中国农业科学院农业信息研究所 中国农学会
Agricultural Information Institute of CAAS China Association of Agricultural Science Societies
En
Ch
中国农业科学院农业信息研究所 中国农学会 主办

AI检索

订阅

返回顶部
当前位置: 首页 > 文章 > 夜间环境下树上柑橘表征缺陷深度学习检测方法 林业工程学报 2021 (6) 148-155
Position: Home > Articles > Fluorescence detection of citrus characterization defects using SVM Journal of Forestry Engineering 2021 (6) 148-155

夜间环境下树上柑橘表征缺陷深度学习检测方法

作  者：
孙宝霞;梁翠晓;刘凯;郑镇辉;胡文馨;熊俊涛
单  位：
广东机电职业技术学院电气技术学院;华南农业大学数学与信息学院
关键词：
图像处理;夜间图像;柑橘识别;荧光检测
摘  要：
利用基于深度学习的视觉检测技术对夜间自然环境下成熟柑橘进行识别与表征缺陷检测,从而实现柑橘的产量估计与生长期品质监测,对柑橘果园的生产智能化管理有重要意义。本研究利用YOLO v4深度学习模型,进行夜间树上柑橘表征的视觉检测。首先设计了多光源结合的视觉系统进行夜间树上柑橘图像采集,对采集的柑橘图像进行人工标记,建立夜间柑橘图像的训练集和测试集,通过试验确定训练模型的批处理量和初始学习率,并在训练模型时根据训练次数逐渐降低学习率,最终训练模型在测试集的准确率为95.32%。通过设计的夜间柑橘视觉检测试验结果表明:本研究采用的YOLO v4模型检测方法在柑橘测试集上的精确率、召回率、F_1值以及mAP(mean average precision)值分别为95.32%、94.59%、0.95和90.52%,相比Faster R-CNN,YOLO v4模型检测方法精确率和召回率分别提高了5.14%和6.16%,同时检测速度明显快于Fater R-CNN,表明该方法对夜间自然环境树上柑橘的识别和缺陷检测有较好的准确性和实时性,满足室外夜间环境柑橘荧光缺陷检测的精确性要求,可为农业智能化生产中果蔬产量的估计提供技术支持。
译  名：
Fluorescence detection of citrus characterization defects using SVM
作  者：
SUN Baoxia;LIANG Cuixiao;LIU Kai;ZHENG Zhenhui;HU Wenxin;XIONG Juntao;School of Electrical Engineering, Guangdong Mechanical and Electrical Polytechnic;College of Mathematics and Informatics, South China Agricultural University;
关键词：
image processing;;nighttime image;;citrus recognition;;fluorescence detection
摘  要：
Visual detection technology based on deep learning is used to detect characterization defects of mature citrus in natural environments at nighttime to estimate the yield and monitor quality of the products during the growing period, which are of great significance to the research and development of citrus defect detection. In this study, YOLO v4 and deep learning model were used to conduct visual detection of citrus representation on trees at night. In the first place, a multi-source vision system was used for collecting the citrus ultraviolet images from the natural environments in the night, and the acquisition of the citrus image was manually tagged. The training set and test set of citrus images at night were established. A total of 139 samples were obtained from ultraviolet images, in which, 65% of samples were used as training set and 35% were test set. The training model of batch processing and the initial vector were determined through the test, and the number of training was gradually reduced, and the accuracy of the final training model in the test set was 95.32%. Then the SVM model was used to identify the mature citrus regional and to test the defects, finally the defected citrus region labels were outputted. The experimental results of citrus vision detection at night showed that: the accuracy, recall, F_1 and mAP(mean average precision) values of the YOLO v4 model detection method adopted in this study on the citrus test set were 95.32%, 94.59%, 0.95 and 90.52%, respectively. Compared with the Faster R-CNN model, the detection accuracy rate and recall rate of the YOLO v4 model were increased by 5.14% and 6.16%, respectively. The detection speed was significantly faster than that of the Fater R-CNN model, indicating that this proposed method had better accuracy and real-time performance in the identification and defect detection of citrus trees in natural environments at night, which met the accuracy requirements of fluorescence defect detection of citrus trees in outdoor night environment. It can provide technical support for the estimation of fruit and vegetable yield in the agricultural intelligent production.

相似文章

导出引用
计量
文章访问数: 22
HTML全文浏览量: 0
PDF下载量: 0

所属期刊

推荐期刊

友情链接Friendly link
农业信息研究所Agricultural information institute 农业智能知识服务平台Agricultural intelligent knowledge service platform 农业学术服务平台Agricultural Knowledge Service System 中国知网CNKI 维普网VIP 万方WANFANG DATA
联系我们Contact us

电话Tell：010-82109652-808

邮箱E-mail：agrijournal@caas.cn

服务Service
关于我们About us 意见建议Opinions and suggest
版权说明Copyright
单位：中国农业科学院农业信息研究所 中国农学会 © 2017中国农业科学院农业信息研究所 ICP备案号：京ICP备09089781号-46