深度学习在园林植物学教学中的应用  

Application of Deep Learning in the Teaching of Horticulture and Forestry Botany

作　　者：路海兰[1] LU Hai-lan(Zhongshan College of Vocation and Technology,Nanjing,Jiangsu 210049)

机构地区：[1]钟山职业技术学院,江苏南京210049

出　　处：《安徽农业科学》2020年第24期238-241,共4页Journal of Anhui Agricultural Sciences

摘　　要：采用深度学习算法实现植物图片的自动识别,并将该算法嵌入手机APP,应用到园林植物学教学中,辅助教师的教学工作和学生课后的自主学习活动。算法中数据采集与标记采用园林植物专家筛选和标记,数据处理采用随机增强算法避免过拟合,卷积神经网络采用MobileNet以适用于移动终端,损失函数采用正例损失和负例损失,参数优化采用Adam算法。最终训练结果达到90%以上的精度,满足实用需要,达到优化园林植物学课程的课堂与课后教学活动的效果,有效地解决植物识别在种类数量、准确度和速度上的难点。The application of deep learning algorithm was used for automatic recognition of plant pictures,and the algorithm was embeded in mobile APP to horticulture and forestry botany teaching,assisting teachers’teaching and students’self-learning activities after class.In the algorithm,data acquisition and labeling were screened and labeled by experts,data processing was enhanced randomly to avoid over-fitting,convolution neural network was used for mobile terminals,loss function includes positive loss and negative loss,and parameter optimization was based on Adam algorithm.The final training results reached more than 90%accuracy and met the practical needs,so as to optimize the effect of classroom and after-class teaching activities.Deep learning could effectively solve the difficulties of plant recognition in the number,accuracy and speed of species.

关 键 词：深度学习 卷积神经网络 园林植物学 植物识别 教学辅助 

分 类 号：S-058[农业科学]