|
|
基于MobileNetV3的陈皮年份鉴别
|
Abstract:
陈皮(学名柑橘皮)的贮藏年限与其品质特征及市场价值呈现显著正相关性。然而，传统的陈皮年份鉴别方法主要依赖经验判断或化学分析，存在主观性强、操作复杂、耗时耗力等诸多局限。为此，本文提出一种基于轻量化深度学习框架MobileNetV3的陈皮年份智能鉴别方法。通过构建涵盖不同贮藏年限的陈皮图像数据集，采用区域裁剪对原始图像样本进行标准化处理，生成分辨率为224 × 224像素的黑色背景规范图像。在数据预处理与增强阶段，应用像素值归一化、随机旋转等策略，以提升模型的泛化能力与鲁棒性。实验环节对比了MobileNetV3、ResNet、DenseNet、FasterNet、AlexNet、VGG16以及EfficientFormerV2共七种经典与轻量级卷积神经网络模型。结果表明，MobileNetV3-large在模型参数量仅为16.3 MB的前提下，实现了86.96%的分类准确率与87.55%的精确率，单张图像推理时间仅为5.36毫秒，综合性能显著优于其他对比模型。总之，本文提出了一种更为简易且准确的陈皮年份识别方法，为在边缘计算设备的部署提供了高效可靠的技术支持。
The storage age of dried tangerine peel (Citrus reticulata peel) is significantly positively correlated with its quality characteristics and market value. However, traditional methods for identifying the age of dried tangerine peel mainly rely on expert judgment or chemical analysis, which are subjective, complex, time-consuming, and labor-intensive. To address these limitations, this paper proposes an intelligent method for identifying the age of dried tangerine peel based on the lightweight deep learning framework MobileNetV3. A dataset of tangerine peel images with varying storage ages was constructed, and regional cropping was used to standardize the original image samples, generating standardized images with a resolution of 224 × 224 pixels and a black background. During the data preprocessing and augmentation phase, strategies such as pixel value normalization and random rotation were applied to enhance the model’s generalization ability and robustness. The experimental section compares seven classic and lightweight convolutional neural network models, including MobileNetV3, ResNet, DenseNet, FasterNet, AlexNet, VGG16, and EfficientFormerV2. The results show that MobileNetV3-large achieves a classification accuracy of 86.96% and a precision of 87.55%, with a single image inference time of only 5.36 milliseconds, while having only 16.3 MB of model parameters. Its overall performance significantly outperforms the other models. In conclusion, this paper presents a simpler and more accurate method for identifying the age of dried tangerine peel, providing efficient and reliable technical support for deployment on edge computing devices.
| [1] | 国家药典委员会. 中华人民共和国药典(一部) [M]. 北京: 中国医药科技出版社, 2020. |
| [2] | Yin, J., Wang, Y., He, L., Jiang, W. and Liang, Y. (2022) Chemical Variation of Chenpi (Citrus Peels) and Corresponding Antioxidant Activities during the Aging Process. Frontiers in Nutrition, 9, Article ID: 825381. |
| [3] | Chen, Y., Sun, J., He, X., Yu, L. and Lin, Y. (2021) UPLC-Q-TOF-MS with Chemometrics Approach Analysis of Citri Reticulatae Pericarpium. Phytochemical Analysis, 32, 793-802. |
| [4] | Howard, A., Sandler, M., Chen, B., Wang, W., Chen, L., Tan, M., et al. (2019) Searching for MobileNetV3. 2019 IEEE/ CVF International Conference on Computer Vision (ICCV), Seoul, 27 October-2 November 2019, 1314-1324. https://doi.org/10.1109/iccv.2019.00140 |
| [5] | Pan, S.J. and Yang, Q. (2010) A Survey on Transfer Learning. IEEE Transactions on Knowledge and Data Engineering, 22, 1345-1359. https://doi.org/10.1109/TKDE.2009.191 |
| [6] | Sandler, M., Howard, A., Zhu, M., et al. (2018) MobileNetV2: Inverted Residuals and Linear Bottlenecks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, 18-22 June 2018, 4510-4520. https://doi.org/10.1109/CVPR.2018.00474 |
| [7] | Krizhevsky, A., Sutskever, I. and Hinton, G.E. (2012) ImageNet Classification with Deep Convolutional Neural Networks. Advances in Neural Information Processing Systems (NeurIPS), Vol. 25, 1097-1105. |
| [8] | Simonyan, K. and Zisserman, A. (2015) Very Deep Convolutional Networks for Large-Scale Image Recognition. International Conference on Learning Representations (ICLR), San Diego, 7-9 May 2015, 1-14. https://arxiv.org/abs/1409.1556 |
| [9] | He, K., Zhang, X., Ren, S. and Sun, J. (2016) Deep Residual Learning for Image Recognition. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, 27-30 June 2016, 770-778. https://doi.org/10.1109/cvpr.2016.90 |
| [10] | Huang, G., Liu, Z., Van Der Maaten, L. and Weinberger, K.Q. (2017) Densely Connected Convolutional Networks. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, 21-26 July 2017, 4700-4708. https://doi.org/10.1109/cvpr.2017.243 |
| [11] | Li, Y., Du, J., Song, G. and Qian, C. (2022) EfficientFormerV2: Smaller, Faster and Stronger. https://arxiv.org/abs/2212.08059 |
| [12] | Chen, Y., Dai, X., Liu, M., et al. (2023) FasterNet: Rethinking Bottleneck Structure for Real-Time Efficient Image Recognition. https://arxiv.org/abs/2206.03492 |
