基于深度学习的近地面草原鼠洞识别计数关键问题研究与应用

doi:10.3969/j.issn.1004-1524.20230858

浙江农业学报 ›› 2024, Vol. 36 ›› Issue (9): 2146-2154.DOI: 10.3969/j.issn.1004-1524.20230858

基于深度学习的近地面草原鼠洞识别计数关键问题研究与应用

郭秀明¹(), 王大伟²^,³^,⁴^,^*(), 刘升平¹^,^*(), 诸叶平¹, 刘晓辉², 林克剑⁴, 王佳宇⁵, 李非⁵

1.中国农业科学院农业信息研究所,农业农村部农业信息服务技术重点实验室,北京 100081
2.中国农业科学院植物保护研究所,植物病虫害综合治理全国重点实验室,北京 100081
3.中国农业科学院西部农业研究中心,新疆昌吉 831100
4.中国农业科学院草原研究所,农业农村部人工草地生物灾害监测与绿色防控重点实验室,内蒙古呼和浩特 010010
5.锡林郭勒盟草原工作站,内蒙古锡林浩特 026000

收稿日期:2023-07-10 出版日期:2024-09-25 发布日期:2024-09-30
作者简介:*王大伟,E-mail:wangdawei02@caas.cn;
郭秀明(1981—),女,河北沧州人,博士,副研究员,研究方向为农业智能感知。E-mail:guoxiuming@caas.cn
通讯作者: *王大伟,E-mail:wangdawei02@caas.cn; 刘升平,E-mail:liushengping@caas.cn
基金资助:
内蒙古自治区科技计划(2022YFSJ0010);内蒙古自治区科技计划(2020GG0112);中国农业科学院平台提质增效项目(Y2021PT03);中国农业科学院创新工程(CAAS-ASTIP-2016-AII);中国农业科学院北方农牧业科技创新中心项目(BFGJ2022007);“天池英才”引进计划

Study on key problems for rat hole recognition and count near ground based on deep learning and its application

GUO Xiuming¹(), WANG Dawei²^,³^,⁴^,^*(), LIU Shengping¹^,^*(), ZHU Yeping¹, LIU Xiaohui², LIN Kejian⁴, WANG Jiayu⁵, LI Fei⁵

1. Key Laboratory of Agri-Information Service Technology, Ministry of Agriculture and Rural Affairs, Agricultural Information Institute, Chinese Academy of Agricultural Sciences, Beijing 100081, China
2. State Key Laboratory for Biology of Plant Diseases and Insect Pests, Institute of Plant Protection, Chinese Academy of Agricultural Sciences, Beijing 100081, China
3. Western Agricultural Research Center, Chinese Academy of Agricultural Sciences, Changji 831100, Xinjiang, China
4. Key Laboratory of Biohazard Monitoring and Green Prevention and Control in Artificial Grassland, Ministry of Agriculture and Rural Affairs, Institute of Grassland Research, Chinese Academy of Agricultural Sciences, Hohhot 010010, Inner Mongolia, China
5. Xilingol League Grassland Workstation, Xilinhot 026000, Inner Mongolia, China

Received:2023-07-10 Online:2024-09-25 Published:2024-09-30

摘要/Abstract

摘要：

鼠洞密度可用于评估草原鼠害发生程度。在近地面鼠洞图片采集与识别中,图像分辨率和拍摄倾角是影响时效性和识别性能的关键因素。为此,设计了带有倾角传感器的图像采集装置,在锡林郭勒草原采集2 325张鼠洞图片并进行手工标注,对比分析了3种图片输入尺寸(416 pixel×416 pixel、608 pixel×608 pixel、1 024 pixel×1 024 pixel)、4类拍摄倾角(21°、32°、41°、51°)、2种目标识别模型(YOLOv3和YOLOv4)对识别性能的影响。结果表明:YOLOv4模型在输入图片尺寸为416 pixel×416 pixel时能取得最优的性能。当拍摄倾角为41°时,识别性能最优;当拍摄倾角为32°时,识别性能最差。与近3年发表的鼠洞识别方法进行对比分析,验证了本文方法的性能先进性。研究结果可为草原鼠害的智能化监测提供技术支撑。

关键词: 目标检测, 草原生态, 机器视觉, 鼠洞, YOLOv3模型, YOLOv4模型

Abstract:

Rat hole density serves as a significant indicator for assessing the extent of rat damage in grasslands. In the image acquisition for rat hole recognition and count near ground by deep learning, the optimal image input size and shooting angle are crucial. To address these considerations, an image acquisition device equipped with an inclination sensor was designed, and a dataset of 2 325 rat hole images was collected from Xilingol grassland and manually annotated. The recognition performance was compared under 3 image input sizes (416 pixel×416 pixel, 608 pixel×608 pixel, 1 024 pixel×1 024 pixel), 4 shooting angles (21°, 32°, 41°, and 51°), and 2 target recognition models (YOLOv3 and YOLOv4). The findings revealed that YOLOv4 achieved superior performance when the image input size was set as 416 pixel×416 pixel. Furthermore, the recognition performance was optimal at the shooting angle of 41°, while it was the poorest at 32°. The proposed method was validated by comparing it with relevant approaches published within the past three years. These results offered valuable insights to support the development of intelligent monitoring technologies for assessing rat damage in grassland ecosystems.

Key words: object detection, grassland ecosystem, machine vision, rat hole, YOLOv3 model, YOLOv4 model

中图分类号:

S24
TP391.41

郭秀明, 王大伟, 刘升平, 诸叶平, 刘晓辉, 林克剑, 王佳宇, 李非. 基于深度学习的近地面草原鼠洞识别计数关键问题研究与应用[J]. 浙江农业学报, 2024, 36(9): 2146-2154.

GUO Xiuming, WANG Dawei, LIU Shengping, ZHU Yeping, LIU Xiaohui, LIN Kejian, WANG Jiayu, LI Fei. Study on key problems for rat hole recognition and count near ground based on deep learning and its application[J]. Acta Agriculturae Zhejiangensis, 2024, 36(9): 2146-2154.

图/表 9

图1 草原鼠洞图片采集设备的系统组成

Fig.1 Systematic composition of grassland rat hole image acquisition equipment

图2 草原鼠洞图片采集设备结构示意图

Fig.2 Sketch diagram of grassland rat hole image acquisition equipment

表1 不同图片尺寸下YOLOv3和YOLOv4的性能指标

Table 1 Performance indicators of YOLOv3 and YOLOv4 under different image sizes

图片尺寸 Image size	不同模型的精度 Precision of models/%		不同模型的召回率 Recall of models/%		不同模型的推理时间 Time consumption of models/ms
图片尺寸 Image size	YOLOv3	YOLOv4	YOLOv3	YOLOv4	YOLOv3	YOLOv4
416 pixel×416 pixel	89.4	98.2	89.3	91.8	22.8	32.0
608 pixel×608 pixel	90.7	96.0	94.0	90.2	36.4	54.0
1 024 pixel×1 024 pixel	94.6	97.3	98.7	90.6	80.7	120.0

表2 不同拍摄倾角下YOLOv4模型的鼠洞识别性能

Table 2 Performance of YOLOv4 under different shooting angles

拍摄倾角 Shooting angle/(°)	精度 Precision/%	召回率 Recall/%	F1分数 F1 score
21	81.7	81.8	0.896
32	80.1	80.3	0.887
41	91.8	92.1	0.940
51	87.3	89.9	0.920

表3 不同模型的鼠洞识别性能对比

Table 3 Comparison of performance of different models %

模型 Models	精度 Precision	召回率 Recall
参考文献[7]模型Model in Ref. [7]	91.30	94.08
参考文献[8]方法Model in Ref. [8]	53.32	77.12
参考文献[9]方法Model in Ref. [9]	91.00	77.00
参考文献[10]方法Model in Ref. [10]	91.30	85.90
本文方法Proposed model	98.02	91.80

图3 不同图片输入尺寸下YOLOv4的鼠洞识别效果

Fig.3 Rat hole recognition performance of YOLOv4 under different image sizes

图4 不同图片输入尺寸下YOLOv3的鼠洞识别效果

Fig.4 Rat hole recognition performance of YOLOv3 under different image sizes

图5 鼠洞识别及鼠害面积评估系统的照片

Fig.5 Photo of rat hole identification and rodent area assessment system

图6 鼠洞识别结果展示界面的照片

Fig.6 Photo of display view of rat hole identification results

参考文献 20

[1]	李燕妮, 袁帅, 付和平, 等. 典型草原啮齿动物密度与牧草损失量的关系[J]. 兽类学报, 2018, 38(4): 369-376.
	LI Y N, YUAN S, FU H P, et al. Relationship of rodent densities with forage loss in typical steppe[J]. Acta Theriologica Sinica, 2018, 38(4): 369-376. (in Chinese with English abstract) DOI
[2]	沈海花, 朱言坤, 赵霞, 等. 中国草地资源的现状分析[J]. 科学通报, 2016, 61(2): 139-154.
	SHEN H H, ZHU Y K, ZHAO X, et al. Analysis of current grassland resources in China[J]. Chinese Science Bulletin, 2016, 61(2): 139-154. (in Chinese with English abstract)
[3]	张涛, 杜健民, 张海军, 等. 基于无人机高光谱荒漠草原鼠洞识别方法研究[J]. 光电子·激光, 2022, 33(2): 120-126.
	ZHANG T, DU J M, ZHANG H J, et al. Research on recognition method of desert steppe rat hole based on unmanned aerial vehicle hyperspectral[J]. Journal of Optoelectronics·Laser, 2022, 33(2): 120-126. (in Chinese with English abstract)
[4]	温阿敏, 郑江华, 陈梦, 等. 荒漠生态林区大沙鼠鼠洞密度的无人机遥感监测技术初探[J]. 林业科学, 2018, 54(4): 186-192.
	WEN A M, ZHENG J H, CHEN M, et al. Monitoring mouse-hole density by Rhombomys opimus in desert forests with UAV remote sensing technology[J]. Scientia Silvae Sinicae, 2018, 54(4): 186-192. (in Chinese with English abstract)
[5]	周晓琳, 安如, 陈跃红, 等. 三江源典型区鼠洞无人机遥感识别研究[J]. 亚热带资源与环境学院, 2018, 13(4): 85-92.
	ZHOU X L, AN R, CHEN Y H, et al. Identification of rat holes in the typical area of “Three-River Headwaters” region by UAV remote sensing[J]. Journal of Subtropical Resources and Environment, 2018, 13(4): 85-92. (in Chinese with English abstract)
[6]	孙迪, 倪亦非, 陈吉军, 等. 应用无人机(UAV)低空影像监测黄兔尾鼠鼠洞初探[J]. 中国植保导刊, 2019, 39(4): 35-43.
	SUN D, NI Y F, CHEN J J, et al. Application of UAV low-altitude image on rat hole monitoring of Eolagurus luteus[J]. China Plant Protection. 2019, 39 (4): 35-43. (in Chinese with English abstract)
[7]	崔博超, 郑江华, 刘忠军, 等. 无人机遥感影像的YOLOv3鼠洞识别技术[J]. 林业科学, 2020, 56(10): 199-208.
	CUI B C, ZHENG J H, LIU Z J, et al. YOLOv3 mouse hole recognition based on remote sensing images from technology for unmanned aerial vehicle[J]. Scientia Silvae Sinicae, 2020, 56(10): 199-208. (in Chinese with English abstract)
[8]	周俗, 韩立亮, 杨思维, 等. 基于卷积神经网络的若尔盖草原鼠害监测应用研究[J]. 草学, 2021(2): 15-25.
	ZHOU S, HAN L L, YANG S W, et al. A study of rodent monitoring in Ruoergai grassland based on convolutional neural network[J]. Journal of Grassland and Forage Science, 2021(2): 15-25. (in Chinese with English abstract)
[9]	EZZY H, CHARTER M, BONFANTE A, et al. How the small object detection via machine learning and UAS-based remote-sensing imagery can support the achievement of SDG2: a case study of vole burrows[J]. Remote Sensing, 2021, 13(16): 3191.
[10]	DU M Z, WANG D W, LIU S P, et al. Rodent hole detection in a typical steppe ecosystem using UAS and deep learning[J]. Frontiers in Plant Science, 2022, 13: 992789.
[11]	李鹤. 基于无人机高光谱遥感的典型荒漠草原的鼠洞识别研究[D]. 呼和浩特: 内蒙古农业大学, 2021.
	LI H. Research on rat hole recognition of typical desertification grassland based on UAV hyperspectral remote sensing[D]. Hohhot: Inner Mongolia Agricultural University, 2021. (in Chinese with English abstract)
[12]	ZHANG X, HAN L X, HAN L H, et al. How well do deep learning-based methods for land cover classification and object detection perform on high resolution remote sensing imagery?[J]. Remote Sensing, 2020, 12(3): 417.
[13]	WAN J X, JIAN D F, YU D Z. Research on the method of grass mouse hole target detection based on deep learning[J]. Journal of Physics: Conference Series, 2021, 1952(2): 022061.
[14]	郭秀明, 诸叶平, 李世娟, 等. 农业复杂环境下尺度自适应小目标识别算法: 以蜜蜂为研究对象[J]. 智慧农业(中英文), 2022, 4(1): 140-149.
	GUO X M, ZHU Y P, LI S J, et al. Scale adaptive small objects detection method in complex agricultural environment: taking bees as research object[J]. Smart Agriculture, 2022, 4(1): 140-149. (in Chinese with English abstract) DOI
[15]	HEYDARI M, MOHAMADZAMANI D, PARASHKOUHI M G, et al. An algorithm for detecting the location of rodent-made holes through aerial filming by drones[J]. Archives of Pharmacy Practice, 2020, 11(1): 55-60.
[16]	REDMON J and FARHADI A. YOLOv3: an incremental improvement[EB/OL]. (2018-04-08) [2023-07-10]. https://arxiv.org/pdf/1804.02767.
[17]	BOCHKOVSKIY A, WANG C Y, LIAO H. YOLOv4: optimal speed and accuracy of object detection[EB/OL]. (2020-04-23) [2023-07-10]. https://arxiv.org/pdf/2004.10934.
[18]	WANG C Y, MARK LIAO H Y, WU Y H, et al. CSPNet: a new backbone that can enhance learning capability of CNN[C]// 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). Seattle, WA, USA: IEEE, 2020: 1571-1580.
[19]	HE K M, ZHANG X Y, REN S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916. DOI PMID
[20]	LIU S, QI L, QIN H F, et al. Path aggregation network for instance segmentation[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA: IEEE, 2018: 8759-8768.

编辑推荐

Metrics

阅读次数

全文

306

HTML			PDF

最新录用	在线预览	正式出版	最新录用	在线预览	正式出版
0	0	19	0	0	287

来源	本网站	其他网站

次数	238	68
比例	78%	22%

摘要

154

最新录用	在线预览	正式出版

0	0	154

	来源	本网站

	次数	154
	比例	100%

基于深度学习的近地面草原鼠洞识别计数关键问题研究与应用

Study on key problems for rat hole recognition and count near ground based on deep learning and its application

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 20

相关文章 10

编辑推荐

Metrics

本文评价

[1]	宁文楷, 李静, 沈晓东, 吴鑫, 李臻锋. 南瓜干燥过程中β-胡萝卜素的多源融合预测[J]. 浙江农业学报, 2023, 35(8): 1876-1887.
[2]	李斌, 刘东阳, 时国龙, 慕京生, 徐浩然, 辜丽川, 焦俊. 基于改进YOLOv4模型的群养生猪姿态检测[J]. 浙江农业学报, 2023, 35(1): 215-225.
[3]	闫宁, 张晗, 董宏图, 康凯, 罗斌. 基于透射光和反射光图像同位分割的小麦品种识别方法研究[J]. 浙江农业学报, 2022, 34(3): 590-598.
[4]	包晓敏, 盛家文. 粘虫板害虫自动识别计数研究[J]. 浙江农业学报, 2019, 31(9): 1516-1522.
[5]	伍渊远, 尚欣, 张呈彬, 谢新义. 自然光照下智能叶菜收获机作业参数的获取[J]. 浙江农业学报, 2017, 29(11): 1930-1937.
[6]	田海韬, 赵军, 蒲富鹏. 马铃薯芽眼图像的分割与定位方法[J]. 浙江农业学报, 2016, 28(11): 1947-1953.
[7]	任磊1，张俊2，*，陆胜民2. 脱囊衣橘片自动分拣机器视觉算法研究[J]. 浙江农业学报, 2015, 27(12): 2212-.
[8]	刘建军;姚立健;彭樟林. 基于机器视觉的山核桃等级检测技术[J]. , 2010, 22(6): 854-858.
[9]	梁琨;罗汉亚;沈明霞;*;何瑞银;张璐. 水稻播种质量检测技术的研究进展及展望[J]. , 2010, 22(2): 0-257.
[10]	王勇;沈明霞;姬长英. 基于色差信息的田间成熟棉花识别[J]. , 2007, 19(5): 0-388.