Scholarlink

Background <p>Cultural heritage requires intelligent frameworks for value assessment and activation to ensure preservation, sustainable utilization, and risk management. Integrating deep learning with Internet of Things (IoT) technologies enables comprehensive monitoring of structural, environmental, and contextual characteristics of heritage sites.</p> Problem statement <p>Conventional cultural heritage value assessment methods largely depend on subjective expert judgments and limited or fragmented data sources, resulting in inconsistent evaluations, poor scalability, and reduced reliability in decision-making processes.</p> Objective <p>To overcome these limitations, this research aims to develop an advanced computational framework that enables accurate, robust, and scalable cultural heritage value assessment by integrating deep learning models with evolutionary optimization techniques.</p> Proposed methodology <p>A Pooling-enriched Transformer Graph Neural Network with Scalable Reptile Search Algorithm (PTGNN-SRA) was proposed. IoT sensors collected environmental and structural parameters, supplemented with archival records. Data preprocessing included handling missing values via interpolation and imputation, and Z-score normalization for numerical sensor data. PTGNN captured structural and contextual relationships in multimodal data, combining global attention with local pooling for precise representation learning. The SRA optimized PTGNN parameters efficiently, ensuring fast convergence and scalability.</p> Results <p>The Python-based model achieved an overall efficiency of the PTGNN-SRA achieved 97.9% accuracy, 96.5% precision, 97.3% recall, and 96.8% F1-score, with 1.92% maximum relative error, 0.6% average relative error, and 0.3257&#xa0;mm average absolute error.</p> Conclusion <p>It provides efficiency of using Transformer-based GNNs with evolutionary optimization to evaluate heritages. Combination of multimodal IoT information, effective preprocessing, and scalable algorithms assist in better conservation, sustainable activation and informed decision-making.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Algorithms for cultural heritage value assessment and activation utilization based on deep learning and the Internet of Things

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月23日

查看全文
  • Haijing Fan

摘要

Background

Cultural heritage requires intelligent frameworks for value assessment and activation to ensure preservation, sustainable utilization, and risk management. Integrating deep learning with Internet of Things (IoT) technologies enables comprehensive monitoring of structural, environmental, and contextual characteristics of heritage sites.

Problem statement

Conventional cultural heritage value assessment methods largely depend on subjective expert judgments and limited or fragmented data sources, resulting in inconsistent evaluations, poor scalability, and reduced reliability in decision-making processes.

Objective

To overcome these limitations, this research aims to develop an advanced computational framework that enables accurate, robust, and scalable cultural heritage value assessment by integrating deep learning models with evolutionary optimization techniques.

Proposed methodology

A Pooling-enriched Transformer Graph Neural Network with Scalable Reptile Search Algorithm (PTGNN-SRA) was proposed. IoT sensors collected environmental and structural parameters, supplemented with archival records. Data preprocessing included handling missing values via interpolation and imputation, and Z-score normalization for numerical sensor data. PTGNN captured structural and contextual relationships in multimodal data, combining global attention with local pooling for precise representation learning. The SRA optimized PTGNN parameters efficiently, ensuring fast convergence and scalability.

Results

The Python-based model achieved an overall efficiency of the PTGNN-SRA achieved 97.9% accuracy, 96.5% precision, 97.3% recall, and 96.8% F1-score, with 1.92% maximum relative error, 0.6% average relative error, and 0.3257 mm average absolute error.

Conclusion

It provides efficiency of using Transformer-based GNNs with evolutionary optimization to evaluate heritages. Combination of multimodal IoT information, effective preprocessing, and scalable algorithms assist in better conservation, sustainable activation and informed decision-making.