基于卷积神经网络的花卉识别技术

Flower Recognition Based on Convolutional Neural Networks

由于花卉所处背景复杂和本身类别之间的相似性,传统的人工提取特征进行图像识别的方法不能很好的解决其识别问题。随着科技的发展与进步,深度学习逐渐步入图像识别问题中,并取得了很好的成绩。本文针对现在主流的卷积神经网络深度较深而存在的参数多、训练时间长、收敛缓慢的缺陷,提出了一种基于卷积神经网络的花卉图像识别模型。该模型借鉴LeNet-5网络进行搭建,利用Dropout和Adam优化算法,减少过拟合,加速网络收敛。最终在自建数据集上进行训练后,测试集的准确率达到了92.76%。实验结果表明,本文提出的方法具有收敛速度快、识别准确率高的特点。

Due to the complexity of the background in which the flowers are located and the similarity between their own categories, the traditional method of manually extracting features for image recognition cannot solve its recognition problem well. With the development and progress of science and technology, deep learning has gradually stepped into the image recognition problem and achieved good results. This paper proposes a flower image recognition model based on convolutional neural network for the defects of more parameters, long training time and slow convergence of the mainstream convolutional neural network with deeper depth. The model is constructed by drawing on LeNet-5 network, using Dropout and Adam optimization algorithms to reduce overfitting and accelerate network convergence. Finally, after training on the self-constructed dataset, the accuracy of the test set reaches 92.76%. The experimental results show that the method proposed in this paper is characterized by fast convergence speed and high recognition accuracy.