| 摘 要: 为了解决蘑菇图像分类的问题,实现野生菌毒性快速识别,以7种蘑菇作为研究对象,提出了一种基于深 度学习的蘑菇图像分类的方法。所提出的分类方法在考虑了自然场景图像的特点下,利用图像像素信息进行特征提取, 提取到的特征向量具有可辨别性、独立性和鲁棒性;轻量级ShuffleNetV2模型与作为其他常用CNN模型相比具有更高 的精度。实验表明,基于ShuffleNetV2的蘑菇分类模型的Top-1准确率为55.18%,Top-5准确率为93.55%,能够一定 程度上解决蘑菇图像分类困难的问题。未来结合移动设备和嵌入式开发,将能够用于自然环境下野生菌的自动分类,为 蘑菇产业智能化和自动化提供新的思路。 |
| 关键词: 深度学习;图像分类;蘑菇;卷积神经网络 |
|
中图分类号: TP399
文献标识码: A
|
|
| Research on Mushroom Image Classi cation based on Deep Learning |
|
XIAO Jiewen1, ZHAO Chengbo1, LI Xinjie2, LIU Zhongyu3, PANG Bo3, YANG Yihua4, WANG Jianxin3
|
( 1.College of Environmental Science and Engineering, Beijing Forestry University, Beijing 100083, China;
2.College of Science, Beijing Forestry University, Beijing 100083, China;
3.School of Information Science and Technology, Beijing Forestry University, Beijing 100083, China;
4.Guizhou institute of biology, Guizhou Academy of Sciences, Guiyang 550009, China)
xjw17@bjfu.edu.cn; z840922704@bjfu.edu.cn; lixj_6117@bjfu.edu.cn; lzy@noqaqs.cn; pxfwxhpb@live.com; wangjx@bjfu.edu.cn; yangyh_009@163.com
|
| Abstract: In order to cope with the problem of mushroom image classi cation and facilitate the rapid identi cation of wild mushroom toxicity, a method of mushroom image classi cation based on deep learning is proposed for seven kinds of mushrooms. The proposed classi cation method takes into account the characteristics of natural scene images and uses image pixel information for feature extraction. The extracted feature vectors are distinguishable, independent and robust; the lightweight Shuf eNetV2 model is used and has higher accuracy compared with other commonly used models such as CNN. Experiment results show that the mushroom classi cation model based on Shuf eNetV2 has a Top-1 accuracy of 55.18% and a Top-5 accuracy of 93.55%, which can solve, to some extent, the dif culty of mushroom image classi cation. If combined with mobile devices and embedded development in the future, it can be used for automatic classi cation of wild fungi in natural environments. |
| Keywords: deep learning; image classification; mushroom; convolutional neural networks |
