Abstract: Addressing the issues of poor adaptabilities to complex geometric structures, low computational efficiencies, and insufficient accuracies in through-hole feature recognition of injection molded parts based on attributed adjacency graph (AAG) and neural network, a hybrid recognition method combining Open CASCADE (OCC) topological analysis and image recognition is proposed. In the proposed method, firstly OCC geometric topological information is utilized, and candidate feature regions are filtered through face-to-face and edge-to-face feature matching. Subsequently, these regions are rendered into two-dimensional grayscale images, and the shapes and positions of holes are precisely extracted using image recognition algorithms such as edge detection and contour analysis from OpenCV. Finally, combining a spatial mapping algorithm, the two-dimensional recognition results are mapped to the three-dimensional model to obtain feature parameters. Compared with AAG and OCC topological analysis recognition algorithms, the average recognition times of the proposed algorithm are shortened by 67. 21 % and 8. 52 % respectively, and the average accuracies are improved by 39. 39 % and 12. 88 % respectively, which verifies the effectiveness of the proposed algorithm. The proposed algorithm provides a new solution for the automated inspection of injection molded parts.