锐化处理和优化器对VGG16在印文种类鉴别中影响的研究  

Study on the Effect of Image Sharpening and Optimizer on VGG16 in Seal Impression Identification

作　　者：张询 于彬[1] ZHANG Xun;YU Bin(School of Forensic Science and Technology,Criminal Investigation Police University of China,Shenyang 110035,China)

机构地区：[1]中国刑事警察学院刑事科学技术学院,辽宁沈阳110035

出　　处：《中国人民公安大学学报(自然科学版)》2025年第1期7-16,共10页Journal of People’s Public Security University of China(Science and Technology)

基　　金：公安部科技强警基础工作专项项目(2018GABJC05);公安部科技强警基础工作专项项目(2019GABJ04)。

摘　　要：探讨了图像锐化处理和优化器选择对VGG16模型在印章印文种类鉴别任务中的影响。结合OpenCV进行图像预处理和锐化处理,评估了SGD与Adam优化器的性能表现。实验结果表明,图像锐化显著提升了模型的特征提取能力和分类精度,而SGD优化器在验证集和测试集上的准确率均超过95%,且在收敛速度和稳定性方面优于Adam。因此,适当的图像锐化处理和优化器选择能够有效提高VGG16模型的分类性能,未来可通过扩展数据集和优化算法进一步提升印章印文识别的效率和鲁棒性。The impact of image sharpening and optimizer selection on the performance of the VGG16 model for stamp inscription classification tasks were explored.By combining image preprocessing and sharpening with OpenCV,the performance of the SGD and Adam optimizers were evaluated.The experimental results demonstrated that image sharpening significantly enhanced the feature extraction capability and classification accuracy of the model.Additionally,the SGD optimizer achieved an accuracy of over 95%on both the validation and test sets,with superior convergence speed and stability compared to Adam.Therefore,appropriate image sharpening and optimizer selection could effectively improve the classification performance of VGG16 model,and the stamp inscription recognition efficiency and robustness could be improved through dataset expansion and algorithm optimization in the future.

关 键 词：印章印文检验 卷积神经网络 OPENCV VGG16模型 图像锐化 优化器 

分 类 号：D918.92[政治法律—法学]