A fidelity evaluation model for watermarked vector maps based on improved fuzzy comprehensive evaluation and random forest

Digital watermarking is crucial for securing vector maps, requiring imperceptible embedding to preserve map integrity and data quality. Traditional assessments of watermarked vector map fidelity, relying on visual inspection and error analysis, often overlook the intrinsic map characteristics, which results in inaccurate evaluations. To address the issue, this study uses topological consistency, geometric feature similarity, and coordinate error as the first-level indicators, and ten indicators such as element closure, topological error rate, angle change rate, maximum error, graphic complexity and so on as the second-level indicators to describe the variation in vector maps after watermark embedding. Considering the uncertainty and importance of indicators, a fidelity assessment model for watermarked vector maps is constructed based on fuzzy comprehensive evaluation. Furthermore, the model optimizes the indicator weights using the Random Forest algorithm, thereby mitigating subjectivity and improving the accuracy of fidelity assessments. This approach provides a scientifically robust evaluation framework aligned with real-world data, offering a more rational alternative to traditional error-based evaluations. Additionally, it provides guidance for determining the watermark embedding strength, facilitating the balance between robustness and invisibility, thus ensuring quality standards and optimal algorithm performance.