, the reconstructed quad-pol SAR data of the proposed method achieve a 97.08% classification accuracy, which is 1.25% higher than that of dual-pol SAR data. [ABSTRACT FROM AUTHOR]

Abstract Time-averaged shear-wave velocity to 30 m (V S30) is commonly used in ground motion models as a parameter for evaluating site effects. This study used a collection of boreholes in Beijing, Tianjin, Guangxi, Guangdong, and three other municipalities and provinces, which were divided into three regions with reference to the seismic ground motion parameter zonation map of China, to establish V S30 prediction models based on terrain categories. Regional effects were verified by comparing morphometric parameter (topographic slope, surface texture, and local convexity) thresholds and terrain classification maps obtained from global digital elevation model (DEM) data and regional DEM data of the three regions. Additionally, V S30 prediction models for the three regions using both types of terrain classification maps were established and analyzed comparatively to provide credible regional V S30 models for China. Through analysis of the correlations between the measured V S30 values and the predicted V S30 values, calculation of the mean squared error and mean absolute percentage error in each region, and with consideration of the geological characteristics of the boreholes, the V S30 prediction models based on terrain classification maps from regional data were finally applied in developing regional V S30 models for China. Intercomparison of the V S30 prediction models for the three regions indicated that subregional consideration is necessary in terrain classification. Finally, a spatial analysis method adopting inverse distance weighting of the residuals was used to update the initial V S30 models. The developed V S30 models could be used both in developing regional ground motion models and in the construction of earthquake disaster scenarios. Graphical Abstract

The significant differences in data domains between SAR images and the expensive and time-consuming process of data labeling pose significant challenges to terrain classification. Current terrain classification methodologies face challenges in addressing domain disparities and detecting uncommon terrain effectively. Based on Style Transformation and Domain Metrics (STDMs), we propose an unsupervised domain adaptive framework named STDM-UDA for terrain classification in this paper, which consists of two steps: image style transfer and domain adaptive segmentation. As a first step, image style transfer is performed within the image space to mitigate the differences in low-level features between SAR image domains. Subsequently, leveraging this process, the segmentation network extracts image features, employing domain metrics and adversarial training to enhance alignment between domain gaps in the semantic feature space. Finally, experiments conducted on several pairs of SAR images, each exhibiting varying degrees of differences in key imaging parameters such as source, resolution, band, and polarization, demonstrate the robustness of the proposed method. It achieves remarkably competitive classification accuracy, particularly for unlabeled, high-resolution broad scenes, effectively overcoming the domain gaps introduced by the diverse imaging parameters under studies.

Effective features play an important role in synthetic aperture radar (SAR) image interpretation. However, since SAR images contain a variety of terrain types, it is not easy to extract effective features of different terrains from SAR images. Deep learning methods require a large amount of labeled data, but the difficulty of SAR image annotation limits the performance of deep learning models. SAR images have inevitable geometric distortion and coherence speckle noise, which makes it difficult to extract effective features from SAR images. If effective semantic context features cannot be learned for SAR images, the extracted features struggle to distinguish different terrain categories. Some existing terrain classification methods are very limited and can only be applied to some specified SAR images. To solve these problems, a jigsaw puzzle self-supervised learning (JPSSL) framework is proposed. The framework comprises a jigsaw puzzle pretext task and a terrain classification downstream task. In the pretext task, the information in the SAR image is learned by completing the SAR image jigsaw puzzle to extract effective features. The terrain classification downstream task is trained using only a small number of labeled data. Finally, fully connected conditional random field processing is performed to eliminate noise points and obtain a high-quality terrain classification result. Experimental results on three large-scene high-resolution SAR images confirm the effectiveness and generalization of our method. Compared with the supervised methods, the features learned in JPSSL are highly discriminative, and the JPSSL achieves good classification accuracy when using only a small amount of labeled data.

Real-time gait event detection (GED) system can be utilized for gait analysis and tracking fitness activities. GED for various types of terrains (e.g., stair-walk, uneven surfaces, etc.) is still an open research problem. This study presents an inertial sensor-based approach for real-time GED system that works for diverse terrains in an uncontrolled environment. The GED system classifies three types of terrains, i.e., flat-walk, stair-ascend and stair-descend, with an average classification accuracy of 99%. It also accurately detects various gait events, including, toe-strike, heel-rise, toe-off, and heel-strike. It is computationally efficient, implemented on a low-cost microcontroller, works in real-time and can be used in portable rehabilitation devices for use in dynamic environments.

• Develop a dataset of Mars scenes from numerous raw Mars images. • Compare performance of three common deep learning models to select the appropriate one. • Conduct transfer learning and data augmentation to enhance selected model's generation capacity. • Deep learning can realize autonomous, real-time Martian terrain classification. Scene information plays a crucial role in motion control, attitude perception, and path planning for wheeled planetary rovers (WPRs). Terrain recognition is the fundamental component of scene recognition. Due to the rich information, visual sensors are usually used in terrain classification. However, teleoperation delay prevents WPRs from using visual information efficiently. End-to-end learning method of deep learning (DL) that does not need complex image preprocessing was proposed to deal with this issue. This paper first built a terrain dataset (consists of loose sand, bedrock, small rock, large rock, and outcrop) using real Mars images to directly support You Only Look Once (YOLOv5) to test its performance on terrain classification. Because the capability of end-to-end training scheme is positively correlated with dataset, the performance of YOLOv5 can be significantly improved by exploiting orders of magnitude more data. The best combination of hyperparameters and models was achieved by slightly tuning YOLOv5, and data augmentation was also applied to optimize its accuracy. Furthermore, its performance was compared with two other end-to-end network architectures. Deep learning algorithms can be used in the future planetary exploration missions, such as WPRs autonomy improvement, traversability analysis, and avoiding getting trapped. [ABSTRACT FROM AUTHOR]

Terrain classification provides valuable information for both control and navigation algorithms of wheeled mobile robots. In this paper, a novel online outdoor terrain classification algorithm is proposed for wheeled mobile robots. The algorithm is based on only time-domain features with both low computational and low memory requirements, which are extracted from the inertial and magnetic sensor signals. Multilayer perceptron (MLP) neural networks are applied as classifiers. The algorithm is tested on a measurement database collected using a prototype measurement system for various outdoor terrain types. Different datasets were constructed based on various setups of processing window sizes, used sensor types, and robot speeds. To examine the possibilities of the three applied sensor types in the application, the features extracted from the measurement data of the different sensors were tested alone, in pairs and fused together. The algorithm is suitable to operate online on the embedded system of the mobile robot. The achieved results show that using the applied time-domain feature set the highest classification efficiencies on unknown data can be above 98%. It is also shown that the gyroscope provides higher classification rates than the widely used accelerometer. The magnetic sensor alone cannot be effectively used but fusing the data of this sensor with the data of the inertial sensors can improve the performance. [ABSTRACT FROM AUTHOR]

For terrain recognition needs during vehicle driving, this paper carries out terrain classification research based on vibration and image information. Twenty time-domain features and eight frequency-domain features of vibration signals that are highly correlated with terrain are selected, and principal component analysis (PCA) is used to reduce the dimensionality of the time-domain and frequency-domain features and retain the main information. Meanwhile, the texture features of the terrain images are extracted using the gray-level co-occurrence matrix (GLCM) technique, and the feature information of the vibration and images are fused in the feature layer. Then, the improved weighted K-nearest neighbor (WKNN) algorithm is used to achieve the terrain classification during the travel process of tracked robots. Finally, the experimental results verify that the proposed method improves the terrain classification accuracy of the tracked robot and provides a basis for improving the stable autonomous driving of tracked vehicles. [ABSTRACT FROM AUTHOR]

With the development of science and technology, although more and more polarimetric synthetic aperture radar (PolSAR) data are collected, marking PolSAR data still requires a lot of costs. Moreover, the datasets between different domains have the class distribution shift problem, which reduces the reusability of labeled samples between cross-domain images. To address this issue, this letter proposed an unsupervised domain adaptation (UDA) network based on coordinate attention (CA) and weighted clustering. First, an adversarial UDA network with a biclassifier is introduced to eliminate the problem of class distribution shift and achieve alignment of data distribution between different domains. Second, the CA mechanism is introduced to select important features to enhance the utilization of spatial information among pixels. Finally, to improve the utilization of semantic and classification information of the target domain, and to align same class samples, a weighted clustering algorithm is introduced. Experimental results show that compared with the existing UDA method, the proposed method can achieve better PolSAR image classification. [ABSTRACT FROM AUTHOR]

Computer vision can provide upcoming walking environment information for lower limb-assisted robots, thereby enabling more accurate and robust decisions for high-level control. However, current computer vision systems in lower extremity devices are still constrained by the disruptions that occur in the interaction between human, machine, and the environment, which hinder optimal performance. In this paper, we propose a gimbal-based terrain classification system that can be adapted to different lower limb movements, different walking speeds, and gait phases. We use a linear active disturbance rejection controller to realize fast response and anti-disturbance control of the gimbal, which allows computer vision to continuously and stably focus on the desired field of view angle during lower limb motion interaction. We also deployed a lightweight MobileNetV2 model in an embedded vision module for real-time and highly accurate inference performance. By using the proposed terrain classification system, it can provide the ability to classify and predict terrain independent of mounting position (thighs and shanks), gait phase, and walking speed. This also makes our system applicable to subjects with different physical conditions (e.g., non-disabled subjects and individuals with transfemoral amputation) without tuning the parameters, which will contribute to the plug-and-play functionality of terrain classification. Finally, our approach is promising to improve the adaptability of lower limb assisted robots in complex terrain, allowing the wearer to walk more safely.

Time-averaged shear-wave velocity to 30 m (VS30) is commonly used in ground motion models as a parameter for evaluating site effects. This study used a collection of boreholes in Beijing, Tianjin, Guangxi, Guangdong, and three other municipalities and provinces, which were divided into three regions with reference to the seismic ground motion parameter zonation map of China, to establish VS30 prediction models based on terrain categories. Regional effects were verified by comparing morphometric parameter (topographic slope, surface texture, and local convexity) thresholds and terrain classification maps obtained from global digital elevation model (DEM) data and regional DEM data of the three regions. Additionally, VS30 prediction models for the three regions using both types of terrain classification maps were established and analyzed comparatively to provide credible regional VS30 models for China. Through analysis of the correlations between the measured VS30 values and the predicted VS30 values, calculation of the mean squared error and mean absolute percentage error in each region, and with consideration of the geological characteristics of the boreholes, the VS30 prediction models based on terrain classification maps from regional data were finally applied in developing regional VS30 models for China. Intercomparison of the VS30 prediction models for the three regions indicated that subregional consideration is necessary in terrain classification. Finally, a spatial analysis method adopting inverse distance weighting of the residuals was used to update the initial VS30 models. The developed VS30 models could be used both in developing regional ground motion models and in the construction of earthquake disaster scenarios. [ABSTRACT FROM AUTHOR]

Deep learning and convolutional neural networks (CNN) have been widely applied in polarimetric synthetic aperture radar (PolSAR) image classification, and satisfactory results have been obtained. However, there is one crucial issue that still has not been solved. These methods require abundant labeled samples and obtaining the labeled samples of PolSAR images is usually time-consuming and labor-intensive. To obtain better classification results with fewer labeled samples, a new attention-based 3D residual relation network (3D-ARRN) is proposed for PolSAR image. Firstly, a multilayer CNN with residual structure is used to extract depth polarimetric features. Secondly, to extract more important feature information and improve the classification results, a spatial weighted attention network (SWANet) is introduced to concentrate the feature information, which is more favorable for a classification task. Then, the features of training and test samples are integrated and CNN is utilized to compute the score of similarity between training and test samples. Finally, the similarity score is used to determine the category of test samples. Studies on four different PolSAR datasets illustrate that the proposed 3D-ARRN model can achieve higher classification results than other comparison methods with few labeled data. [ABSTRACT FROM AUTHOR]

Terrain classification information is of great significance for legged robots to traverse various terrains. Therefore, this communication presents an online terrain classification framework for legged robots, utilizing the acoustic signals produced during locomotion. The Mel-Frequency Cepstral Coefficient (MFCC) feature vectors are extracted from the acoustic data recorded by an on-board microphone. Then the Gaussian mixture models (GMMs) are used to classify the MFCC features into different terrain type categories. The proposed framework was validated on a quadruped robot. Overall, our investigations achieved a classification time-resolution of 1 s when the robot trotted over three kinds of terrains, thus recording a comprehensive success rate of 92.7%.

Identifying terrain parameters is important for high-fidelity simulation and high-performance control of planetary rovers. The wheel–terrain interaction classes (WTICs) are usually different for rovers traversing various types of terrain. Every terramechanics model corresponds to its wheel–terrain interaction class (WTIC). Therefore, for terrain parameter identification of the terramechanics model when rovers traverse various terrains, terramechanics model switching corresponding to the WTIC needs to be solved. This paper proposes a speed-independent vibration-based method for WTIC recognition to switch the terramechanics model and then identify its terrain parameters. In order to switch terramechanics models, wheel–terrain interactions are divided into three classes. Three vibration models of wheels under three WTICs have been built and analyzed. Vibration features in the models are extracted and non-dimensionalized to be independent of wheel speed. A vibration-feature-based recognition method of the WTIC is proposed. Then, the terrain parameters of the terramechanics model corresponding to the recognized WTIC are identified. Experiment results obtained using a Planetary Rover Prototype show that the identification method of terrain parameters is effective for rovers traversing various terrains. The relative errors of estimated wheel–terrain interaction force with identified terrain parameters are less than 16%, 12%, and 9% for rovers traversing hard, gravel, and sandy terrain, respectively.

Abstract Global terrain classification data have been used for various issues related to topography such as the estimation of soil types and of ground vulnerability to earthquakes and the creation of seismic hazard maps. However, due to the resolution of digital elevation models (DEMs), the terrain classification data from previous studies could not discriminate small landforms such as plains at the bottom of narrow valleys and small rises in plains. Owing to the greater regional variation of small landforms, there is trade-off between DEMs of higher resolution and the creation of global geomorphological legends. To address this problem, we first merged regions with similar topographic characteristics using slope gradients and HAND (height above the nearest drainage) calculated by the 90-m-spatial-resolution DEMs interpolated from the multi-error-removed improved-terrain DEM (MERIT DEM), and united the polygons with the unit catchments of the MERIT-Basins dataset, so that the polygons contain calculated terrain measurements (slope gradient, HAND, surface texture, local convexity, sinks) and noise types as attributes, as well as the ID number of the unit catchment. In addition, we performed k-means clustering on the dataset using slope gradient, HAND, and surface texture, which can be combined with the dataset as a simple terrain classification. The clustering results were prepared in 15 and 40 global uniform clusters and 15 and 40 clusters for each basin to understand the global appearance of the terrain and provide zoning data for regional problem-solving. The 15 clusters were prepared to observe the outline of the terrain without any processing, whereas the 40 clusters were prepared to group and reclassify the polygons to create zoning data for each region. This dataset showed improvements in terms of capturing the small rises in plains compared to the authors' previous global terrain classification data. This dataset can be used as a proxy and is expected to contribute to modeling and estimation in various applications that are known to be related to topography. The datasets of this article are available at [ https://gisstar.gsi.go.jp/terrain2021/ ].

• Processing Complex-valued PolSAR Data Using Complex-valued Convolutional Neural Network (CV-CNN). • Uses a Gaussian-type activation function (GTAF) that preserves the integrity of complex-valued operations. • Introduces learnable Gaussian parameters for GTAF, and designs two multi-channel activation methods. • The classification accuracy is better than that of existing state-of-the-art methods in three datasets. To process complex-valued information such as SAR signals conveniently, the complex-valued convolutional neural network (CV-CNN) has been proposed in recent years, and it has achieved great success in SAR image recognition. This paper proposes an activation function with learnable parameters based on the Gaussian-type activation function (GTAF) in CV-CNN to improve the utilization of information in the real and imaginary parts of the neuro. For the multi-channel input of the feature map, this paper discusses two ways to set the parameters of the Gaussian-type activation function. One is that all channels share the same parameters, called the channel-sharing Gaussian-type activation function (CSGTAF). The other is that each channel has its independent parameters, called the channel-exclusive Gaussian-type activation function (CEGTAF). In addition, this paper derives the backpropagation formula of both CSGTAF and CEGTAF in detail for the training process of CV-CNN. This paper performs experimental analysis on three L-band standard PolSAR datasets. The experimental results show that, compared with the traditional method and the Gaussian activation function with fixed parameters, both CSGTAF and CEGTAF achieve higher recognition accuracy, and the difference in the recognition effect of different targets in the same dataset is little. Both show good recognition performance and have good stability and versatility. [ABSTRACT FROM AUTHOR]

Slope units represent surface slopes by means of polygons delimited by drainage and divide lines obtained on a digital topography. Objective slope unit delineation for a given digital elevation model is still an open issue and, often, a limitation that may dictate the use of a more traditional pixel-based approach for spatial analysis. Availability of slope unit maps facilitates many kinds of studies and allows scholars to focus on specific scientific issues rather than on preparing sound mapping units from scratch for their research. Here, we present a slope unit map of a large portion of the Himalayas. The map is prepared following a widely tested, parameter-free optimization algorithm. The area encompassed by the map is relevant to studies of the well-known 2015 Gorkha earthquake and monsoons, which makes it relevant to a vast portion of the scientific community working in natural hazards including, but not limited to, landslide scientists and practitioners. The map contains 112,674 polygons with average area of 0.38 km[Formula: see text] and is published in vector form. The map is accompanied by a selection of data including morphometric and thematic quantities. In addition to describing the rationale behind the delineation of the polygonal map and selected data, we describe an application devoted to unsupervised terrain classification. We applied a k-means clustering procedure with two strategies: one at (coarser) basin scale and one at (finer) slope unit scale. We show similarities and differences between the two classification strategies, highlighting the role of the slope unit subdivision in the two cases.

Hyperspectral remote sensing has been an important technical means to obtain more refined information and provide rich, accurate, and reasonable data for the quantitative analysis and delicacy management of a ''smart city''. To better understand and use the hyperspectral data to help the construction of a digital city, the study of the feature and characteristics of hyperspectral remote sensing images is introduced in this paper. Then how to collect the hyperspectral information of urban ground objects through the unmanned aerial vehicle (UAV) and hyperspectral imager was described, which greatly improves the efficiency of urban data acquisition. Finally, various application cases of UAV-based hyperspectral remote sensing and deep information mining of urban ground objects were analyzed and discussed in detail, such as terrain classification, urban greening analysis, etc. The research result shows that airborne hyperspectral imagery (HIS) has unique advantages over color photography and multispectral remote sensing, with a richer and higher level of spectral details and physical & chemical properties.

Recently, convolutional neural networks (CNNs) have been successfully utilized in polarimetric synthetic aperture radar (PolSAR) image classification and obtained promising results. However, most CNN-based classification methods require a large number of labeled samples and it is difficult to obtain sufficient labeled samples. For this reason, an ensemble dual-branch CNN (EDb-CNN) is proposed for PolSAR image classification with small samples. First, to solve the problem of the small sample in PolSAR image classification, a new data enhancement method based on the superpixel algorithm is proposed to expand the number of labeled samples. Second, to obtain different scales of features from PolSAR images, a Db-CNN model is proposed. This model contains two parallel CNN structures. One CNN branch is used to extract the polarization features from the complex coherency matrix. The other branch is utilized to extract the spatial features based on weighted spatial neighborhood. On the top of these two branches, a feature fusion model is adopted to combine these two deep features, and a weighted loss function is employed to improve the learning procedure. Then, the ensemble learning algorithm is used for each CNN branch and Db-CNN network to obtain the better classification results. Finally, a postprocess algorithm based on the superpixel algorithm is proposed to improve the consistency of classification results. Experiments on two PolSAR datasets show that the proposed method achieves a much better performance than other classification methods, especially when only a few labeled samples are available.

“Camouflage” is a natural or nature identical phenomenon where the sensory route of vision is delayed to avoid visual detection. Reducing detection capability and hiding in the background environment is critical for Army vehicles, equipment, and soldiers. This research aims to implement a process that will generate digital camouflage patterns specific to the terrain. The adapted digital pattern helps an object blend symmetrically into the background environment. Pixelated textures combine macro and micro designs that blend with ambient shrubs, trees, branches, and shadows. The technique presented in this paper consists of the following main modules: terrain classification model, pixelated camouflage texture generation, and texture evaluation. Experiments have been conducted to detect camouflage objects in the scene to evaluate the performance of the resultant camouflage texture generated for a natural environment. Photo simulation and saliency maps for hidden object detection have been used to evaluate the effectiveness of generated textures for a given terrain. [ABSTRACT FROM AUTHOR]

It is important for Mars exploration rovers to achieve autonomous and safe mobility over rough terrain. Terrain classification can help rovers to select a safe terrain to traverse and avoid sinking and/or damaging the vehicle. Mars terrains are often classified using visual methods. However, the accuracy of terrain classification has been less than 90% in read operations. A high-accuracy vision-based method for Mars terrain classification is presented in this paper. By analyzing Mars terrain characteristics, novel image features, including multiscale gray gradient-grade features, multiscale edges strength-grade features, multiscale frequency-domain mean amplitude features, multiscale spectrum symmetry features, and multiscale spectrum amplitude-moment features, are proposed that are specifically targeted for terrain classification. Three classifiers, K-nearest neighbor (KNN), support vector machine (SVM), and random forests (RF), are adopted to classify the terrain using the proposed features. The Mars image dataset MSLNet that was collected by the Mars Science Laboratory (MSL, Curiosity) rover is used to conduct terrain classification experiments. The resolution of Mars images in the dataset is 256 × 256. Experimental results indicate that the RF classifies Mars terrain at the highest level of accuracy of 94.66%. [ABSTRACT FROM AUTHOR]

Polarimetric synthetic aperture radar (PolSAR) is widely used in remote sensing applications due to its ability to obtain full-polarization information. Compared to the quad-pol SAR, the dual-pol SAR mode has a wider observation swath and is more common in most SAR systems. The goal of reconstructing quad-pol SAR data from the dual-pol SAR mode is to learn the contextual information of dual-pol SAR images and the relationships among polarimetric channels. This work is dedicated to addressing this issue, and a multiscale feature aggregation network has been established to achieve the reconstruction task. Firstly, multiscale spatial and polarimetric features are extracted from the dual-pol SAR images using the pretrained VGG16 network. Then, a group-attention module (GAM) is designed to progressively fuse the multiscale features extracted by different layers. The fused feature maps are interpolated and aggregated with dual-pol SAR images to form a compact feature representation, which integrates the high- and low-level information of the network. Finally, a three-layer convolutional neural network (CNN) with a 1 × 1 convolutional kernel is employed to establish the mapping relationship between the feature representation and polarimetric covariance matrices. To evaluate the quad-pol SAR data reconstruction performance, both polarimetric target decomposition and terrain classification are adopted. Experimental studies are conducted on the ALOS/PALSAR and UAVSAR datasets. The qualitative and quantitative experimental results demonstrate the superiority of the proposed method. The reconstructed quad-pol SAR data can better sense buildings’ double-bounce scattering changes before and after a disaster. Furthermore, the reconstructed quad-pol SAR data of the proposed method achieve a 97.08% classification accuracy, which is 1.25% higher than that of dual-pol SAR data.

For terrain recognition needs during vehicle driving, this paper carries out terrain classification research based on vibration and image information. Twenty time-domain features and eight frequency-domain features of vibration signals that are highly correlated with terrain are selected, and principal component analysis (PCA) is used to reduce the dimensionality of the time-domain and frequency-domain features and retain the main information. Meanwhile, the texture features of the terrain images are extracted using the gray-level co-occurrence matrix (GLCM) technique, and the feature information of the vibration and images are fused in the feature layer. Then, the improved weighted K-nearest neighbor (WKNN) algorithm is used to achieve the terrain classification during the travel process of tracked robots. Finally, the experimental results verify that the proposed method improves the terrain classification accuracy of the tracked robot and provides a basis for improving the stable autonomous driving of tracked vehicles.

Global terrain classification data have been used for various issues related to topography such as the estimation of soil types and of ground vulnerability to earthquakes and the creation of seismic hazard maps. However, due to the resolution of digital elevation models (DEMs), the terrain classification data from previous studies could not discriminate small landforms such as plains at the bottom of narrow valleys and small rises in plains. Owing to the greater regional variation of small landforms, there is trade-off between DEMs of higher resolution and the creation of global geomorphological legends. To address this problem, we first merged regions with similar topographic characteristics using slope gradients and HAND (height above the nearest drainage) calculated by the 90-m-spatial-resolution DEMs interpolated from the multi-error-removed improved-terrain DEM (MERIT DEM), and united the polygons with the unit catchments of the MERIT-Basins dataset, so that the polygons contain calculated terrain measurements (slope gradient, HAND, surface texture, local convexity, sinks) and noise types as attributes, as well as the ID number of the unit catchment. In addition, we performed k-means clustering on the dataset using slope gradient, HAND, and surface texture, which can be combined with the dataset as a simple terrain classification. The clustering results were prepared in 15 and 40 global uniform clusters and 15 and 40 clusters for each basin to understand the global appearance of the terrain and provide zoning data for regional problem-solving. The 15 clusters were prepared to observe the outline of the terrain without any processing, whereas the 40 clusters were prepared to group and reclassify the polygons to create zoning data for each region. This dataset showed improvements in terms of capturing the small rises in plains compared to the authors' previous global terrain classification data. This dataset can be used as a proxy and is expected to contribute to modeling and estimation in various applications that are known to be related to topography. The datasets of this article are available at [https://gisstar.gsi.go.jp/terrain2021/]. [ABSTRACT FROM AUTHOR]

Increasing the terrain awareness of planetary exploration rovers is one key technology for future space robotics to successfully accomplish long‐distance and long‐duration missions. In contrast to most of the existing algorithms that use visual or depth data for terrain classification, the approach presented in this study tackles the problem using proprioceptive sensing, for example, vibration or force measurements. The underlying assumption is that these signals, being directly modulated by the terrain properties, are well descriptive of a given surface. Therefore, terrain signature can be inferred via learning algorithms that are trained on either the signals directly or a signal‐derived feature set. Following the latter approach, first, a physics‐based signal augmentation process is presented that aims at maximizing the information content. Then, a feature selection algorithm based on a scoring system and an iterative search is developed to decrease the computational cost while preserving high classification accuracy. The resulting most informative feature subspace can be used to train a support vector machine (SVM) classifier. For comparison, the time histories of the selected proprioceptive signals are used to train a deep convolutional neural network (CNN). Results obtained from real experiments using the SherpaTT rover confirm that proprioceptive sensing is effective in predicting terrain type with an accuracy higher than 90% for both algorithms in generalization tasks. When the two learning approaches are contrasted in extrapolation problems, for example, predicting observations acquired at previously unseen velocity or terrain, CNN outperforms the standard SVM. Furthermore, CNN holds the additional advantage of learning features automatically from signal spectrograms, reducing the need of a priori knowledge at the expense of higher computational efforts. [ABSTRACT FROM AUTHOR]

Hyperspectral imaging technology produces images that capture both spatial and spectral details of features, constituting a key data source in remote sensing detection. However, the accuracy of classifying hyperspectral remote sensing images is influenced by factors such as extensive data volume, high redundancy, and numerous spectral bands exhibiting significant correlation. To address these challenges, we introduce a two-layer with dual-branch 3D/2D-Attention cascade network model named DBANet. Firstly, the model employs principal component analysis to extract the raw data's primary features to reduce the data's complexity. Subsequently, the data undergoes two layers of 3D convolution and 2D convolution combined with a cascade structure of attention mechanism to fully extract the data's spatial, spectral, and textural feature information. Finally, the classification of input images is computed using a fully connected layer combined with the LogSoftmax function. Our DBANet demonstrates notable advantages in feature extraction and classification performance compared to other methods, as evidenced by the comprehensive qualitative and quantitative results of three public hyperspectral remote sensing datasets. It exhibits superior performance in the classification of hyperspectral remote-sensing images. [ABSTRACT FROM AUTHOR]

Deep learning and convolutional neural networks (CNN) have been widely applied in polarimetric synthetic aperture radar (PolSAR) image classification, and satisfactory results have been obtained. However, there is one crucial issue that still has not been solved. These methods require abundant labeled samples and obtaining the labeled samples of PolSAR images is usually time-consuming and labor-intensive. To obtain better classification results with fewer labeled samples, a new attention-based 3D residual relation network (3D-ARRN) is proposed for PolSAR image. Firstly, a multilayer CNN with residual structure is used to extract depth polarimetric features. Secondly, to extract more important feature information and improve the classification results, a spatial weighted attention network (SWANet) is introduced to concentrate the feature information, which is more favorable for a classification task. Then, the features of training and test samples are integrated and CNN is utilized to compute the score of similarity between training and test samples. Finally, the similarity score is used to determine the category of test samples. Studies on four different PolSAR datasets illustrate that the proposed 3D-ARRN model can achieve higher classification results than other comparison methods with few labeled data.

Slope units represent surface slopes by means of polygons delimited by drainage and divide lines obtained on a digital topography. Objective slope unit delineation for a given digital elevation model is still an open issue and, often, a limitation that may dictate the use of a more traditional pixel-based approach for spatial analysis. Availability of slope unit maps facilitates many kinds of studies and allows scholars to focus on specific scientific issues rather than on preparing sound mapping units from scratch for their research. Here, we present a slope unit map of a large portion of the Himalayas. The map is prepared following a widely tested, parameter-free optimization algorithm. The area encompassed by the map is relevant to studies of the well-known 2015 Gorkha earthquake and monsoons, which makes it relevant to a vast portion of the scientific community working in natural hazards including, but not limited to, landslide scientists and practitioners. The map contains 112,674 polygons with average area of 0.38 km² and is published in vector form. The map is accompanied by a selection of data including morphometric and thematic quantities. In addition to describing the rationale behind the delineation of the polygonal map and selected data, we describe an application devoted to unsupervised terrain classification. We applied a k-means clustering procedure with two strategies: one at (coarser) basin scale and one at (finer) slope unit scale. We show similarities and differences between the two classification strategies, highlighting the role of the slope unit subdivision in the two cases. [ABSTRACT FROM AUTHOR]

Unmanned mobile robots in rough terrains are a key technology for achieving smart agriculture and smart construction. The mobility performance of robots highly depends on the moisture content of soil, and past few studies have focused on terrain classification using moisture content. In this study, we demonstrate a convolutional neural network-based terrain classification method using RGB-infrared (IR) images. The method first classifies soil types and then categorizes the moisture content of the terrain. A three-step image preprocessing for RGB-IR images is also integrated into the method that is applicable to an actual environment. An experimental study of the terrain classification confirmed that the proposed method achieved an accuracy of more than 99% in classifying the soil type. Furthermore, the classification accuracy of the moisture content was approximately 69% for pumice and 100% for dark soil. The proposed method can be useful for different scenarios, such as small-scale agriculture with mobile robots, smart agriculture for monitoring the moisture content, and earthworks in small areas. [ABSTRACT FROM AUTHOR]

Wide-area land subsidence occurred across the Kashiwazaki Plain, Niigata, Japan after the 2007 Niigata-ken Chuetsu-Oki Earthquake. The stratified ground subsidence meter clearly showed that a thick clayey layer had caused this subsidence after the earthquake. To examine the ground conditions, a boring survey was conducted at the site and the physical and mechanical properties of the ground were tested. Although leveling surveys are insufficient for surveying the behavior of the ground, the measurement data obtained at various points from these surveys are meritorious for grasping the spatial distribution and the time-dependent property of the land subsidence. In this study, by analyzing the instantaneous and the long-term amounts of land subsidence at each point from the time-series observation data, the correlation between geology and topography is reported and the cause of the land subsidence of the clayey soils after the earthquake is discussed. It is clarified that the existence of a marine clay layer greatly affected the long-term subsidence of the land. [ABSTRACT FROM AUTHOR]

To improve the accuracy of terrain classification during mobile robot operation, an adaptive online terrain classification method based on vibration signals is proposed. First, the time domain and the combined features of the time, frequency, and time–frequency domains in the original vibration signal are extracted. These are adopted as the input of the random forest algorithm to generate classification models with different dimensions. Then, by judging the relationship between the current speed of the mobile robot and its critical speed, the classification model of different dimensions is adaptively selected for online classification. Offline and online experiments are conducted for four different terrains. The experimental results show that the proposed method can effectively avoid the self-vibration interference caused by an increase in the robot's moving speed and achieve higher terrain classification accuracy. [ABSTRACT FROM AUTHOR]

It is important for Mars exploration rovers to achieve autonomous and safe mobility over rough terrain. Terrain classification can help rovers to select a safe terrain to traverse and avoid sinking and/or damaging the vehicle. Mars terrains are often classified using visual methods. However, the accuracy of terrain classification has been less than 90% in read operations. A high-accuracy vision-based method for Mars terrain classification is presented in this paper. By analyzing Mars terrain characteristics, novel image features, including multiscale gray gradient-grade features, multiscale edges strength-grade features, multiscale frequency-domain mean amplitude features, multiscale spectrum symmetry features, and multiscale spectrum amplitude-moment features, are proposed that are specifically targeted for terrain classification. Three classifiers, K-nearest neighbor (KNN), support vector machine (SVM), and random forests (RF), are adopted to classify the terrain using the proposed features. The Mars image dataset MSLNet that was collected by the Mars Science Laboratory (MSL, Curiosity) rover is used to conduct terrain classification experiments. The resolution of Mars images in the dataset is 256 × 256. Experimental results indicate that the RF classifies Mars terrain at the highest level of accuracy of 94.66%.