中国人民公安大学学报（自然科学版）

印章印文鉴定是文书检验研究的关键环节，旨在通过对印章或印文进行同一认定辨别文件真伪，进而为案情研判及侦查范围的确定提供支撑。当前，印章真伪鉴定与材质分类普遍依赖于人工对印迹特征的分析，其鉴定成本高、检验周期长且准确性易受主观因素的影响。针对上述问题，基于YOLOv8模型提出一种轻量化快速鉴定印章真伪并识别印章材质的方法。首先，收集并手工标注了包含印章印文图像的数据集；然后，通过标注后的数据集对YOLOv8进行预训练，并根据测试结果存储性能最优的模型；最后，对印章真伪鉴定和材质分类的结果进行展示。在鉴定过程中，主要根据印文的墨迹、边缘等印迹特征判断印章的真伪；而材质分类方面则关注印章的制印方式，结合相应印迹特征来确定印章的材质组成。实验结果显示，YOLOv8模型在印章真伪鉴定与材质分类任务中的准确率分别达到96.4%和95.3%,具有在真实案件中广泛应用的潜力。

Seal identification plays a pivotal role in document examination, aiming to identify the authenticity of documents through the identification of the seal or seal text for consistency, thereby offering robust support for case analysis and the determination of the scope of investigation. Currently, the identification of seal authenticity and material classification has relied heavily on manual analysis of imprinting characteristics, resulting in high costs, long inspection cycles, and accuracy susceptible to subjective factors. To address these challenges, a lightweight method utilizing the YOLOv8 model was proposed for rapid identification of seal authenticity and materials. Initially, datasets consisting of seal images were collected and manually labeled. Subsequently, YOLOv8 was pre-trained on the labeled datasets, and the model with the best performance was selected according to the test results. Finally, the results of seal authenticity identification and material classification were presented. Specifically, seal authenticity was mainly evaluated by analyzing imprinting characteristics such as ink and edge features. In terms of material classification, the focus was on the printing method of the seal, and the material composition of the seal was determined based on the corresponding imprint characteristics. Experimental results demonstrate that the YOLOv8 model achieves the accuracy rates of 96.4% and 95.3% in seal authenticity identification and material classification tasks respectively, highlighting its broad application potential in real-world scenarios.