摘要
文章提出一种结合语义分割与目标检测的人员识别方法,利用Mask2Former实现洪水区域高精度分割,YOLOv8识别图像中人员目标,通过空间重叠策略判断其是否处于水中。为验证方法有效性,构建复合数据集并开展多阈值对比实验。结果表明,该融合方法在复杂洪水场景下具备较高准确性与实用性,明显减少了误检与漏检,展现出良好适应性与应用价值。
The article proposes a personnel identification method that combines semantic segmentation and object detection,using Mask2Former to achieve high-precision segmentation of flooded areas and YOLOv8 to detect personnel targets in images,determining whether they are in water through a spatial overlap strategy.To verify the effectiveness of the method,a composite dataset was constructed,and multi-threshold comparative experiments were conducted.The results show that this integrated method has high accuracy and practicality in complex flood scenarios,significantly reducing false positives and false negatives,and demonstrating good adaptability and application value.
作者
侯愷
杨亮珠
HOU Kai;YANG Liangzhu
出处
《智能城市》
2026年第2期141-144,共4页
Intelligent City
