Article
Computer Science, Information Systems
Yixin Sun, Yusen Luo, Xiaoyu Chai, Pengpeng Zhang, Qian Zhang, Lizhang Xu, Lele Wei
Summary: The proposed crop density estimation method based on LiDAR, double-threshold segmentation, and clustering algorithms effectively estimates the density of mature rice plants. Experimental results show its accuracy with RMSE of 9.968 and 5.877, and MAPE of 5.67% and 3.37%, laying a foundation for intelligent harvest.
Article
Environmental Sciences
Anjin Chang, Jinha Jung, Junho Yeom, Juan Landivar
Summary: This study developed a method for detecting and characterizing individual sorghum panicles using a 3D point cloud derived from UAV images, and proposed strategies to estimate panicle volumes. Experimental results showed a high correlation between UAV-derived panicle parameters and ground measurements, providing important information for studying genetic diversity and yield estimation in sorghum.
Article
Environmental Sciences
Zexin Yang, Qin Ye, Jantien Stoter, Liangliang Nan
Summary: This paper presents a new method that integrates continuous implicit representations with point clouds, by parameterizing the continuous unsigned distance field around each point and concatenating it with the Cartesian coordinates of the point as the network input, to better leverage implicit representations. It also introduces a novel local canonicalization approach to ensure the transformation-invariance of the encoded implicit features. Experiments have demonstrated the effectiveness of the proposed method in object-level classification and scene-level semantic segmentation tasks.
Article
Agronomy
Dan Wu, Lejun Yu, Junli Ye, Ruifang Zhai, Lingfeng Duan, Lingbo Liu, Nai Wu, Zedong Geng, Jingbo Fu, Chenglong Huang, Shangbin Chen, Qian Liu, Wanneng Yang
Summary: This research presents an automatic and nondestructive method for 3D panicle modeling in rice plants. The method integrates various techniques such as shoot rice reconstruction, shape from silhouette, deep convolutional neural network, ray tracing, and supervoxel clustering. It demonstrates high efficiency and performance in recovering the 3D shapes of rice panicles from multiview images, and is adaptable to diverse rice plants. The proposed algorithm outperforms the classical structure-from-motion method in terms of texture preservation and computational efficiency.
Article
Agriculture, Multidisciplinary
Juntao Xiong, Junhao Liang, Yanyun Zhuang, Dan Hong, Zhenhui Zheng, Shisheng Liao, Wenxin Hu, Zhengang Yang
Summary: This paper proposes a real-time localization and semantic map reconstruction method for smart agriculture, integrating the visual-inertial SLAM VINS-RGBD framework with the semantic segmentation algorithm BiSeNetV1. The method achieves high accuracy and real-time performance in semantic map reconstruction, providing essential technical support for orchard management.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2023)
Article
Plant Sciences
Xindong Guo, Yu Sun, Hua Yang
Summary: Semantic segmentation of 3D point clouds has been important in plant phenotyping. Existing methods down-sample point clouds when processing large-scale plant point clouds, failing to fully utilize high-resolution scanning devices. To address this, we propose FF-Net, a feature-fusion-based method with voxel and point branches. Our method outperformed three commonly used models on maize and tomato datasets, achieving the best mIoU of 80.95% and 86.65% respectively. Cross-validation experiments confirmed the generalization ability of our method.
Article
Computer Science, Artificial Intelligence
Luis Roldao, Raoul de Charette, Anne Verroust-Blondet
Summary: This paper surveys the progress of semantic scene completion (SSC), highlighting the unresolved challenges and evaluating the performance of state-of-the-art techniques on popular datasets.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2022)
Article
Computer Science, Artificial Intelligence
Yupeng Song, Fazhi He, Linkun Fan, Jicheng Dai, Qing Guo
Summary: In this paper, a novel approach is proposed to enhance the machine perception of 3D semantic information in point clouds. The approach includes a dynamic local self-attention mechanism and a dynamic self-attention learning block, which can handle unordered and irregular point cloud data, learn global and local features, and dynamically learn important local semantic information. The method shows advantages in point cloud tasks.
ADVANCED ENGINEERING INFORMATICS
(2022)
Article
Environmental Sciences
Fengjiao Gao, Yiming Yan, Hemin Lin, Ruiyao Shi
Summary: This paper proposes a novel 3D semantic segmentation method called PIIE-DSA-net, which combines low-level features and deep features for more reliable feature extraction. Experimental results demonstrate that this method achieves good segmentation results on both indoor and outdoor datasets.
Article
Computer Science, Information Systems
Muhammad Ibrahim, Naveed Akhtar, Michael Wise, Ajmal Mian
Summary: The translation introduces a public annotation tool PC-Annotate for research on 3D point clouds, and a large outdoor dataset PC-Urban for 3D semantic segmentation.
Article
Construction & Building Technology
Yunxiang Zhou, Ankang Ji, Limao Zhang, Xiaolong Xue
Summary: This paper presents a method called ASPCNet, which uses point cloud technique to segment 3D tunnel point clouds for accurate and efficient processing. The developed model demonstrates superior performance and efficiency and the adopted sampling method strengthens the model performance.
AUTOMATION IN CONSTRUCTION
(2023)
Article
Environmental Sciences
Reza Mahmoudi Kouhi, Sylvie Daniel, Philippe Giguere
Summary: Currently, 3D point clouds are widely used for presenting 3D objects and accurately localizing them. However, the lack of semantic information in raw point clouds has led to the development of deep neural networks for semantic segmentation. Few prior works have studied the impact of data preparation on network performance. Therefore, this study proposes novel data preparation methods that improve the performance of deep neural networks for point cloud semantic segmentation.
Article
Environmental Sciences
Tengfei Wang, Qingdong Wang, Haibin Ai, Li Zhang
Summary: In this paper, a semantics-and-primitives-guided method for automatically reconstructing indoor scenes is proposed. It includes semantic segmentation using a local, fully connected graph neural network, primitive-based reconstruction based on enumerable features, and a coarse-to-fine registration method. The results demonstrate high-quality reconstruction with better resilience to incomplete and noisy point clouds.
Article
Computer Science, Software Engineering
Yushuang Wu, Zizheng Yan, Shengcai Cai, Guanbin Li, Xiaoguang Han, Shuguang Cui
Summary: Semantic segmentation of point cloud usually requires dense annotation, but our PointMatch framework explores data and label information simultaneously, achieving better representation learning and robustness to label sparsity.
COMPUTERS & GRAPHICS-UK
(2023)
Article
Agriculture, Multidisciplinary
Luolin Xiao, Zhibin Pan, Xiaoyong Du, Wei Chen, Wenpeng Qu, Youda Bai, Tian Xu
Summary: Unmanned aerial vehicles (UAVs) have the potential to enhance precision agriculture in unmanned farms by reducing manual interventions and improving data collection efficiency. This study proposes a new method called weighted skip-connection feature fusion (WSFF) to augment UAV rice panicle image segmentation. The constructed model WSUNet, which combines WSFF and UNet, shows improved segmentation performance without additional computational cost.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2023)
Article
Plant Sciences
Jie Zong, Li Wang, Lu Zhu, Lianle Bian, Bo Zhang, Xiaofei Chen, Guoqiang Huang, Xuelian Zhang, Junyi Fan, Liming Cao, George Coupland, Wanqi Liang, Dabing Zhang, Zheng Yuan
Summary: This study used high-throughput single-cell RNA sequencing to construct a gene expression resource and revealed the transcriptomic atlas of early rice inflorescence development, providing insights into axillary meristem differentiation and floret development.
Article
Chemistry, Analytical
Liang Gong, Binhao Chen, Wenbin Xu, Chengliang Liu, Xudong Li, Zelin Zhao, Lujie Zhao
Summary: This paper proposes a human-computer interaction method for remotely manipulating life-size humanoid robots by mapping motion capture data to each joint motion angle of the robot. It also introduces a DTW-based trajectory evaluation method to quantitatively evaluate the difference between robot trajectory and human motion. Experimental results demonstrate the feasibility and real-time performance of the control method.
Article
Chemistry, Analytical
Xueqi Li, Rong Li, Zheng Yuan, Zaobing Zhu, Wenting Xu, Yijie Wang, Dabing Zhang, Litao Yang
Summary: In this study, a new method called Cc-qPCR was developed for rapid screening of mutants and identification of their genotypes, as well as evaluating gene editing efficiency. Testing on rice samples showed that Cc-qPCR is an accurate and effective approach, which can enhance the efficiency and range of molecular breeding techniques.
ANALYTICAL CHEMISTRY
(2022)
Article
Chemistry, Analytical
Yijie Wang, Rong Li, Zaobing Zhu, Zheng Yuan, Chen Wang, Li Wang, Dabing Zhang, Litao Yang
Summary: This study presents a novel SMART approach for simultaneous analysis of CRISPR/Cas-induced mutants, genotypes, and gene editing frequency (GEF). The SMART method is highly specific, sensitive, and accurate, capable of detecting various types of mutations without strict requirements on target sequences. It enables versatile screening, genotyping, and quantification of CRISPR/Cas-induced mutants, and has potential for clinical detection of rare mutations.
SENSORS AND ACTUATORS B-CHEMICAL
(2023)
Article
Agronomy
Yixiang Huang, Pengcheng Xia, Liang Gong, Binhao Chen, Yanming Li, Chengliang Liu
Summary: This paper presents a new in-field interactive cognition phenotyping paradigm to solve occlusion and observation pose problems in field phenotyping. By introducing an active interactive cognition method and an attentional residual network, automatic high-throughput phenotyping is achieved with accurate tiller counting data.
Article
Agronomy
Lin Liu, Jin Yuan, Liang Gong, Xing Wang, Xuemei Liu
Summary: A novel method for predicting the dynamic growth of substrate-cultivated leafy vegetables in a solar greenhouse is proposed, based on in situ sensing of phenotypic and environmental data. The results show that the method can accurately predict the fresh weight of lettuce in the future, and the accuracy can be improved by introducing more data batches.
Article
Plant Sciences
Ming Yan, Fangjun Feng, Xiaoyan Xu, Peiqing Fan, Qiaojun Lou, Liang Chen, Anning Zhang, Lijun Luo, Hanwei Mei
Summary: Phosphate (Pi) is essential for plant growth, but low-Pi stress limits crop growth and yield. The tolerance to low-Pi stress in rice varies among different germplasm resources. This study identified significant loci associated with biomass and grain yield under low-Pi supply through a genome-wide association study (GWAS) using a diverse collection of 191 rice accessions. The candidate gene OsAAD was found to be up-regulated after low-Pi stress and suppressing its expression improved phosphorus use efficiency (PPUE) and grain yields in rice.
FRONTIERS IN PLANT SCIENCE
(2023)
Article
Computer Science, Artificial Intelligence
Liang Gong, Yingxin Wu, Bishu Gao, Yefeng Sun, Xinyi Le, Chengliang Liu
Summary: This paper proposes a real-time planning and tracking control method for wireless charged nonholonomic autonomous vehicles, which addresses the challenges of dynamic obstacle avoidance and precise targeting posture control. By considering vehicle jerk and targeting posture constraints, a kinematic and dynamic model is established as the foundation of planning and tracking control. A real-time layered planner and a fast nonlinear model predictive control algorithm are designed to generate reference trajectory and realize high precision trajectory tracking during dynamic obstacle avoidance.
IEEE TRANSACTIONS ON INTELLIGENT VEHICLES
(2023)
Article
Engineering, Electrical & Electronic
Pengcheng Xia, Yixiang Huang, Chengjin Qin, Dengyu Xiao, Liang Gong, Chengliang Liu, Wenliao Du
Summary: This article proposes an adaptive feature utilization method and a global temporal convolutional network for predicting the remaining useful life (RUL) of machinery. The effectiveness and superiority of the method are validated through case studies.
IEEE SENSORS JOURNAL
(2023)
Article
Engineering, Multidisciplinary
Wei Zhang, Liang Gong, Yefeng Sun, Bishu Gao, Chenrui Yu, Chengliang Liu
Summary: This paper proposes a precise visual positioning method for agricultural mobile robots in the greenhouse, which improves their positioning accuracy via minimizing fiducial marker reprojection errors. The method uses fiducial markers to enhance environment features and formulates a markers-based visual positioning task as a Perspective-n-Point problem. A reprojection error minimization approach is proposed, considering the markers' distance and image noise, to ensure higher positioning accuracy. Synthetic and field experiments are conducted to evaluate the performance of the proposed method.
MEASUREMENT SCIENCE AND TECHNOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Liang Gong, Zhiyu Yang, Yihang Yao, Binhao Chen, Wenjie Wang, Xiaofeng Du, Yidong He, Chengliang Liu
Summary: This paper proposes a method for on-site image acquisition and semi-automatic annotation based on eye-tracking, aiming to improve annotation efficiency and overcome the bottleneck of existing methods. The method combines human-machine interaction to make full use of human perception and recognition intelligence. Experimental results show that the method achieves annotation quality comparable to manual methods while significantly improving efficiency.
IEEE TRANSACTIONS ON HUMAN-MACHINE SYSTEMS
(2023)
Article
Plant Sciences
Fangjun Feng, Xiaosong Ma, Ming Yan, Hong Zhang, Daoliang Mei, Peiqing Fan, Xiaoyan Xu, Chunlong Wei, Qiaojun Lou, Tianfei Li, Hongyan Liu, Lijun Luo, Hanwei Mei
Summary: Mesocotyl elongation is a crucial trait for deep sowing tolerance and well seedling establishment in dry direct sowing rice (DDSR). Using dark germination, 294 accessions from Rice Diversity Panel 1 (RDP1) and 312 lines from the Hanyou 73 (HY73) recombinant inbred line (RIL) population were screened for mesocotyl length (ML). Through GWAS and QTL analysis, multiple associated SNPs and QTLs were identified on various chromosomes, and candidate genes were predicted and partially validated using RNA-seq data. Strategies for donor parent selection in DDSR breeding were also discussed.
Article
Agronomy
Liang Gong, Fei Huang, Wei Zhang, Yanming Li, Chengliang Liu
Summary: This study proposes a system to predict seedbed-sized sunshine using a cross-scale approach, aiming to improve the efficiency of photosynthesis in plant factories. The research utilizes a hybrid modeling algorithm to predict hourly sunshine and achieves uniform solar energy absorption through a dynamic seedbed scheduling scheme.
Article
Agronomy
Yanming Li, Yibo Guo, Liang Gong, Chengliang Liu
Summary: This paper proposes a harvesting operation image segmentation method based on SLIC superpixel segmentation and the AdaBoost ensemble learning algorithm, which addresses the challenges of reduced robustness and limited perception of crop lodging in the current harvester route detection method. Experimental results demonstrate the effectiveness of this method in successfully segmenting the harvested and unharvested areas of the farmland.
Article
Computer Science, Artificial Intelligence
Ke Lin, Duantengchuan Li, Yanjie Li, Shiyu Chen, Qi Liu, Jianqi Gao, Yanrui Jin, Liang Gong
Summary: Reinforcement learning suffers from sample inefficiency and exploration issues. Learning from demonstration (LfD) was proposed to address these problems, but often requires a large number of demonstrations. This study presents a sample efficient teacher-advice mechanism with Gaussian process (TAG) that leverages a few expert demonstrations. The TAG mechanism helps the agent explore the environment more intentionally and guides the agent accurately using a guided policy. Experiments show that TAG helps RL algorithms achieve significant performance gains and outperforms other LfD methods on delayed reward and continuous control environments.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)