基于激光-视觉融合的配怀猪舍内导航建图技术研究

doi:10.3969/j.issn.1004-1524.20231148

浙江农业学报 ›› 2024, Vol. 36 ›› Issue (10): 2358-2367.DOI: 10.3969/j.issn.1004-1524.20231148

基于激光-视觉融合的配怀猪舍内导航建图技术研究

潘梓博¹^,²(), 周昕², 徐杏², 刘凯歌¹^,², 吉洪湖³, 路伏增², 叶春林¹^,^*(), 周卫东²^,^*()

1.浙江科技大学生物与化学工程学院,浙江杭州 310023
2.浙江省农业科学院畜牧兽医研究所,浙江杭州 310021
3.浙江省金华市农业科学研究院,浙江金华 321017

收稿日期:2023-09-26 出版日期:2024-10-25 发布日期:2024-10-30
作者简介:潘梓博(1998—),男,浙江宁海人,硕士研究生,研究方向为畜牧智能装备。E-mail: 222103855051@zust.edu.cn
通讯作者: ^*叶春林,E-mail:chlye2005@126.com;周卫东,E-mail: zhouwd@zaas.ac.cn
基金资助:
浙江省“三农九方”科技协作计划“揭榜挂帅”项目(2023SNJF060);宁波市科技计划项目(2022Z178)

Research on navigation and mapping in pregnancy pig house based on laser-vision fusion technology

PAN Zibo¹^,²(), ZHOU Xin², XU Xing², LIU Kaige¹^,², JI Honghu³, LU Fuzeng², YE Chunlin¹^,^*(), ZHOU Weidong²^,^*()

1. School of Biological & Chemical Engineering, Zhejiang University of Science and Technology, Hangzhou 310023, China
2. Institute of Animal Husbandry and Veterinary, Zhejiang Academy of Agricultural Sciences, Hangzhou 310021, China
3. Zhejiang Jinhua Academy of Agricultural Sciences, Jinhua 321017, Zhejiang, China

Received:2023-09-26 Online:2024-10-25 Published:2024-10-30

摘要/Abstract

摘要：

针对规模生猪养殖场猪舍内复杂设施环境下,现有的纯激光或纯视觉主流导航技术精度不够,影响移动智能装备作业效果的问题,研究建立了深度相机内参标定、激光雷达与深度相机数据融合的前端里程计、后端全面优化、回环检测和建图等方法,比较研究了利用激光-视觉融合建图技术构建的配怀猪舍通道和走廊二维栅格地图和点云地图的效果。结果显示: 融合建立的二维栅格地图较单激光雷达建图的信息量更多,能显示直观的点云图,较完整地展示实际场景中的道路信息与色彩信息,起点与终点坐标的相对误差范围分别为0.02%~0.33%、0.48%~1.62%,明显低于3种算法的激光建图,具有良好的精度和建图效率。

关键词: 激光雷达, 深度相机, 同步定位与地图构建, 融合导航建图, 猪舍

Abstract:

Research has been conducted to address the problem of insufficient accuracy of the laser-based or vision-based navigation technologies in the complex facility environment of large-scale pig farms. In the present study, a new method was established by calibrating the intrinsic parameters of depth cameras, fusing data from laser radar and depth cameras for front-end odometry, and comprehensive optimization for back-end processing, loop closure detection, and mapping. A comparative study was conducted to evaluate the effectiveness of constructing 2D grid maps and point cloud maps by using laser-vision fusion mapping technology for passages and corridors. The results demonstrated that the fused 2D grid map provided more information than the single laser radar mapping, and it also displayed an intuitive point cloud representation, which could effectively present the road and color information from the actual scene. The relative error ranges for the coornidates of the starting potint and the terminal point were 0.02%-0.33% and 0.48%-1.62%, respectively, which were lower than those of the three algorithms used for laser mapping, indicating good accuracy and mapping efficiency.

Key words: lidar, depth camera, simultaneous localization and mapping, fusion navigation mapping, pig house

中图分类号:

S818.9

潘梓博, 周昕, 徐杏, 刘凯歌, 吉洪湖, 路伏增, 叶春林, 周卫东. 基于激光-视觉融合的配怀猪舍内导航建图技术研究[J]. 浙江农业学报, 2024, 36(10): 2358-2367.

PAN Zibo, ZHOU Xin, XU Xing, LIU Kaige, JI Honghu, LU Fuzeng, YE Chunlin, ZHOU Weidong. Research on navigation and mapping in pregnancy pig house based on laser-vision fusion technology[J]. Acta Agriculturae Zhejiangensis, 2024, 36(10): 2358-2367.

图/表 17

图1 实验装置的控制系统框架 SLAM,同步定位与地图构建;DC,直流电。

Fig.1 Diagram of experimental device control system framework SLAM, Simultaneous localization and mapping; DC, Direct current.

图2 STM32 F407单片机(左)和Jetson nano开发板(右)的实物照片

Fig.2 Photos of STM32 F407 microcontroller (left) and Jetson nano development board (right)

图3 镭神M10P激光雷达(左)和RealSense L515深度相机(右)的实物照片

Fig.3 Photos of Riegl M10P laser scanner (left) and RealSense L515 depth camera (right)

图4 实验装置的正视图(左)和后视图(右)

Fig.4 The front view (left) and rear view (right) of the experimental apparatus

图5 激光-视觉融合技术框架

Fig.5 Framework of laser-vision fusion technology

图6 标定板照片

Fig.6 Photo of calibration board

图7 限位栏实物(左)及其特征点提取图(右)

Fig.7 Physical diagram (left) and its feature point extraction diagram (right) of limit switch

图8 重投影误差示意图

Fig.8 Diagram of reprojection error

图9 k-means算法聚类流程图

Fig.9 Flow char of classification based on k-means algorithm

图10 建图实验现场照片

Fig.10 Photo of construction experiment site

图11 基于Gmapping算法的配怀舍走廊(上)、通道(下)二维栅格地图

Fig.11 The two-dimensional raster map of corridor (up) and passage (down) of pregnancy house based on Gmapping algorithm

图12 基于Cartographer算法的配怀舍走廊(上)、通道(下)二维栅格地图

Fig.12 The two-dimensional raster map of corridor (up) and passage (down) of pregnancy house based on Cartographer algorithm

图13 基于Katro算法的配怀舍走廊(上)、通道(下)二维栅格地图

Fig.13 The two-dimensional raster map of corridor (up) and passage (down) of pregnancy house based on Katro algorithm

表1 不同算法建图的效果对比

Table 1 Comparison of mapping effect of different algorithms

建图算法 Mapping algorithm	建图时间 Mapping time/s	平均偏移角度 Mean deviation angle/(°)	局部建图范围 Local mapping scope	地图完整度 Map integrity
Gmapping	630	10	大Wide	较为完整Relatively complete
Cartographer	803	5	小Small	不太完整Less complete
Katro	686	2	中Medium	较为完整Relatively complete
融合方法Fusion method	615	3	大Wide	十分完整Quite complete

图14 走廊的二维栅格地图(A)和点云图(B) 图中的绿色线条表示移动机器人的运动轨迹。下同。

Fig.14 2D grid map (A) and point cloud map (B) of corridor The green line represents the motion trajectory of the mobile robot. The same as below.

图15 通道的二维栅格地图(A)和点云图(B)

Fig.15 2D grid map (A) and point cloud map (B) of passage

表2 不同算法建图的坐标及其相对误差

Table 2 Coordinates of different mapping algorithms and relative errors

建图算法 Mapping algorithm	地点 Site	位置 Location	坐标 Coordinate(x,y)/m	相对误差 Relative error/%
Gmapping	走廊Corridor	起始Starting point	(0.15,-4.00)	0.37
		终点Terminal point	(60.00, 17.00)	2.03
	舍内Hoggery	起始Starting point	(0.20,-4.00)	0.98
		终点Terminal point	(45.00,-5.00)	0.96
Cartographer	走廊Corridor	起始Starting point	(0.30,-5.00)	0.40
		终点Terminal point	(64.00,-11.00)	0.51
	舍内Hoggery	起始Starting point	(-0.30, 6.00)	1.03
		终点Terminal point	(45.00,-6.00)	0.48
Katro	走廊Corridor	起始Starting point	(0.18,-4.00)	0.35
		终点Terminal point	(66.00,-2.00)	1.76
	舍内Hoggery	起始Starting point	(0.20,-5.00)	0.02
		终点Terminal point	(46.00,-6.00)	0.96
融合建图Fusion mapping	走廊Corridor	起始Starting point	(0.20,-5.00)	0.33
		终点Terminal point	(65.00,-6.00)	1.62
	舍内Hoggery	起始Starting point	(0.20,-5.00)	0.02
		终点Terminal point	(45.00,-6.00)	0.48

参考文献 15

[1]	丁伟, 张文爱, 冯青春, 等. 畜禽舍防疫消毒机器人控制系统设计[J]. 中国农机化学报, 2019, 40(10): 175-181.
	DING W, ZHANG W A, FENG Q C, et al. Design of robot control system for disinfection of livestock breeding[J]. Journal of Chinese Agricultural Mechanization, 2019, 40(10): 175-181. (in Chinese with English abstract)
[2]	高翔. 视觉SLAM十四讲: 从理论到实践[M]. 2版. 北京: 电子工业出版社, 2019.
[3]	SMITH R, SELF M, CHEESEMAN P. Estimating uncertain spatial relationships in robotics[M]. Autonomous robot vehicles. New York: Springer New York, 1990: 167-193.
[4]	ZHANG J, SINGH S. LOAM: lidar odometry and mapping in real-time[EB/OL]. (2014-07-12) [2023-09-26]. https://www.roboticsproceedings.org/rss10/p07.pdf.
[5]	赵伟卓. 基于视觉与激光雷达的ROS机器人自主导航研究[D]. 沈阳: 沈阳理工大学, 2022.
	ZHAO W Z. Research on autonomous navigation of ROS robot based on vision and LiDAR[D]. Shenyang: Shenyang Ligong University, 2022. (in Chinese with English abstract)
[6]	VELAS M, SPANEL M, MATERNA Z, et al. Calibration of RGB camera with velodyne LiDAR[EB/OL]. [2023-09-26]. https://dspace5.zcu.cz/bitstream/11025/26408/1/Velas.pdf.
[7]	代云勇. 畜牧养殖场环境巡检机器人的建图及导航研究[D]. 赣州: 江西理工大学, 2020.
	DAI Y Y. Research on mapping and navigation of environment inspection robot in animal husbandry farm[D]. Ganzhou: Jiangxi University of Science and Technology, 2020. (in Chinese with English abstract)
[8]	喻良友. 基于激光雷达与视觉融合的移动机器人SLAM研究[D]. 南昌: 南昌大学, 2022.
	YU L Y. Research on SLAM of mobile robot based on lidar and vision fusion[D]. Nanchang: Nanchang University, 2022. (in Chinese with English abstract)
[9]	SICILIANO B, SCIAVICCO L, VILLANI L, et al. Robotics:modelling, planning and control[M]. London: Springer Science & Business Media, 2010.
[10]	SCARAMUZZA D, FRAUNDORFER F. Visual odometry[tutorial[J]. IEEE Robotics & Automation Magazine, 2011, 18(4): 80-92.
[11]	PIRE T, FISCHER T, CASTRO G, et al. S-PTAM: stereo parallel tracking and mapping[J]. Robotics and Autonomous Systems, 2017, 93: 27-42.
[12]	李志梅. 深度相机和2D激光融合SLAM安防巡检机器人[D]. 广州: 广东工业大学, 2020.
	LI Z M. Depth camera and 2D laser fusion SLAM security inspection robot[D]. Guangzhou: Guangdong University of Technology, 2020. (in Chinese with English abstract)
[13]	王佳龙. 基于激光视觉融合的机器人位姿估计及建图技术研究[D]. 长沙: 湖南大学, 2021.
	WANG J L. Research on pose estimation and mapping technology of mobile robot based on laser and visual fusion[D]. Changsha: Hunan University, 2021. (in Chinese with English abstract)
[14]	孟令波. 基于视觉激光融合的无人车自主定位方法研究[D]. 哈尔滨: 哈尔滨工业大学, 2020.
	MENG L B. Autonomous localization method for driverless cars based on fusion of vision and LiDAR[D]. Harbin: Harbin Institute of Technology, 2020. (in Chinese with English abstract)
[15]	钟敏. 基于视觉与激光雷达融合的移动机器人自主导航研究[D]. 哈尔滨: 哈尔滨工业大学, 2021.
	ZHONG M. Research on autonomous navigation of mobile robot based on fusion of vision and LiDAR[D]. Harbin: Harbin Institute of Technology, 2021. (in Chinese with English abstract)

基于激光-视觉融合的配怀猪舍内导航建图技术研究

Research on navigation and mapping in pregnancy pig house based on laser-vision fusion technology

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 17

参考文献 15

相关文章 2

编辑推荐

Metrics

本文评价

[1]	李颀, 李煜哲. 弱光条件下猪舍清洗目标检测[J]. 浙江农业学报, 2023, 35(9): 2240-2249.
[2]	陈绍孟;路伏增;*;华卫东;徐如海;褚晓红;戴丽荷;胡锦平;宋维龙. 舍内喷洒中草药制剂预防断奶仔猪腹泻的试验效果[J]. , 2013, 25(6): 0-1227.