With the continuous advancement of 3D modeling technology, 3D models have become increasingly complex and large in volume to achieve finer and more realistic representations, presenting significant challenges for transmission and visualization. To address this, 3D model lightweight methods have emerged, with mesh simplification algorithms being a crucial strategy that focuses on reducing the number of triangular facets on model surfaces. Among various mesh simplification algorithms, the Quadric Error Metric (QEM) algorithm stands out for its high simplification efficiency and superior results. However, models simplified by the original QEM algorithm may suffer from the loss or even distortion of certain features. To resolve these issues, this paper proposes an optimized QEM algorithm that incorporates Gaussian curvature and relative curvature error to enhance edge collapse cost calculation. This approach maintains model details while effectively reducing triangle counts. Experimental validation demonstrates the algorithm's performance compared with traditional QEM methods. The optimized algorithm achieves approximately 11.2% reduction in Hausdorff distance. Although its mean square error and average error are slightly higher than those of the original QEM algorithm, it shows enhanced visual quality and better preservation of original model details and characteristic features.