文化遗产场景下花山岩画智能检测与可视化分析系统研究与实现

Research and Implementation of an Intelligent Detection and Visual Analysis System for Huashan Rock Art in Cultural Heritage Contexts

摘要: 针对花山岩画姿态识别依赖人工、效率偏低且易受主观判断影响的问题，本文提出了一种基于YOLOv8目标检测的可视化分析系统。系统采用CustomTkinter构建交互界面，集成模型推理、结果渲染、批量处理、数据统计与场景识别功能。通过引入自适应标签布局策略，根据目标框位置自动调整标签显示区域，有效避免密集人像区域的标签重叠遮挡；提供置信度阈值实时调节功能，用户可根据图像质量与检测需求动态调整灵敏度，实现检出数量与准确率的灵活平衡；基于考古学研究成果构建场景识别规则体系，系统支持普通正身、首领正身、巫师正身、普通侧身、持鼓侧身、群聚正身六类姿态检测，可根据检测结果自动判定祭祀、集会、狩猎等场景类型，辅助参观人员快速理解岩画文化内涵。系统在常规硬件环境下能够实现快速检测，界面交互流畅，批量处理流程稳定。本系统将高精度检测模型转化为考古人员可直接操作的桌面工具，为花山岩画数字化保护提供了可用的技术实现方案。

Abstract: Addressing the issues of manual reliance, low efficiency, and susceptibility to subjective judgment in the pose recognition of Huashan Rock Art, this paper proposes a visual analysis system based on YOLOv8 object detection. The system utilizes CustomTkinter to build an interactive interface and integrates functions such as model inference, result rendering, batch processing, data statistics, and scene recognition. By introducing an adaptive label layout strategy, the system automatically adjusts label display areas based on bounding box positions, effectively preventing label overlap and occlusion in areas with dense figures. It offers real-time adjustment of confidence thresholds, allowing users to dynamically tune sensitivity according to image quality and detection requirements, thereby achieving a flexible balance between the number of detections and accuracy. A scene recognition rule system, constructed based on archaeological research findings, enables the detection of six pose categories—ordinary frontal, leader frontal, shaman frontal, ordinary profile, profile holding a drum, and clustered frontal—and automatically classifies scenes such as rituals, gatherings, and hunting, helping visitors quickly grasp the cultural significance of the rock art. The system achieves rapid detection on standard hardware, features a smooth interactive interface, and ensures stable batch processing workflows. By transforming a high-precision detection model into a desktop tool directly operable by archaeologists, this system provides a practical technical solution for the digital preservation of Huashan Rock Art.