Abstract: Aiming at the problems of low detection precision and missed detection due to poor resolution and fuzzy features of infrared images, a Vehicle infrared image detection algorithm based on SCW-YOLO (Shared-Context-Wise-YOLO) is proposed. Firstly, a Shared Convolution Pyramid Pooling (SCPP) structure is proposed. By using shared convolution layers and convolution with different expansion rates, the multi-scale feature extraction capability of the model is improved. Secondly, a Context Information Fuision (CIF) module based on adaptive fine grained channel attention is designed to enhance the weight of key information and strengthen the feature fusion ability by using adaptive adjustment strategies. Finally, the WIoU (Wise-IoU) loss function is used to adjust the gradient gain distribution to reduce the effect of possible mislabeling in the image. Experiments on FLIR datasets show that SCW-YOLO is 1.5, 1.7 and 1.8 percentage points higher in precision, recall and mean Average Precision than the YOLO11s benchmark algorithm, respectively, and the detection speed reaches 85.5 FPS (Frames Per Second).