<p>An improved YOLOv11 circuit component detection algorithm (CSTM-YOLOv11) is proposed to address the lack of publicly available circuit diagram datasets and the problems of insufficient accuracy, easy occurrence of false positives and false negatives in existing circuit component detection algorithms. Firstly, combined with ECA attention mechanism, a Convolutional Channel Collaboration Module (ConCA) is designed to efficiently capture inter channel dependencies and improve the sensitivity of the model to different target features; Secondly, reconstruct the feature interaction module (C3k2-SE) to adaptively focus on key information and enhance the expressive ability of target features while maintaining lightweight. Integrate Transformer Blocks into the C2PSA component in the backbone network to enhance the model’s ability to handle contextual relationships and reduce the interference of confusing components; Finally, a new detection head (MBCHead) based on MobileNetV2 BottleNeck structure is proposed to enhance the classification and localization capabilities of the model. At the same time, an Inner-Focal IOU is designed to replace the original CIOU loss function and optimize the regression of bounding boxes to improve the detection rate of the model. Compared to the YOLOv11 model, the improved CSTM-YOLOv11 model has improved accuracy, recall mAP50 and the real-time detection speed FPS has increased by 2.8%, 3.4%, 5.1%, and 24.67% respectively. The generalization test on multiple public datasets shows that the algorithm can effectively improve the detection performance of circuit components in complex scenes, and has good robustness and universality.</p>