摘要
为了提高水敏试纸图像处理算法评估喷雾质量的精确性。提出1种雾滴轮廓参数差异识别雾滴类型的方法,并基于机器视觉对K近邻分类模型、逻辑回归分类模型、决策树分类模型和支持向量机识别模型进行对比研究。以水代替农药利用无人植保飞机喷洒,选取10张不同稀疏程度的水敏试纸验证逻辑模型识别的准确性。结果表明,经接受者工作特征(receiver operating characteristic,ROC)曲线评估,ROC曲线与横轴围成的面积(area under the curve of ROC,AUC)取值以逻辑回归分类模型0.98最高。模型识别平均相对误差为4.05%,雾滴识别平均正确率为95.95%,最大相对误差6.62%。基于逻辑回归分类模型构建的雾滴图像处理算法能显著提高雾滴分辨准确率,快速了解田间农药分布情况,为后期田间植保作业提供有力的数据评估。
To improve the accuracy of the water-sensitive test paper image processing algorithm for assessing spray quality.A method for identifying the type of droplets by differences in droplet profile parameters is proposed,and a comparative study of K-nearest neighbor classification models,logistic regression classification models,decision tree classification models and support vector machine recognition models was conducted based on machine vision.Water was used instead of pesticide for spraying using unmanned plant protection aircraft,and 10 water-sensitive test papers with different sparsity were selected to verify the accuracy of the logistic model recognition.The AUC took the highest value of 0.98 for the logistic regression classification model as assessed by the receiver operating characteristic curve.The mean relative error in model recognition was 4.05%and the mean correct rate of fog drop recognition was 95.95%,with a maximum relative error of 6.62%.Droplet image processing algorithm based on logistic regression classification model can significantly improve the accuracy of droplet discrimination,quickly understand the distribution of pesticides in the field,and can provide powerful data evaluation for later field plant protection operations.
作者
张开飞
程上上
张志
丁力
赵弋秋
李赫
ZHANG Kaifei;CHENG Shangshang;ZHANG Zhi;DING Li;ZHAO Yiqiu;LI He(College of Mechanical and Electrical Engineering,Henan Agricultural University,Zhengzhou 450002,China;School of Agricultural Engineering, Jiangsu University, Zhenjiang 212013, China)
出处
《河南农业大学学报》
CAS
CSCD
2021年第6期1109-1117,共9页
Journal of Henan Agricultural University
基金
现代农业产业技术体系建设专项资金项目(CARS-04)
河南省科技攻关项目(212102110219)。
关键词
雾滴图像
轮廓参数
分类模型
机器视觉
ROC曲线
droplet image
shape parameter
classification model
machine vision
ROC curve
