<p>This study focuses on the soundscape and spatial structure of urban parks. Four representative parks in Fuzhou City were selected as case studies. Data was collected through a combination of questionnaire surveys, sound level measurements, and on-site recordings. Deep learning code was used to implement audio automatic classification, image semantic segmentation, and depth estimation processes. Based on this, a multi-dimensional spatial analysis indicator system was constructed, and a subjective and objective soundscape evaluation system was established simultaneously. Using methods such as correlation analysis, stepwise regression, cluster analysis, and geographically weighted regression, the study progressively revealed significant differences in the composition of soundscapes and their influencing factors between park boundaries and internal spaces. The results indicate that the soundscapes of boundary spaces are more significantly influenced by urban traffic, facilities, and layout, while internal spaces are more dependent on vegetation structure and visual perception elements. Soundscape characteristics exhibit obvious scale heterogeneity and structural dependency across different spatial types. This provides theoretical basis and data support for the management and optimization of acoustic environments in urban landscape spaces.</p>