Image cropping aims to create visually appealing pictures aligned with user preferences. As social media develops, user demand for visual content in images has become more diverse. Previous cropping methods struggle to capture the semantics of images, failing to highlight representative content and adapt to the varying preferences of users for non-representative content. To address these issues, we propose a novel visual-semantic-aware cropping method that uses a visual semantic aggregation approach to identify key visual patches in the image, then integrates these patch features and basic image features through partition enhancement and graph-based feature interaction, thereby extracting representative content with high aesthetic value. To consider user preferences for non-representative content, we introduce a user-adjustable semantic coordination proportional mechanism, allowing users to adjust the visual richness of non-representative content in cropped images. Experiments show our method outperforms state-of-the-art methods in achieving aesthetic crops, while providing users with adjustable cropping options.