Measuring Nozzle Image Area Using Deep Learning and Image Segmentation

Nozzle characteristics of aircraft engines are one of the important indicators for measuring their performance. It is necessary to conduct measurement and analysis in order to obtain the performance parameters of the nozzle. This paper proposes a measurement method for nozzle images. This method is based on YOLO V11 object detection and GrabCut image segmentation techniques, and mainly includes the following points: To enhance the YOLO V11 model, we utilize the re-parameterization module, RepConv, to improve its C3k2 module, and thus propose the RepC3k2 module. This allows the model to have better feature representation capabilities during the training process. A loss function based on MPDIoU with better convergence properties is introduced to optimize the model’s training process. A dual attention mechanism, CBAM, is added to the backbone network of the model to improve its accuracy and robustness during detection. A novel method for measuring the area of a nozzle image is proposed: utilizing GrabCut image segmentation and image processing techniques to segment the nozzle image, and then obtaining the area based on the proportion of image pixels. Finally, experiments were conducted to verify the reliability and accuracy of the proposed scheme.