Laser deposition manufacturing (LDM) technology has unique advantages in additive manufacturing of large aircraft frames and beams. However, stress and deformation have become bottlenecks that hinder the application of this technology. Therefore, the islands process is widely used to discrete the residual stress and alleviate the deformation of the parts. However, the traditional islands process does not take into account the geometric structural features of parts easily lead to irregular partition lap, which introduced pores, poor fusion and other defects. In order to solve this problem, a feature islands method is proposed. According to the geometric shape characteristics of the slicing layers of typical frame and beam structural parts, The features are classified into four types: “十”-shape, T-shape, L-shape and “一”-shape features, and all kinds of features are limited from three aspects :shape, pose and size to complete the definition of island features. A feature recognition algorithm based on region skeleton line detection is proposed, where the skeletonization is used to effectively simplify the features and retain the part characteristics. The vector cross-product and fixed-ratio point method are used to calculate the relevant parameters such as feature angle, plane attitude angle and number of feature branches. Feature type identification is achieved by comparing calculated values with defined values. The algorithm is verified by slice data of a typical aircraft frame model. The results show that the algorithm can quickly and accurately identify various features to realize automatic feature islands of parts, which lays a foundation for intelligent additive manufacturing technology.