| 基于注意力机制的遥感图像车辆目标检测 |
 点此下载全文 |
| 引用本文：肖美波，孟维伟，刘轩宏，张丹.基于注意力机制的遥感图像车辆目标检测[J].计算技术与自动化,2025,(1):29-34 |
| 摘要点击次数: 1126 |
| 全文下载次数: 294 |
|
|
| 中文摘要:为了更好地通过遥感图像监测地面交通流量，应对日趋复杂的路况问题，提出了一种基于YOLOX的全新的一阶段目标检测网络AE-YOLO(Attention-Enhanced YOLO)。采用混合注意力加强的SCACSP特征增强模块，很大程度上抑制了背景信息，突出了目标的显著特征，用解耦头部代替耦合头，大大提高了网络的收敛速度和性能。对损失函数进行改进，使用CIoU Loss代替原来的IoU Loss，从而解决了树木遮挡等造成的车辆漏检问题。试验结果表明，与主流的目标检测网络相比，该方法在精度上有更好的表现。 |
| 中文关键词:深度学习 目标检测 遥感图像 混合注意力机制 YOLOX |
| |
| Vehicle Target Detection in Remote Sensing Image Based on Attention Mechanism |
|
|
| Abstract:In order to better monitor ground traffic flow through remote sensing images and deal with increasingly complex road conditions, a new one-stage target detection network AE-YOLO (Attention-Enhanced YOLO) based on YOLOX is proposed. The SCACSP feature enhancement module with mixed attention enhancement suppresses the background information to a large extent, highlights the salient features of the target. Replacing the coupling head with the decupled head, which greatly improves the convergence speed and performance of the network. Using CIoU Loss instead of the original IoU Loss to improve the loss function, thus solving the problem of vehicle missed detection caused by trees occlusion and other reasons. Experimental results show that the proposed method outperforms the mainstream object detection networks in accuracy. |
| keywords:deep learning target detection remote sensing image hybrid attention mechanism YOLOX |
| 查看全文 查看/发表评论 下载pdf阅读器 |
