摘要
针对木薯叶片疾病较难分辨、当前疾病检测机制智能化程度低的问题,提出了一种基于HSV色彩空间与EfficientNet的木薯疾病检测方法,设计了木薯疾病自动化监测与处理系统。首先利用HSV色彩空间进行图像预处理,提高目标区域的检测精度,减少图像预处理阶段的信息丢失;然后,引入改进型EfficientNet模型对预处理后的叶片图像进行训练,提取多维度的深度、宽度、分辨率特征,并利用几何方法将患病部位的坐标输出。最后,将该网络部署于嵌入式平台上,机械臂根据输出的坐标准确标定处理患病木薯。结果表明,疾病检测综合准确率为88.4%,其中木薯褐条病检测准确率可达96%,最低的木薯绿斑病可达82%。
Aiming at the problems that cassava leaf disease is difficult to distinguish and the current disease detection mechanism is low intelligent,this paper proposes a variety of cassava disease detection methods based on HSV color space and EfficientNet Networks,and designs an automatic monitoring and processing system for cassava disease.Firstly,the HSV color space is used for image preprocessing to improve the detection accuracy of the target area and reduce the information loss in the image preprocessing stage;Then,the improved EfficientNet model is introduced to train the preprocessed leaf images,extract the multi-dimensional depth,width and resolution features,and output the coordinates of the diseased parts using geometric methods.Finally,the network will be deployed on the embedded platform,and the robot arm will accurately calibrate and process the diseased cassava according to the output coordinates.The results showed that the comprehensive accuracy rate of disease detection was 88.4%,of which the accuracy rate of cassava brown streak disease was 96%,and the lowest cassava green spot disease was 82%.
作者
彭强
涂赛飞
赵中雨
王卓尔
王细桃
高菲
PENG Qiang;TU Sai-fei;ZHAO Zhong-yu;WANG Zhuo-er;WANG Xi-tao;GAO Fei(Undergraduate College,Wuhan University of Technology,Wuhan 430070,China;School of Information Engineering,Wuhan University of Technology,Wuhan 430070,China;Archives,Wuhan University of Technology,Wuhan 430070,China)
出处
《武汉理工大学学报》
CAS
2022年第9期95-100,共6页
Journal of Wuhan University of Technology
基金
大学生创新创业训练计划(202110497071).