[1]陈旭君,王承祥,朱德泉,等.基于YOLO卷积神经网络的水稻秧苗行线检测[J].江苏农业学报,2020,(04):930-935.[doi:doi:10.3969/j.issn.1000-4440.2020.04.017]
　CHEN Xu-jun,WANG Cheng-xiang,ZHU De-quan,et al.Detection of rice seedling row lines based on YOLO convolutional neural network[J].,2020,(04):930-935.[doi:doi:10.3969/j.issn.1000-4440.2020.04.017]
点击复制

基于YOLO卷积神经网络的水稻秧苗行线检测()
分享到：

江苏农业学报[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2020年04期
页码:
930-935
栏目:
耕作栽培·资源环境
出版日期:
2020-08-31

文章信息/Info

Title:
Detection of rice seedling row lines based on YOLO convolutional neural network
作者:
陈旭君1; 王承祥1; 朱德泉1; 刘晓丽1; 邹禹2; 张顺1; 廖娟1
(1.安徽农业大学工学院,安徽合肥230036;2.安徽省农业科学院水稻研究所，安徽合肥230031)
Author(s):
CHEN Xu-jun1; WANG Cheng-xiang1; ZHU De-quan1; LIU Xiao-li1; ZOU Yu2; ZHANG Shun1; LIAO Juan1
(1.School of Engineering， Anhui Agricultural University， Hefei 230036， China；2.Rice Research Institute， Anhui Academy of Agricultural Sciences， Hefei 230031， China)
关键词:
视觉导航; 秧苗行线; YOLO网络; 秧苗行定位点
Keywords:
visual navigation; rice seedling row lines; YOLO network; locating points of seedling row
分类号:
S126
DOI:
doi:10.3969/j.issn.1000-4440.2020.04.017
文献标志码:
A
摘要:
为了提高插秧机视觉导航精度，提出了一种基于YOLO卷积神经网络的水稻秧苗行线检测方法。首先对原始图像进行裁剪和拉伸预处理，划分为子图,然后由YOLO网络训练子图，根据秧苗目标框的位置信息确定秧苗行定位点，并拼接子图定位点获取全图秧苗行定位点，连接秧苗行定位点生成秧苗行线，计算相邻定位点间线段斜率以获得行线角度。试验结果表明，YOLO网络的秧苗区域检测性能优于Faster R-CNN和ResNet101，且秧苗行线的检测准确率高于Hough变换和聚类算法。该方法泛化性能强，可以准确检测秧苗行线，为插秧机视觉导航的路径规划提供可靠的定位信息。
Abstract:
In order to improve the positioning accuracy in visual navigation systems of rice transplanter, a detection method of rice seedling row lines based on YOLO convolutional neural network was proposed in this study. Each original image was firstly divided into sub-images by clipping and stretching preprocessing operations. According to the position information of seedling target blobs, the locating points of seedling row in the sub-images were determined. And the seedling row positioning points for sub-images were jointed to get the seedling row positioning points for the original image. Afterwards, the positioning points were connected to be the seedling lines, and the slope between two positioning points was calculated to get the line angle. The experimental results show at that the YOLO network had the advantage of detection performance and could accurately extract seedling target blobs, compared with Faster R-CNN and ResNet101. In addition, the detection accuracy of seedling lines was better than that of Hough transform and clustering algorithm. This method has strong generalization performance, can detect accurately seedling lines, and can provide reliable location information for the path planning in visual navigation systems of rice transplanter.

参考文献/References:

[1]谭晨佼,李轶林,王东飞,等.农业机械自动导航技术研究进展[J].农机化研究,2020,42(5):7-14，32.
[2]周航,杜志龙,武占元,等.机器视觉技术在现代农业装备领域的应用进展[J].中国农机化学报,2017,38(11):86-92.
[3]姬长英,周俊.农业机械导航技术发展分析[J].农业机械学报,2014,45(9):44-54.
[4]ARAVIND K R, RAJA P, PEREZ M. Task-based agricultural mobile robots in arable farming: A review [J]. Spanish Journal of Agricultural Research, 2017, 15(1): 1-16.
[5]彭顺正,坎杂,李景彬. 矮化密植枣园收获作业视觉导航路径提取[J].农业工程学报, 2017, 33(9) : 45-52.
[6]蔡道清,李彦明,覃程锦,等.水田田埂边界支持向量机检测方法[J].农业机械学报,2019,50(6):22-27，109.
[7]刁智华,赵明珍,宋寅卯,等.基于机器视觉的玉米精准施药系统作物行识别算法及系统实现[J].农业工程学报，2015,31(7):47-52.
[8]姜国权,杨小亚,王志衡,等.基于图像特征点粒子群聚类算法的麦田作物行检测[J].农业工程学报,2017,33(11):165-170.
[9]赵瑞娇,李民赞,张漫,等.基于改进Hough变换的农田作物行快速检测算法[J].农业机械学报,2009,40(7):163-165，221.
[10]廖娟,汪鹞,尹俊楠,等.基于分区域特征点聚类的秧苗行中心线提取[J].农业机械学报,2019,50(11):34-41.
[11]孟笑天,徐艳蕾,王新东,等.基于改进K均值特征点聚类算法的作物行检测[J].农机化研究,2020,42(8):26-30.
[12]张善文,谢泽奇,张晴晴. 卷积神经网络在黄瓜叶部病害识别中的应用[J]. 江苏农业学报, 2018, 34(1): 56-61.
[13]KAMILARIS A, PRENAFETA-BOLDU F X. Deep learning in agriculture: A survey [J]. Computers and Electronics in Agriculture, 2018, 147: 70-90.
[14]王丹丹，何东建. 基于R-FCN深度卷积神经网络的机器人疏果前苹果目标的识别[J]. 农业工程学报，2019，35(3):156-163.
[15]BODHWANI V, ACHARJYA D P, BODHWANI U. Deep residual networks for plant identification[J]. Procedia Computer Science, 2019, 152: 186-194.
[16]杨观赐,杨静,苏志东,等.改进的YOLO特征提取算法及其在服务机器人隐私情境检测中的应用[J].自动化学报,2018,44(12):2238-2249.

备注/Memo

备注/Memo:
收稿日期：2020-03-31基金项目：安徽省重点研发计划项目(1804a07020111);安徽省科技重大专项(18030701204); 安徽省自然科学基金项目(1808085ME158)；安徽省高等学校自然科学研究项目(KJ2017A134）;安徽省大学生创新创业教育训练计划项目(201910364083)作者简介：陈旭君(1999-)，男，湖北大冶人，本科生，研究方向为机器学习与图像处理, (E-mail)chenxujun173@163.com通讯作者：廖娟，(E-mail)liaojuan308@163.com
更新日期/Last Update: 2020-09-08