基于深度学习的番茄叶部病害识别模型

番茄叶病种类多、成因复杂,其预防和识别难度较大。传统基于机器学习的方法多靠人工识别,需要一定的专家经验,且具有主观性强、识别准确率不高等缺点。为实现番茄叶病特征的自动提取,并提高识别准确率,提出一种基于深度学习的番茄叶病识别模型。该模型基于卷积神经网络对番茄叶部病害特征进行自动提取,获得高维特征后,采用PCA降维算法去除冗余特征;从增大类间距离并减小类内距离的角度改进了softmax分类器,提高了识别准确率。将该模型在CrowdAI提供的数据集上进行了仿真验证,结果表明,该模型能够对番茄叶部常见10种病害进行自动提取特征和识别,综合识别准确率达到95%以上。

Recognition Model of Tomato Leaf Diseases Based on Deep Learning

Tomato leaf disease has many kinds and complicated causes, and its prevention and identification are difficult. Traditional methods based on machine learning rely on manual recognition and require some expert experience, and have the disadvantages of strong subjectivity and low recognition accuracy. In order to realize automatic extraction of tomato leaf disease features and improve recognition accuracy, a recognition model based on depth learning is proposed. The model automatically extracts tomato leaf disease features based on convolution neural network, and after obtaining high-dimensional features, PCA dimensionality reduction algorithm is used to remove redundant features. The softmax classifier is improved by the way of increasing the distance between classes and decreasing the distance within classes, thus improving the recognition accuracy. The model is verified by simulation on the data set provided by CrowdAI. The results show that the model can automatically extract features and identify10 common diseases of tomato leaves, and the comprehensive identification accuracy rate is above 95%.