Detection of Pine Trees Infected by Pine Wood Nematode with UAV Images Based on Improved SSD

doi:10.13466/j.cnki.lyzygl.2022.03.021

Abstract

Abstract:

It is an effective means to prevent the further spread of pine wilt disease by quickly and accurately obtaining the number and location information of pine wilt discoloration wood and combining with the corresponding control measures.The centimeter-level image of pine wilted wood forest area was obtained by UAV,and the automatic detection of discoloration wood was realized by SSD,YOLO v4 and Faster R-CNN.The experimental results showed that compared with YOLO v4 and Faster R-CNN,the overall accuracy of SSD for discoloration wood was 75.0%.A method combining feature fusion module and channel attention mechanism module was proposed to improve SSD.The overall detection accuracy of improved SSD was 79.0%,which was 4.0% higher than that of SSD,indicating that the improved SSD was more suitable for discoloration wood detection than SSD.In addition,the number of discoloration trees in the verification area was 87,and the number of correct detection trees in the improved SSD was 81.The accuracy of detection was as high as 93.1%.The accurate detection of pine wilt discoloration trees in UAV images of forest areas was realized,which can provide technical support for the prevention and control of pine wilt discoloration trees.

Key words: Bursaphelenchus xylophilus, feature fusion, squeeze and excitation block, improved SSD

CLC Number:

S763
S771.8

LIU Shunli, LIU Changhua, ZHANG Lei, PENG Ciqing, XUE Dongdong. Detection of Pine Trees Infected by Pine Wood Nematode with UAV Images Based on Improved SSD[J]. FOREST RESOURCES WANAGEMENT, 2022, 0(3): 135-141.

Figures/Tables 4

References 21

[1]	杨宝君, 潘宏阳, 汤坚, 等. 松材线虫病[M]. 北京: 中国林业出版社, 2003.
[2]	国家林业和草原局公告[EB/OL].(2022-03-14)[2022-04-01]. http://www.forestry.gov.cn/main/3457/20220318/111403492220586.html.
[3]	蒋丽雅, 江顺利, 包锦芳. 松材线虫病疫情人工地面调查方法[J]. 安徽农业科学, 2006(13):3053-3054.
[4]	Kim S R, Lee W K, Lim C H, et al. Hyperspectral analysis of pine wilt disease to determine an optimal detection Index[J]. Forests, 2018, 9(3):115. doi: 10.3390/f9030115
[5]	宋以宁, 刘文萍, 骆有庆, 等. 基于线性谱聚类的林地图像中枯死树监测[J]. 林业科学, 2019, 55(4):187-195.
[6]	胡根生, 张学敏, 梁栋, 等. 基于加权支持向量数据描述的遥感图像病害松树识别[J]. 农业机械学报, 2013, 44(5):258-263.
[7]	李卫正, 申世广, 何鹏, 等. 低成本小型无人机遥感定位病死木方法[J]. 林业科技开发, 2014, 28(6):102-106.
[8]	陶欢, 李存军, 谢春春, 等. 基于HSV阈值法的无人机影像变色松树识别[J]. 南京林业大学学报:自然科学版, 2019, 43(3):99-106.
[9]	刘遐龄, 程多祥, 李涛, 等. 无人机遥感影像的松材线虫病危害木自动监测技术初探[J]. 中国森林病虫, 2018, 37(5):16-21.
[10]	张瑞瑞, 夏浪, 陈立平, 等. 基于U-Net网络和无人机影像的松材线虫病变色木识别[J]. 农业工程学报, 2020, 36(12):61-68.
[11]	徐信罗, 陶欢, 李存军, 等. 基于Faster R-CNN的松材线虫病受害木识别与定位[J]. 农业机械学报, 2020, 51(7):228-236.
[12]	李凤迪, 申卫星, 吴杰芳, 等. 基于YOLOv3-CIoU的松材线虫病树检测方法研究[J]. 山东农业大学学报:自然科学版, 2021, 52(2):224-233.
[13]	黄丽明, 王懿祥, 徐琪, 等. 采用YOLO算法和无人机影像的松材线虫病异常变色木识别[J]. 农业工程学报, 2021, 37(14):197-203.
[14]	成喆, 吕京国, 白颖奇, 等. 结合RPN网络与SSD算法的遥感影像目标检测算法[J]. 测绘科学, 2021, 46(4):75-82.
[15]	张泽宁, 郑树彬, 李立明. 基于SSD的轨道扣件快速弹条状态检测[J]. 计算机与数字工程, 2021, 49(8):1560-1565.
[16]	王震, 张晓丽, 安树杰. 松材线虫病危害的马尾松林木光谱特征分析[J]. 遥感技术与应用, 2007(3):367-370.
[17]	徐华潮, 骆有庆, 张廷廷, 等. 松材线虫自然侵染后松树不同感病阶段针叶光谱特征变化[J]. 光谱学与光谱分析, 2011, 31(5):1352-1356.
[18]	Liu W, Anguelov D, Erhan D, et al. SSD:Single shot multibox detector[C]. European conference on computer vision.Springer,Cham, 2016:21-37.
[19]	Andrea V, And Rew Z. VGG Convolutional Neural Networks Practical[EB/OL].[2022-04-01]. https://www.robots.ox.ac.uk/-vgg/practicals/cnn/index.html.
[20]	Bochkovskiy A, Wang C Y, Liao H Y M. Yolov4:Optimal speed and accuracy of object detection[J/OL]. https://doi.org/10.48550/arXiv.2004.10934.
[21]	Ren S, He K, Girshick R, et al. Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6):1137-1149. doi: 10.1109/TPAMI.2016.2577031

算法	IOU阈值	精确率/%	召回率/%	平均精度/%	总体精度/%	每秒检测图片数量/s
YOLO v4	0.50	90.53	60.78	77.44	73.11	39.68
Faster R-CNN	0.50	60.11	76.68	73.98	67.23	6.37
SSD	0.50	78.86	71.85	79.43	75.00	57.89
改进SSD	0.50	86.09	73.33	80.40	79.00	40.81