Stiffened thin-walled structures are widely used in aerospace and related engineering applications due to their high load-carrying capacity. In practical design, a sequential optimization strategy is commonly adopted, in which the stiffener layout is first determined through topology or layout optimization, followed by the optimization of stiffener cross-sectional shape and dimensions. Since stiffeners can be regarded as beam-like components embedded in thin-walled structures, their cross-sectional configuration plays a critical role in governing the global stiffness distribution and local stability. Existing studies on stiffener cross-sectional optimization mainly focus on shape and size optimization of beam sections [1], while the evolution of cross-sectional topological configurations has received limited attention. Although traditional topology optimization methods have been introduced into beam cross-section design to explore novel configurations [2], the resulting designs often lack clear geometric features and require complex post-processing. Alternatively, section-type selection approaches based on predefined standard profiles can produce geometrically clear and directly manufacturable designs [3], but they are usually computationally expensive and rely heavily on heuristic optimization algorithms. To address these limitations, this work proposes a continuous variable cross-section optimization framework for stiffeners in thin-walled structures. Specifically, a transition-based modeling strategy is introduced to enable smooth and continuous variation among different candidate cross-sectional profiles, including T-shaped, L-shaped, and rectangular sections. Moreover, the original discrete 0/1 section selection problem is reformulated as a continuous optimization problem without intermediate values, ensuring physically meaningful and manufacturable optimal designs. In addition, Isogeometric analysis (IGA) is employed for structural analysis to accurately represent the geometry of both the skin and the stiffeners, while Feature-Driven Optimization (FDO) method is employed to drive the optimization of stiffener cross-sectional configurations. Numerical examples demonstrate that the proposed method can effectively improve the structural stiffness under prescribed volume constraints, confirming its potential for efficient and practical stiffener cross-sectional optimization in thin-walled structures.