Aluminum droplet, oxide cap and flame segmentation in burning Al/AP propellants by combining YOLOv7 and two-stage cluster

Aluminum additives have complex effects on propellant combustion, and high-speed microscopic imaging is a valuable tool to investigate these effects. However, challenges arise from issues like out-of-focus images and grayscale variations, hindering structural information extraction. This study introduces a segmentation method to segment the oxide cap, aluminum droplet, and enveloping flame, combining YOLOv7 detection and two-stage cluster segmentation, integrating geometrical data into the primary cluster. The method is rigorously evaluated with metrics, yielding impressive results: 84.4% Mean Intersection over Union (MIoU), 91.1% Precision (Pr), 92.4% Recall (Re), and 89.3% F1 score. These metrics affirm its effectiveness. Accurate segmentation facilitates the extraction of essential information, including position, shape, and motion data. This information is vital for understanding combustion mechanisms, such as reaction nonuniformity, combustion rate, and motion impetus and the further enlightenment of the investigation of propellents.