Article
Computer Science, Artificial Intelligence
Tong Peng, Kun He, Yao Su, Ziwei Hui
Summary: A top-down segmentation model is proposed in this study, which combines visual perception and local features to address the issues in traditional segmentation models. Experimental results demonstrate the improvement in segmentation performance achieved by this model.
IET IMAGE PROCESSING
(2022)
Article
Agriculture, Multidisciplinary
Jiacai Liao, Ibrahim Babiker, Wen-fang Xie, Wei Li, Libo Cao
Summary: By combining background transfer learning and color-attention module methods, the dandelion segmentation method can effectively segment the dandelion with a satisfactory accuracy rate.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2022)
Article
Chemistry, Analytical
Takashi Anazawa, Motohiro Yamazaki, Shuhei Yamamoto, Ryoji Inaba
Summary: This study developed a novel method for Sanger DNA sequencing using the RGB image sensor of a digital color camera. By improving the spectral response of the sensor, the researchers successfully quantified four or more fluorophores in a mixed state, achieving accurate DNA sequencing.
SENSORS AND ACTUATORS B-CHEMICAL
(2022)
Article
Optics
Yuhyun Ji, Yunsang Kwak, Sang Mok Park, Young L. Kim
Summary: This study discusses the mapping of spectral stimuli to RGB color values using spectral response functions of a three-color image sensor, highlighting the device-dependent differences between models. It introduces a compressive sensing framework in the frequency domain for accurately predicting RGB spectral response functions with several primary colors.
Article
Energy & Fuels
Robinson Cavieres, Rodrigo Barraza, Danilo Estay, Jose Bilbao, Patricio Valdivia-Lefort
Summary: This article introduces an artificial neural network tool to quantify power loss in solar photovoltaic modules due to soiling and partial shading effects. The proposed method consists of three main stages: segmentation, resizing, and performance prediction. Compared to state-of-the-art computer vision architectures, the approach achieves similar results with a significant reduction in computational cost.
Article
Engineering, Electrical & Electronic
Miaohui Wang, Yijing Huang, Jiaxin Lin, Wuyuan Xie, Guanghui Yue, Shiqi Wang, Leida Li
Summary: This article addresses the challenges of foreground distortion and background artifacts in screen sharing on visual quality perception, proposing a method that includes foreground perception and background suppression. Experimental results show that this method achieves state-of-the-art results on two latest benchmark databases.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Computer Science, Artificial Intelligence
Lukasz Karbowiak, Janusz Bobulski
Summary: This article introduces the importance of background segmentation and proposes a method to compare algorithms under severe weather conditions. Through testing in different weather conditions, interesting differences in detail detection and detection noise were observed.
PEERJ COMPUTER SCIENCE
(2022)
Article
Computer Science, Artificial Intelligence
Jiawei Wu, Guolin Zheng, Kun Zeng, Haoyi Fan, Zuoyong Li
Summary: Image matting is the process of extracting specific objects from an image, widely used in various applications. Researchers have recently focused on trimap-free matting methods to address the dependence on prior information. This paper proposes a new set of matting subtasks and a novel matting network, achieving superior results compared to the state-of-the-art methods. The proposed method addresses issues of stagewise modeling, uncorrectable errors, and subtasks bottleneck, by decoupling the foreground and background segmentation and employing local disambiguation modules.
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS
(2022)
Article
Computer Science, Information Systems
Jae-Yeul Kim, Jong-Eun Ha
Summary: The paper proposes an algorithm using a deep-learning-based segmenter to generate a background model image, addressing the challenges of considering long-term temporal information. The new method shows significant improvements in error rate compared to traditional and the latest deep learning algorithms.
Article
Chemistry, Multidisciplinary
Ting-Hui Chiang, Meng-Hsiu Chiang, Ming-Han Tsai, Che-Cheng Chang
Summary: This paper proposes an image segmentation-based monocular depth estimation model with attention mechanisms to address depth variations in outdoor scenes. The model segments images into foreground and background regions and predicts depth maps individually. Attention mechanisms are adopted to extract meaningful features from complex scenes and improve the prediction of foreground and background depth maps. Experimental results show that the proposed model outperforms previous methods by 27.5% on the KITTI dataset.
APPLIED SCIENCES-BASEL
(2022)
Article
Materials Science, Multidisciplinary
Ying-Jun Quan, Soo-Hong Min, Sungjin Hong, Ji Ho Jeon, Won-Jin Kim, Sung-Hoon Ahn
Summary: This article introduces a camera-based sensing method for structural color sensors, which involves simple image processing to convert RGB data into hue data. Through experimental verification, this method has been proven effective and cost-efficient in application, potentially accelerating the adoption of the Internet of Things in various engineering fields.
ADVANCED MATERIALS TECHNOLOGIES
(2022)
Review
Agriculture, Multidisciplinary
Polina Kurtser, Stephanie Lowry
Summary: Fusing RGB images with depth data (RGB-D) is a growing modality in agriculture, but collecting appropriate data and ground truth information is challenging. This paper surveys existing RGB-D datasets for agricultural robotics, summarizes trends and challenges, evaluates sensor advantages, and analyzes the role of RGB-D data in vision-based machine learning tasks.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2023)
Article
Computer Science, Artificial Intelligence
Zongxin Yang, Yunchao Wei, Yi Yang
Summary: This paper investigates embedding learning principles for tackling the challenging semi-supervised video object segmentation task. Unlike previous approaches that focus on foreground objects, the paper emphasizes the importance of treating background equally. It proposes a Collaborative video object segmentation by Foreground-Background Integration (CFBI) approach, which separates feature embedding into the foreground object and background regions, leading to improved segmentation results. CFBI also incorporates pixel-level matching processes and instance-level attention mechanisms, making it robust to object scales. Based on CFBI, the paper introduces a multi-scale matching structure and an Atrous Matching strategy, resulting in the CFBI+ framework, which outperforms state-of-the-art methods on benchmark datasets.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Chiranjibi Sitaula, Yong Xiang, Sunil Aryal, Xuequan Lu
Summary: This study introduces the use of hybrid features to represent scene images, which differs from previous methods focusing solely on foreground or background information. By combining information from foreground, background, and hybrid sources, more accurate representation of scene images is achieved.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Chang Liu, Gang Yang, Shuo Wang, Hangxu Wang, Yunhua Zhang, Yutao Wang
Summary: In this study, a Transformer-based asymmetric network is proposed to tackle the issues of existing RGB-D salient object detection methods. The proposed network utilizes the power of Transformer to extract global semantic information from RGB data and a lightweight CNN backbone to extract spatial structure information from depth data. The asymmetric hybrid encoder reduces the number of parameters and increases speed without sacrificing performance. The method achieves superior performance over 14 state-of-the-art RGB-D methods on six public datasets. The authors' code will be released at (website).
IET COMPUTER VISION
(2023)
Article
Automation & Control Systems
Yongchen Guo, Bo Pan, Yanwen Sun, Guojun Niu, Yili Fu, Max Q. -H. Meng
Summary: This paper proposes a motion hysteresis compensation method for cable-driven mechanism, which leverages the relationship between actuate motor current and hysteresis phases to generate compensation curves. The method is shown to improve the accuracy and efficiency of compensation, and has potential applications in large-scale identification or repeated usage scenarios.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2023)
Article
Automation & Control Systems
Jin Wu, Miaomiao Wang, Hassen Fourati, Hui Li, Yilong Zhu, Chengxi Zhang, Yi Jiang, Xiangcheng Hu, Ming Liu
Summary: This study investigates the generalized rigid registration problem in high-dimensional Euclidean spaces and proposes a method to minimize the loss function using the Cayley formula. By deriving a closed-form linear least-square solution, the registration covariances are obtained, providing accurate probabilistic descriptions. The proposed method demonstrates its efficiency in terms of computation time and accuracy compared to previous algorithms. Additionally, it is applied to an interpolation problem on a special Euclidean group and shows practical superiority in covariance-aided Lidar mapping for robotic navigation.
IEEE TRANSACTIONS ON CYBERNETICS
(2023)
Article
Robotics
Bowen Yang, Qingwen Zhang, Ruoyu Geng, Lujia Wang, Ming Liu
Summary: Having good knowledge of terrain information is essential for improving the performance of legged robots in locomotion and navigation on complex terrains. We present a novel framework that generates dense robot-centric elevation maps online from sparse LiDAR observations, and provides uncertainty estimations. Our approach ensures high robustness and computational efficiency by using a novel pre-processing and point features representation approach. The generative Bayesian model recovers detailed terrain structures and provides pixel-wise reconstruction uncertainty, benefiting the downstream tasks of legged robots.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Automation & Control Systems
Kuanqi Cai, Weinan Chen, Chaoqun Wang, Shuang Song, Max Q. -H. Meng
Summary: In this paper, an integrated framework for finding the optimal path in complex environments is proposed, considering collision risk, social norms, and crowded areas. The framework includes a dynamic group model, a collision risk and human space model, and an improved navigation method. Experimental results demonstrate that this method can generate the optimal human-aware collision-free path in complex environments.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2023)
Article
Automation & Control Systems
Erli Lyu, Tingting Liu, Jiaole Wang, Shuang Song, Max Q-H Meng
Summary: This paper proposes a points-guided sampling net (PGSN) that utilizes geometric information to guide the sampling process in a sampling-based motion planner. By extracting geometric features from point clouds, a VAE feature extraction net and a multi-modal sampling net are designed. Additionally, a sampling-based motion planning algorithm called PG-RRT is presented based on PGSN, and its effectiveness is demonstrated through experiments.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2023)
Article
Chemistry, Analytical
Kwai-Wa Tse, Rendong Pi, Yuxiang Sun, Chih-Yung Wen, Yurong Feng
Summary: Traditional methods for crack inspection in large infrastructures are time-consuming and costly as they require multiple devices and instruments. In this study, we propose a real-time crack inspection system based on unmanned aerial vehicles, which successfully detects and classifies various types of cracks. The system accurately identifies the crack positions in the world coordinate system. Our detector, an improved YOLOv4 with an attention module, achieves 90.02% mean average precision (mAP) and outperforms the YOLOv4-original by 5.23% in terms of mAP. The proposed system is low-cost, lightweight, and not restricted by navigation trajectories. Experimental results demonstrate its robustness and effectiveness in real-world crack inspection tasks.
Article
Robotics
Zhenyu Xu, Yuehua Li, Shiqiang Zhu, Yuxiang Sun
Summary: Dense depth estimation is crucial for various applications and stereo matching has become popular for this task. However, stereo-LiDAR fusion is a promising method to overcome challenges like low textures and occlusions. To address the issue of sparse and uneven distribution of LiDAR data, a semi-dense depth expansion method is proposed using RGB images as a reference. This method shows superior performance over traditional sparse invariant convolution methods in terms of accuracy and robustness, as proven by experimental results on different datasets.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Robotics
Zhenhua Xu, Yuxuan Liu, Yuxiang Sun, Ming Liu, Lujia Wang
Summary: The road network graph plays a critical role in downstream tasks of autonomous driving. To annotate road network graphs effectively and efficiently, automatic algorithms are demanded. However, existing methods suffer from hard-coded algorithms and poor performance. We propose RNGDet++, an improved method with an instance segmentation head and the ability to leverage multi-scale features, which outperforms baseline methods on large-scale public datasets.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Robotics
Zhen Feng, Yanning Guo, Yuxiang Sun
Summary: Semantic scene understanding using thermal images is challenging due to the lack of color information and blurred edges. To tackle this, we propose a cross-modal edge-privileged knowledge distillation framework that utilizes a fusion-based segmentation network as a teacher to guide a student network that only uses thermal images. Experimental results demonstrate that the student network achieves superior performance with thermal images under the guidance of the teacher.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Automation & Control Systems
Ziwen Wang, Baoliang Zhao, Peng Zhang, Liang Yao, Qiong Wang, Bing Li, Max Q. -H. Meng, Ying Hu
Summary: A robotic system for automated breast ultrasound scanning is proposed in this study. The system obtains the point cloud of the breast from multiple angles and registers them together for accurate tissue shape reconstruction. Scan path planning is performed using an isometric 3-D point cloud searching algorithm for full and uniform coverage. A contact force-strain regression model is built for tissue deformation estimation and used to correct the planned scanning path. Additionally, a probe-tissue interaction model with scanning resistance is built for normal probe orientation adjustment.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2023)
Article
Robotics
Keyu Li, Yangxin Xu, Ziqi Zhao, Ang Li, Max Q. -H. Meng
Summary: This paper presents a closed-loop magnetic manipulation framework for robotic transesophageal echocardiography (TEE) acquisitions. The framework utilizes magnetic control methods for more direct, intuitive, and accurate manipulation of the probe. Extensive experiments validate the effectiveness of the framework in terms of localization accuracy, update rate, workspace size, and tracking accuracy.
IEEE TRANSACTIONS ON ROBOTICS
(2023)
Article
Biotechnology & Applied Microbiology
Xilong Cui, Junjun Zhu, Wanmei Yang, Yuxiang Sun, Xiuling Huang, Xiumei Wang, Haiyang Yu, Chengmin Liang, Zikai Hua
Summary: This study aimed to investigate the effects of the sagittal location of the fracture region on the biomechanics of the internal fixation system and surgical strategy. The results showed that sagittal location has an impact on stress distribution of the fixation system, but does not influence the selection of surgical strategy.
FRONTIERS IN BIOENGINEERING AND BIOTECHNOLOGY
(2023)
Article
Automation & Control Systems
Fei Meng, Liangliang Chen, Han Ma, Jiankun Wang, Max Q. -H. Meng
Summary: Building a general and efficient path planning framework in uncertain nonconvex environments is challenging. Traditional methods involve convexifying obstacles and presuming Gaussian distribution, which are not universal. Our novel neural risk-bounded path planner quickly finds near-optimal solutions with acceptable collision probability in complex environments.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2023)
Article
Robotics
Yongchen Guo, Bo Pan, Yili Fu, Max Q. -H. Meng
Summary: Learning-based grip force measurement methods in RAMIS outperform traditional model-based methods and avoid the issues of sensor-based approaches. However, there is limited research on grip force measurement in mass-produced surgical instruments. This letter proposes a novel learning-based method, ACAM-FoC, which considers the differences in motion hysteresis and mechanism friction among mass-produced surgical instruments.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Review
Engineering, Biomedical
Zhe Min, Ang Zhang, Zhengyan Zhang, Jiaole Wang, Shuang Song, Hongliang Ren, Max Q. -H. Meng
Summary: This paper provides a concise review of rigid point set registration methods in computer-assisted orthopedic surgery (CAOS). The challenge lies in establishing point correspondences between two point sets under noise, outliers, and partial overlapping. The paper discusses and compares the advantages and disadvantages of surveyed registration algorithms, and also proposes potential future research directions.
IEEE TRANSACTIONS ON MEDICAL ROBOTICS AND BIONICS
(2023)