计算机科学
网格
天蓬
算法
网格单元
计算机视觉
人工智能
数学
地理
几何学
考古
作者
Huaiqu Feng,Te Xi,Dunhong Yang,Yulei Pan,Guoyang Zhao,Rongkai Shi,Bo Chen,Yongwei Wang,Jun Wang
出处
期刊:
日期:2024-01-01
被引量:3
标识
DOI:10.13031/aim.202400095
摘要
Abstract. This research presents a comprehensive study on the contour mapping of rice canopies for Agricultural Unmanned Ground Vehicles (Agri-UGVs) operating in complex, unknown hybrid rice fields. Accurate mapping is crucial for these vehicles to efficiently plan their operational areas while steering clear of protected zones. The motion equations that underpin the control of Agri-UGVs tasked with intelligent impurity removal are heavily reliant on the precise measurements of hybrid rice height and canopy contour. Therefore, the accurate estimation of these parameters is vital for the effective perception and operation of Agricultural robots. The Agri-UGV, equipped with advanced RGB-D depth sensors, facilitates the collection of distance measurements and the perception of the surrounding hybrid rice canopy's geometrical structure. This introduces an innovative mapping estimation and algorithm tailored for the hybrid rice canopy contour, utilizing a combination of kinematic and inertial measurements to achieve proprioceptive localization of the Agri-UGV. The algorithm incorporates an inference framework for grid mapping, which leverages real-time RGB-D data to iteratively refine the probabilistic distributions of the canopy's morphological properties. This is achieved by correlating the hybrid rice height intercept with in-field data while accounting for drift and uncertainty in the state estimation process. The result is a set of probabilistic contour estimates, formatted as grid-based elevation maps. The proposed algorithm has been designed to operate in real-time within the ROS-Jetson system, and its performance is rigorously evaluated using metrics such as relative and absolute pose error. The algorithm has been successfully deployed on a High-clearance Agri-UGV platform, demonstrating its capability to accurately estimate the morphological properties of rice canopy height in a variety of environments, ranging from controlled indoor settings to dynamic outdoor paddy fields. It also details the experiments conducted to explore the Agri-UGV's mapping capabilities within the cluttered hybrid rice environment. This research significantly contributes to the field of autonomous agricultural operations by enhancing the precision and reliability of canopy contour mapping, which is essential for the effective navigation and task execution of Agri-UGVs in intricate agricultural settings.
科研通智能强力驱动
Strongly Powered by AbleSci AI