薛大暄,张瑞瑞,陈立平,陈梅香,徐刚,2020,基于Faster R-CNN的美国白蛾图像识别模型研究[J].环境昆虫学报,(6):1502-1509
基于Faster R-CNN的美国白蛾图像识别模型研究
Faster R-CNN based image recognition research of Hyphantria cunea
  
DOI:
中文关键词:  美国白蛾  图像识别  人工神经网络  深度学习
英文关键词:Hyphantria Cunea  image recognition  artificial neural network  deep learning
基金项目:国家自然科学基金面上项目(31971581);北京市农林科学院2020创新能力建设项目(KJCX20200108)
作者单位
薛大暄,张瑞瑞,陈立平,陈梅香,徐刚 1. 首都师范大学信息工程学院北京 1000482. 国家农业信息化工程技术研究中心北京 100097 
摘要点击次数: 44
全文下载次数: 45
中文摘要:
      昆虫监测中美国白蛾Hyphantria Cunea的人工辨识、分类费时费力,且主观性强。本文利用RPN人工神经网络模型对美国白蛾图像数据进行特征提取,并对比分析Inception_v2,ResNet50,ResNet101网络模型,设计了一种改进的美国白蛾人工神经网络识别模型IHCDM(Improved Hyphantria Cunea Artificial Neural Network Recognition Model,IHCDM),采用端到端方法在GPU处理器上对该模型进行了训练,并对其进行了实验验证。结果表明:该模型对美国白蛾的识别准确率可达99.5%,相比于ResNet50与ResNet101网络模型,识别准确率提高了0.5%与0.4%。超参数微调后,在置信度阈值为0.85时,识别准确率99.7%,识别速度0.09 ms/张。IHCDM模型为美国白蛾的快速辨识、分类提供了一种新方法。
英文摘要:
      Artificial identification and classification of Hyphantria Cunea in insect monitoring are time-consuming and laborious, and subjective. In this paper, the RPN artificial neural network model is used to extract the features of the Hyphantria Cunea image data, and the ResNet50 and ResNet101 network models are compared and analyzed. An improved artificial neural network Recognition model IHCDM of Hyphantria Cunea is designed (Improved Hyphantria Cunea Artificial Neural Network Recognition Model,IHCDM). The model is trained on GPU processor by end-to-end method, and verified by experiments. The results showed that the recognition accuracy of this model was 99.5%, which was 0.5% and 0.4% higher than that of ResNet50 and ResNet101. After Hyper-parameter adjustment, as if the confidence threshold was 0.75, the recognition accuracy was 99.7%, and the recognition speed was 0.09 ms/piece. The improved image recognition model provides a new method for the rapid identification and classification of Hyphantria Cunea.
查看全文  查看/发表评论  下载PDF阅读器
关闭