Classification model of spraying deposition on citrus canopy based on small-scale data set
-
摘要:目的
提升柑橘果园的智能化管理水平，快速无损获取柑橘树冠层的施药情况，改善小规模数据集导致施药情况分类模型易发生过拟合的问题。
方法提出一种基于卷积神经网络的柑橘树冠层施药情况分类模型——VGG_C模型。模型以VGG模型核心思想为基础进行构建，通过交叉熵损失函数优化，加速概率分布与真实分布的迭代过程，并在输出端引入不确定性度量计算以及在下采样模块中插入Droupout方法，降低由于数据较少而发生过拟合的概率。
结果VGG_C模型针对训练集的分类损失值为0.44%，比ResNet和VGG模型分别降低了87%和91%；准确率为95.3%，比ResNet和VGG模型分别提高了5%和10%；验证集的预测平均准确率为96.4%。
结论VGG_C模型通过多层卷积模型协同实现柑橘冠层热红外图像特征的高效提取，通过优化输出端结构提高了柑橘冠层施药情况分类模型在小数据集规模上的训练测试优度，可为柑橘树施药情况的智能化判断提供有效参考。
Abstract:ObjectiveThe study was aimed to improve the intelligent management level of citrus orchards, quickly and non-destructively evaluate the spraying quality on citrus canopy, and solve the overfitting problem of the spraying quality classification model caused by small-scale data set.
MethodWe proposed a classification model of spraying quality on citrus canopy based on convolutional neural network: Visual geometry group citrus model (VGG_C model). The model was constructed based on the core idea of the VGG model. Through optimization of the cross-entropy loss function, the iterative process of probability distribution and true distribution was accelerated. The uncertainty measurement calculation was introduced at the output end and the Droupout method was inserted in the downsampling module to reduce the probability of overfitting due to small amount of data.
ResultThe loss value of VGG_C model for the training set was 0.44%, which was 87% and 91% lower than that of ResNet and VGG respectively. The accuracy of VGG_C model for the training set was 95.3%, which was 5% and 10% higher than that of ResNet and VGG respectively. The average accuracy of the verification set was 96.4%.
ConclusionVGG_ C model can effectively extract the features of citrus canopy thermal infrared image through multi-layer convolution model and improve the training and testing superiority of citrus canopy application classification model in small data set by optimizing the output structure. VGG_ C model can provide an effective reference for the intelligent judgment of pesticide application on citrus trees.
-
-
![]()
图 4 3种模型的损失函数和准确率曲线图
Figure 4. The curves of loss function and accuracy rate of three models
表 1 试验环境数据
Table 1 Data of test environment
组别
Group
θ/℃
Temperature
相对湿度/%
Relative humidity
光照度/lx
Illuminance相机与植物的间距/cm
Distance between camera and plant1 20~23 45~47 800~1000 150 2 20~23 48~50 800~1000 150 3 20~23 51~53 800~1000 150 4 24~27 45~47 800~1000 180 5 24~27 48~50 800~1000 180 6 24~27 51~53 800~1000 180 7 28~31 45~47 800~1000 200 8 28~31 48~50 800~1000 200 9 28~31 51~53 800~1000 200 
下载: 导出CSV
-
[1] 兰玉彬, 朱梓豪, 邓小玲, 等. 基于无人机高光谱遥感的柑橘黄龙病植株的监测与分类[J]. 农业工程学报, 2019, 35(3): 92-100. doi: 10.11975/j.issn.1002-6819.2019.03.012 [2] 陆健强, 林佳翰, 邓小玲, 等. 多尺度特征融合的柑橘冠层施药沉积量分类模型[J]. 农业工程学报, 2020, 36(23): 70-76. doi: 10.11975/j.issn.1002-6819.2020.23.008 [3] GAO J F, FRENCH A P, POUND M P, et al. Deep convolutional neural networks for image-based Convolvulus sepium detection in sugar beet fields[J]. Plant Methods, 2020, 16(9): 108-116.
[4] STEINMETZ M, SOOSTEN D, HUMMEL J, et al. Validation of the RumiWatch Converter V0.7. 4.5 classification accuracy for the automatic monitoring of behavioural characteristics in dairy cows[J]. Archives of Animal Nutrition, 2020, 74(2): 164-172. doi: 10.1080/1745039X.2020.1721260
[5] 姜红花, 张传银, 张昭, 等. 基于Mask R-CNN的玉米田间杂草检测方法[J]. 农业机械学报, 2020, 51(6): 220-228. doi: 10.6041/j.issn.1000-1298.2020.06.023 [6] 何志谦, 胡明海, 丁黎明, 等. 基于智能四旋翼农业喷药机器人的设计与实现[J]. 山西电子技术, 2020(1): 15-17. [7] 李乐, 刘玉倩. 基于大数据的农业智能机器人开发[J]. 农机化研究, 2019, 41(11): 221-224. doi: 10.3969/j.issn.1003-188X.2019.11.038 [8] KISLAYA A, GOUTAM R. An autonomous UAV for pesticide spraying[J]. International Journal of Trend in Scientific Research and Development, 2019, 3(3): 986-990.
[9] RICARDO A D, SERGIO T D J, MARCELO C F. Deposit of pesticides without and with adjuvants on citrus seedlings following different intervals of artificial rain[J]. Ciencia Rural, 2016, 46(1): 13-19. doi: 10.1590/0103-8478cr20150064
[10] 奉山森, 张燕, 樊军庆, 等. 基于单片机的农药喷洒机械自动调平系统设计[J]. 农机化研究, 2014, 36(11): 104-107. doi: 10.3969/j.issn.1003-188X.2014.11.024 [11] WANG J, LAN Y B, YAO W X, et al. Aerial spraying application of multi-rotor unmanned aerial vehicle on areca trees[J]. International Journal Precision Agricultural Aviation, 2020, 3(4): 51-64.
[12] 周建民, 周其显, 刘燕德, 等. 红外成像技术在农产品加工自动化中的应用[J]. 中国农机化, 2010(6): 69-72. [13] 张小雨, 孙宏勇, 王艳哲, 等. 应用基于红外热画像技术的CWSI简化算法判断作物水分状态[J]. 中国农业气象, 2013, 34(5): 569-575. doi: 10.3969/j.issn.1000-6362.2013.05.011 [14] 张京, 何雄奎, 宋坚利, 等. 无人驾驶直升机航空喷雾参数对雾滴沉积的影响[J]. 农业机械学报, 2012, 43(12): 94-96. doi: 10.6041/j.issn.1000-1298.2012.12.017 [15] 尚垚睿. 人脸部件的提取和分类方法研究[D]. 北京: 中国人民公安大学, 2019. [16] 吴穗岚, 陈乐, 曾涛. 基于卷积神经网络的膝关节炎患者的膝盖红外图像分类[J]. 中国计量大学学报, 2019, 30(2): 185-190. doi: 10.3969/j.issn.2096-2835.2019.02.010 [17] 穆文秀, 洪蕾, 王瀚. 基于机器学习的智能昆虫分目识别算法应用[J]. 数字技术与应用, 2018, 36(11): 118-119. [18] 左羽, 陶倩, 吴恋, 等. 基于卷积神经网络的植物图像分类方法研究[J]. 物联网技术, 2020, 10(3): 72-75. [19] 崔洲涓, 安军社, 崔天舒. 基于多层深度卷积特征的抗遮挡实时跟踪算法[J]. 光学学报, 2019, 39(7): 229-242. [20] 徐超. 基于深度学习的信用卡逾期风险预测的方法研究与实现[D]. 苏州: 苏州大学, 2016. [21] 张伟. 基于深度卷积神经网络自学习特征的地表覆盖分类研究[D]. 北京: 中国科学院大学, 2017. [22] 董慧妍. 基于机器学习的糖尿病性视网膜病变图像分级研究[D]. 北京: 北京邮电大学, 2019.
