In the process of detecting male and female ceramic liners, the randomness in the placement of some ceramic liners, significant edge interference, and varying lighting conditions pose challenges. Therefore, traditional detection methods struggle to efficiently and accurately identify the heads of ceramic liners. Accurate and effective detection of male and female liners is crucial for subsequent encapsulation tasks. An improved optimized multi-posture ceramic liners male and female detection algorithm YOLOv7_OBB-CL (YOLOv7_Oriented Bounding Box-Ceramic liners) based on YOLOv7 algorithm is proposed. Firstly, a rotating frame replaces the conventional horizontal frame to meet the multi-posture task requirements; Then, the network structure is improved, with optimized fusion modules RepGFPN and BIC used in the feature fusion process; Subsequently, the incorporation of the SimAM attention mechanism reduces noise interference while giving the network a stronger focus on critical information; Finally, the utilization of the optimized VFLoss function instead of the traditional loss calculation method significantly enhances the stability of model training. In the experiments, images of ceramic liners from different angles and lighting conditions were used for testing, and the results were evaluated and analyzed. The results show that the mAP of YOLOv7_OBB-CL on the homemade dataset reaches 96.7%, representing a notable enhancement compared to the mainstream algorithm.