| 基于二值化卷积神经网络的农业病虫害识别 |
| |
| 引用本文： | 蒲秀夫,宁芊,雷印杰,陈炳才.基于二值化卷积神经网络的农业病虫害识别[J].中国农机化学报,2020(2):177-182. |
| |
| 作者姓名： | 蒲秀夫 宁芊 雷印杰 陈炳才 |
| |
| 作者单位： | 四川大学电子信息学院;新疆师范大学物理与电子工程学院;大连理工大学计算机科学与技术学院 |
| |
| 基金项目： | 新疆自治区区域协同创新专项(科技援疆计划)(2019E0214)。 |
| |
| 摘 要： | 卷积神经网络的使用提高物体识别的准确率,但还是存在运行时间长、参数量较大的问题。针对这些问题提出使用二值化卷积神经网络模型对植物病虫害进行识别。试验中以VGG16模型为基准,采用深度网络模型对植物病虫害进行分类,相比于传统植物分类方法有效提高准确率;以符号函数和尺度因子α代替浮点型权值参数,将权值二值化以提高模型的计算速度;采用了PlantVillage数据集共54 306张图片,并且设置不同比例、环境下的数据集用以排除固有偏差对实验的影响,并且对原数据集进行图片扩充以消除样本分布不均的情况。试验表明二值化模型达到原模型近两倍的计算速度,且在分割数据集下测试平均识别准确率能达到96.8%。
|
| 关 键 词： | 农业病害 深度学习 二值化模型 图像分类 |
Identification of agricultural plant diseases based on binarized convolutional neural network |
| |
| Pu Xiufu,Ning Qian,Lei Yinjie,Chen Bingcai.Identification of agricultural plant diseases based on binarized convolutional neural network[J].Chinese Agricultural Mechanization,2020(2):177-182. |
| |
| Authors: | Pu Xiufu Ning Qian Lei Yinjie Chen Bingcai |
| |
| Institution: | (College of Electronics and Information,Sichuan University,Chengdu,610065,China;College of Physics and Electronics,Xinjiang Normal University,Urumqi,830054,China;College of Computer Science and Technology,Dalian University of Technology,Dalian,116024,China) |
| |
| Abstract: | The accuracy of object recognition is improved by using convolutional neural networks, but there are still some problems, such as long running time and large parameter quantities. In view of these problems, a binary convolutional neural network model which is used to identify plant diseases is proposed. Based on the VGG16 model, using depth network model to classify plant diseases can effectively improve the accuracy. Symbolic function and scale factor α are used instead of floating-point weight parameters,it can improve the calculation speed of the model. And in order to eliminate the influence of inherent deviation on experiments, a dataset from Plantvillage used in this paper includes 54 306 images with different color and grayscale, also the dataset has different proportions and environments. Experiments show that the computational speed of the binary model is almost twice as fast as that of the original model, and the average recognition accuracy can reach 96.8% under the segmented datasets. |
| |
| Keywords: | agricultural diseases deep learning binarized model image classification |
| 本文献已被 维普 等数据库收录！ |
|
