Many traditional anchor-based methods for aerial remote sensing image target detec-tion suffer from an imbalance problem between positive and negative anchor frames, which reduces detection precision. To address this issue, a new algorithm called BSAV-RepVGG is proposed, which combines Box Slide-Aware Vectors (BSAV) with RepVGG. Firstly, an overall framework consisting of four mapping outputs is constructed using BSAV. Then, the RepVGG network is improved and used as the backbone network to enhance the detection performance of small and dense objects. Finally, the included angle between adjacent vectors is increased in the loss function of the rotating frame parameter, which effectively improves the precision of the boundary vector regression. The proposed BSAV-RepVGG achieves significant mAP (7) and mAP (12) values of 90.51% and 97.44% on the HRSC2016 dataset, respectively, representing at least a 0.31% and 1.43% improvement compared to the state-of-the-art models. Moreover, on the DOTA dataset, the proposed method attains an mAP value of 76.64%, which is at least 1.28% higher than other state-of-the-art models.