Enhancing Outdoor Fruit Localization Accuracy: A Stereo Network Approach with Parallax Attention
文献类型： 会议论文
第一作者： Feng Xie
作者： Feng Xie 1 ; Tao Li 2 ;
作者机构： 1.School of Agricultural Engineering, Jiangsu University, Zhenjiang, China
2.Intelligent Equipment Research Center, Beijing Academy of Agriculture and Forestry Sciences, Beijing, China
关键词： Location awareness;Three-dimensional displays;Cameras;Feature extraction;Data models;Filling;Task analysis
会议名称： Youth Academic Annual Conference of Chinese Association of Automation
主办单位：
页码： 1192-1196
摘要： Accurate fruit localization is a crucial and challenging task for achieving automated fruit harvesting in orchards. The depth information output by the stereo camera suffers from the problems of insufficient filling rate and noise, which will affect the accuracy of fruit localization. In this paper, we propose a stereo matching network based on attention mechanism to improve the quality of depth information. The network first extracts multi-scale features from left and right views, then uses the attention mechanism to generate the disparity cost, and finally the refinement disparity map is output from the network. Based on the disparity, we obtain the depth information and estimate the centroid of the fruit using the 3D frustum method. We collected orchard data and trained the network model to verify the effectiveness of the stereo matching network on the orchard dataset. Finally, fruit localization experiments were conducted based on the network, and the results show that the accuracy of fruit localization has been improved.
分类号： tp3
- 相关文献
[1]Segmentation and Extraction of Maize Phytomers Using 3D Data Acquired by RGB-D Cameras. Zhengqiang Fan,Na Sun,Jian Xu,Tao Li,Quan Qiu. 2023
[2]Multi-vision-based Localization and Pose Estimation of Occluded Apple Fruits for Harvesting Robots. Tao Li,Feng Xie,Qingchun Feng,Quan Qiu. 2022
[3]A Method for Counting Leaves of Cabbage Seedlings Based on Instance Segmentation. Ning Zhang,Huarui Wu,Huaji Zhu,Yisheng Miao,Xiang Sun. 2022
[4]Maize Leaf Biomass Retrieval at Multi-growing Stage Using UAV Multispectral Images Based on 3D Radiative Transfer Process-guided Machine Learning. Dan Zhao,Hao Yang,Guijun Yang,Xingang Xu,Bo Xu. 2024
[5]Design of structured-light vision system for tomato harvesting robot. Feng Qingchun,Zhou Jianjun,Wang Xiu,Cheng Wei. 2014
[6]Vertical features of yellow rust infestation on winter wheat using hyperspectral imaging measurements. Zhao, Jinling,Zhang, Dongyan,Huang, Linsheng,Zhang, Qing,Liu, Wenjing,Yang, Hao. 2016
[7]Geographical classification of apple based on hyperspectral imaging. Guo, Zhiming,Huang, Wenqian,Chen, Liping,Zhao, Chunjiang. 2013
[8]Improved Performance for Multi-Arm Harvest Robot System with Edge Computing and Distributed Communication. Feng Xie,Tao Li,Qingchun Feng. 2023
[9]Fruit Distribution Acquisition With Multi-Vision for Multi-Arm Harvesting Robots. Feng Xie,Na Sun,Jiaheng Li,Qingchun Feng,Tao Li. 2023
作者其他论文 更多>>
-
TriLoc-NetVLAD: Enhancing Long-term Place Recognition in Orchards with a Novel LiDAR-Based Approach
作者：Na Sun;Zhengqiang Fan;Quan Qiu;Tao Li;Qingchun Feng;Chao Ji;Chunjiang Zhao
关键词：
-
Improved Performance for Multi-Arm Harvest Robot System with Edge Computing and Distributed Communication
作者：Feng Xie;Tao Li;Qingchun Feng
关键词：Wireless communication;Visualization;System performance;Robot vision systems;Streaming media;Cameras;Real-time systems
-
Segmentation and Extraction of Maize Phytomers Using 3D Data Acquired by RGB-D Cameras
作者：Zhengqiang Fan;Na Sun;Jian Xu;Tao Li;Quan Qiu
关键词：Point cloud compression;Histograms;Three-dimensional displays;Robot vision systems;Crops;Feature extraction;Cameras
-
Fruit Distribution Acquisition With Multi-Vision for Multi-Arm Harvesting Robots
作者：Feng Xie;Na Sun;Jiaheng Li;Qingchun Feng;Tao Li
关键词：Location awareness;Point cloud compression;Visualization;Filtering;Robot kinematics;Robot vision systems;Process control
-
Multi-vision-based Localization and Pose Estimation of Occluded Apple Fruits for Harvesting Robots
作者：Tao Li;Feng Xie;Qingchun Feng;Quan Qiu
关键词：Deep learning;Location awareness;Three-dimensional displays;Shape;Pose estimation;Grasping;Multitasking
