 HBYOLO is a novel framework that improves upon the original YOLO algorithm by replacing its universal convolution layer with an improved Hornet architecture, replacing all Elon layers with the botinette attention mechanism, adjusting the number of anchors, and integrating image segmentation to improve detection accuracy. Experimental results show that this new framework outperforms the original YOLO algorithm on the satellite video dataset, achieving AF1 score of 0.58 and mean average precision of 0.53. Furthermore, the object tracking performance was also improved by incorporating the image segmentation method. This article was authored by Chouranyu, Zhijunfeng, Zinyanwu, and others.