Engineers try to recognize the features of automobile sheet metal manually by human today. This thesis is to establish hybrid feature recognition algorithms. The system accesses the CAD model based on STEP PART 42 format to recognize features. Differing from the previous feature recognition systems which recognize the general feature in sheet metal, this thesis concentrates on the requirement of CAPP. This thesis divides automobile sheet metal into many groups, and recognizes the features into groups. The process of recognition includes building the relationship among features and collecting the necessary geometric data. The system presents the results by the tree data structure and output the text file based on feature protocol. Besides the general features, this research aims at grouping the main flat and the sub flat. The goal of this research is collecting the feature information from complex surfaces.