Article
Environmental Sciences
Yingpin Yang, Zhifeng Wu, Wenju Xiao, Ya'nan Zhou, Qiting Huang, Tianjun Wu, Jiancheng Luo, Haiyun Wang
Summary: This study developed a new method for mapping abandoned land based on spatiotemporal features extracted from PolSAR Single Look Complex images using deep learning methods. The results showed that the introduction of multitemporal polarimetric parameters and spatial features improved the accuracy of identifying abandoned land. The combination of backscattering features, polarimetric parameters, and spatial features yielded the best performance, with a producer's accuracy of 88.29% and a user's accuracy of 84.03%. This study demonstrated the potential of polarimetric parameters and validated the effectiveness of spatiotemporal features in identifying abandoned land, providing a practical method for reliable mapping in cloudy and rainy areas.
Article
Environmental Sciences
Pengxiang Zhao, Zohreh Masoumi, Maryam Kalantari, Mahtab Aflaki, Ali Mansourian
Summary: This study conducted a GIS-based landslide susceptibility mapping in Zanjan, Iran, and compared different machine learning algorithms. The results show that random forest algorithm achieved the best performance, while slope and topographic curvature are identified as the most important causative factors.
Review
Geosciences, Multidisciplinary
Songlin Liu, Luqi Wang, Wengang Zhang, Yuwei He, Samui Pijush
Summary: Landslide susceptibility mapping (LSM) is crucial for development and construction planning to reduce the socio-economic impact of landslides. The use of machine learning algorithms and big data has greatly improved mapping accuracy and efficiency in this field.
GEOLOGICAL JOURNAL
(2023)
Review
Geochemistry & Geophysics
Yansi Chen, Yunchen Wang, Feng Zhang, Yulong Dong, Zhihong Song, Genyuan Liu
Summary: Remote sensing technology has played a significant role in geological exploration and mineral resource assessment. This paper provides a comprehensive overview of the challenges and opportunities in remote sensing-based lithological identification in vegetated regions. It reviews prior research, remote sensing data sources, and classification methodologies. The paper also addresses limitations and proposes promising avenues for future research, including the integration of multi-source data and exploration of novel remote sensing techniques and algorithms.
Article
Chemistry, Multidisciplinary
Jianhai Jin, Yuhuang Ye, Xiaohe Li, Liang Li, Min Shan, Jun Sun
Summary: In this work, a deep-learning-based mapping model is proposed for simulating and predicting the flow field of propellers using Reynolds-averaged Navier-Stokes (RANS) and large eddy simulation (LES). The model leverages image processing and computer vision techniques to process the two-dimensional propeller RANS and LES simulation data. By using a deep convolutional neural network (CNN) for feature extraction and a nonlinear module for regression and mapping, the proposed model demonstrates high accuracy and effectiveness in flow field prediction. The model's generalization ability, stability, and robustness are also evaluated and verified.
APPLIED SCIENCES-BASEL
(2023)
Article
Environmental Sciences
Emre Gulher, Ugur Alganci
Summary: Satellite-derived bathymetry (SDB) is a process to estimate water depth in shallow coastal and inland waters using satellite imagery. Recent advancements in technology and data processing have improved the accuracy and availability of SDB. This study aims to create an SBD map of Horseshoe Island using optical satellite images and compare the performance of different models and atmospheric correction methods. Machine learning-based models, specifically random forest and XGBoost, provided the highest performance and best fitting ability, followed by deep learning-based models. Landsat 8 performed better for deeper depths, while Sentinel 2 was slightly better for shallower depths. ACOLITE, iCOR, and ATCOR all produced reliable results, with ACOLITE offering the highest level of automation.
Article
Computer Science, Artificial Intelligence
Kyungjune Baek, Duhyeon Bang, Hyunjung Shim
Summary: This paper introduces a method called GridMix for improving network generalization by predicting patch-level labels and utilizing grid-based mixing for local data augmentation. Experimental results show that GridMix outperforms state-of-the-art techniques in classification and adversarial robustness, achieving comparable performance in weakly supervised object localization.
PATTERN RECOGNITION
(2021)
Review
Computer Science, Artificial Intelligence
Xiaoqiang Yan, Shizhe Hu, Yiqiao Mao, Yangdong Ye, Hui Yu
Summary: This paper provides a comprehensive review of deep MVL, covering methods in deep learning scope and extensions of traditional methods. It reviews representative MVL methods in deep learning like multi-view auto-encoder, and investigates advancements when traditional methods meet deep learning models. The paper also summarizes main applications, datasets, performance comparisons, and identifies open challenges for future research.
Article
Mechanics
Behrooz Ghorbani, Song Mei, Theodor Misiakiewicz, Andrea Montanari
Summary: For certain classification tasks, RKHS methods can replace NNs without a large loss in performance, but in special examples, NNs trained with SGD can outperform RKHS, especially when covariates display the same low-dimensional structure.
JOURNAL OF STATISTICAL MECHANICS-THEORY AND EXPERIMENT
(2021)
Article
Remote Sensing
Diogo Nunes Goncalves, Jose Marcato Junior, Andre Caceres Carrilho, Plabiany Rodrigo Acosta, Ana Paula Marques Ramos, Felipe David Georges Gomes, Lucas Prado Osco, Maxwell da Rosa Oliveira, Jose Augusto Correa Martins, Geraldo Alves Damasceno Junior, Marcio Santos de Araujo, Jonathan Li, Fabio Roque, Leonardo de Faria Peres, Wesley Nunes Goncalves, Renata Libonati
Summary: Pantanal, the world's largest continuous wetland, is facing endangerment to its biodiversity due to catastrophic wildfires in recent years. To improve measurements and assist in environmental actions, robust methods using high-spatial-resolution imagery are needed to map burned areas. This study combines Transformer-based deep learning methods with high-resolution planet imagery to accurately map burned areas in the Brazilian Pantanal wetland.
INTERNATIONAL JOURNAL OF APPLIED EARTH OBSERVATION AND GEOINFORMATION
(2023)
Review
Biochemical Research Methods
Jing Wang, Qinglong Zhang, Junshan Han, Yanpeng Zhao, Caiyun Zhao, Bowei Yan, Chong Dai, Lianlian Wu, Yuqi Wen, Yixin Zhang, Dongjin Leng, Zhongming Wang, Xiaoxi Yang, Song He, Xiaochen Bo
Summary: This review article discusses the recent applications of computational methods in synthetic lethality (SL) prediction. It introduces the concept and screening methods of SL, summarizes various SL-related data resources, and provides an overview of computational methods including statistical-based methods, network-based methods, classical machine learning methods, and deep learning methods for SL prediction. The article also highlights the use of negative sampling methods in these models. Representative tools for SL prediction are introduced, and the challenges and future work for SL prediction are discussed.
BRIEFINGS IN BIOINFORMATICS
(2022)
Review
Biochemical Research Methods
Karin Elimelech-Zohar, Yaron Orenstein
Summary: Nucleic-acid G-quadruplexes (G4s) are crucial in cellular processes, and experimental assays have been developed to measure them in high throughput. This has enabled the development of machine-learning-based methods, particularly deep neural networks, to predict G4s in any nucleic-acid sequence and species.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Biochemical Research Methods
Md Rezaul Karim, Tanhim Islam, Md Shajalal, Oya Beyan, Christoph Lange, Michael Cochez, Dietrich Rebholz-Schuhmann, Stefan Decker
Summary: Artificial intelligence (AI) systems are widely used for solving critical problems in bioinformatics, biomedical informatics, and precision medicine. However, the lack of transparency in complex AI models can be a challenge in understanding their decision-making processes. Explainable AI (XAI) aims to provide transparency and fairness in AI systems, which is particularly important in sensitive areas like healthcare. This paper discusses the importance of explainability in bioinformatics and showcases model-specific and model-agnostic interpretable ML methods that can be customized for bioinformatics research problems. Through case studies, the authors demonstrate how XAI methods can improve transparency and decision fairness in bioinformatics.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Agriculture, Multidisciplinary
Helizani Couto Bazame, Jose Paulo Molin, Daniel Althoff, Mauricio Martello, Lucas De Paula Corredo
Summary: This study implemented a computer vision algorithm to quantify the number of coffee fruits and create yield maps. The results showed that this method effectively explained the factors influencing yield variations and had the advantages of low cost and independence from specific coffee harvester brands.
PRECISION AGRICULTURE
(2022)
Article
Computer Science, Information Systems
Jingru Sun, Mu Peng, Hongbo Jiang, Qinghui Hong, Yichuang Sun
Summary: With the proposal of the HMIAN model, it aims to better predict short-term traffic flow by considering the spatial and temporal features and effectively fusing traffic data with external factors. Experimental results demonstrate the effectiveness of the hierarchical mapping structure and the influence of different external factors on traffic prediction, providing valuable insights for future research in this area.
IEEE INTERNET OF THINGS JOURNAL
(2022)
Article
Robotics
Rodrigo Marcuzzi, Lucas Nunes, Louis Wiesmann, Jens Behley, Cyrill Stachniss
Summary: Autonomous vehicles need to understand their surroundings geometrically and semantically in order to plan and act appropriately in the real world. This paper proposes an approach called MaskPLS to perform panoptic segmentation of LiDAR scans by predicting a set of non-overlapping binary masks and semantic classes, fully avoiding the clustering step.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Robotics
Nicky Zimmerman, Tiziano Guadagnino, Xieyuanli Chen, Jens Behley, Cyrill Stachniss
Summary: This article presents a method for long-term localization in a changing indoor environment. By utilizing semantic cues and abstract semantic maps, the article proposes a localization framework that combines object detection and camera data with particle filters.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Robotics
Louis Wiesmann, Lucas Nunes, Jens Behley, Cyrill Stachniss
Summary: This letter focuses on point cloud-based place recognition and proposes a novel neural network architecture to reduce the training time. It extracts local features and computes the similarity between locations based on a global descriptor. By utilizing feature banks, faster training and improved performance are achieved.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Robotics
Haofei Kuang, Xieyuanli Chen, Tiziano Guadagnino, Nicky Zimmerman, Jens Behley, Cyrill Stachniss
Summary: This letter addresses the problem of estimating a mobile robot's pose in an indoor environment using 2D LiDAR data. It proposes a neural occupancy field method to implicitly represent the scene and synthesizes 2D LiDAR scans for arbitrary robot poses through volume rendering. The synthesized scans are used in an MCL system as an observation model for accurate localization.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Robotics
Ignacio Vizzo, Tiziano Guadagnino, Benedikt Mersch, Louis Wiesmann, Jens Behley, Cyrill Stachniss
Summary: This article introduces a simple and efficient sensor-based odometry system for accurate pose estimation of a robotic platform. The system utilizes point-to-point ICP matching, adaptive thresholding for correspondence matching, robust kernel, a simple yet widely applicable motion compensation approach, and point cloud subsampling strategy. It can operate under various environmental conditions using different LiDAR sensors.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Automation & Control Systems
Junyi Ma, Xieyuanli Chen, Jingyi Xu, Guangming Xiong
Summary: In this article, we propose a transformer-based network named SeqOT for place recognition based on sequential 3-D LiDAR scans. Our method exploits temporal and spatial information provided by sequential range images and generates global descriptors using multiscale transformers. The results show that our method outperforms state-of-the-art LiDAR-based place recognition methods and operates faster than the sensor's frame rate.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2023)
Article
Automation & Control Systems
Junyi Ma, Guangming Xiong, Jingyi Xu, Xieyuanli Chen
Summary: In this article, a cross-view transformer-based network called CVTNet is proposed to fuse different views generated by LiDAR data for place recognition in GPS-denied environments. Experimental results show that the method outperforms existing techniques in terms of robustness to viewpoint changes and long-time spans, while also exhibiting better real-time performance.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Environmental Sciences
Yuhan Xiao, Yufei Liu, Kai Luan, Yuwei Cheng, Xieyuanli Chen, Huimin Lu
Summary: This article proposes a novel multi-modal sensor fusion network called LRVFNet for accurate 2D object detection in urban autonomous driving scenarios. By effectively combining data from LiDAR, mmWave radar, and visual sensors through a deep multi-scale attention-based architecture, LRVFNet enhances accuracy and robustness.
Proceedings Paper
Automation & Control Systems
Matthias Zeller, Vardeep S. Sandhu, Benedikt Mersch, Jens Behley, Michael Heidingsfeld, Cyrill Stachniss
Summary: The paper focuses on the issue of moving object segmentation in noisy radar point clouds. A novel transformer-based approach is developed to accurately identify moving objects using radar velocity information and adaptive upsampling. The results show that the proposed method outperforms other state-of-the-art approaches in terms of performance.
2023 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA 2023)
(2023)
Proceedings Paper
Automation & Control Systems
Xingguang Zhong, Yue Pan, Jens Behley, Cyrill Stachniss
Summary: This paper focuses on achieving large-scale 3D reconstruction from 3D LiDAR measurements using implicit representations. By learning and storing implicit features in a hierarchical structure and converting them into signed distance values through a shallow neural network, the authors propose an incremental mapping system to address the issue of forgetting in continual learning. Experimental results demonstrate that their approach outperforms current state-of-the-art 3D mapping methods in terms of accuracy, completeness, and memory efficiency.
2023 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA 2023)
(2023)
Proceedings Paper
Automation & Control Systems
Alessandro Riccardi, Shane Kelly, Elias Marks, Federico Magistri, Tiziano Guadagnino, Jens Behley, Maren Bennewitz, Cyrill Stachniss
Summary: Monitoring the traits of plants and fruits is crucial for agriculture. In this paper, the authors propose a fruit descriptor and a matching cost function to address the challenge of matching fruits recorded at different growth stages. The experiments show that their descriptor achieves high spatio-temporal matching accuracy.
2023 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA 2023)
(2023)
Proceedings Paper
Automation & Control Systems
Gianmarco Roggiolani, Matteo Sodano, Tiziano Guadagnino, Federico Magistri, Jens Behley, Cyrill Stachniss
Summary: Plant phenotyping plays a crucial role in agriculture for understanding plant growth stage and development. This paper proposes a single convolutional neural network that simultaneously addresses the joint semantic, plant instance, and leaf instance segmentation problem in crop fields. The proposed architecture utilizes task-specific skip connections and introduces a novel automatic post-processing to handle spatially close instances commonly found in the agricultural domain. Experimental results show superior performance compared to state-of-the-art approaches, with reduced number of parameters and real-time processing.
2023 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA 2023)
(2023)
Proceedings Paper
Automation & Control Systems
Gianmarco Roggiolani, Federico Magistri, Tiziano Guadagnino, Jan Weyler, Giorgio Grisetti, Cyrill Stachniss, Jens Behley
Summary: This paper investigates the problem of reducing the number of labels without compromising the final segmentation performance in agricultural robots' semantic perception. The authors propose the use of self-supervised pre-training and domain-specific data augmentation strategies. Experimental results show that this method achieves superior performance compared to commonly used pre-trainings and obtains similar performance to fully supervised approaches with less labeled data.
2023 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA 2023)
(2023)