<p>Jiandu served as the primary medium for recording important information in ancient China, and analyzing and interpreting its textual content contributes to a deeper understanding of history and cultural heritage preservation. Jiandu text detection is a prerequisite for subsequent recognition and semantic analysis. To address surface aging, irregular character scales, dense arrangements, and intersecting character adhesion in Jiandu images, we propose a Jiandu text detection method based on character-level region mask fusion and adaptive segmentation to preserve character structures and segment adhered characters. In automatic segmentation, pixel feature analysis and overlap suppression are introduced to reduce background noise and invalid regions. In character region processing, mask clustering, region fusion, and adaptive segmentation are employed to alleviate inter-character occlusion and adhesion. Experimental results demonstrate improvements in character structure preservation, adhered character segmentation, and overall Jiandu text detection performance, supporting non-training-based character-level annotation.</p>