摘要
边防巡逻对于国家安全和经济稳定发展至关重要,边境地区巡逻员老龄化及巡逻效率低下的问题经常导致入侵活动未被及时发现。特别是在复杂地形条件下目标运动遮挡、目标距离远以及昼夜视野受限等挑战。采用热成像图像检测结合改进的YOLOV5框架来增加多头注意力机制,从而改善行人车辆的特征区分能力和稳健性。通过在边境场景中的实验验证,该机器人检测技术展现了显著的实用性和有效性。
Border patrol is very important to the economic and national security of the country,and it is also very necessary to carry it out.However,issues such as the aging of patrol personnel and low patrol efficiency often result in delayed detection of intrusion activities.Challenges include target movement occlusion,long distances,and limited visibility during day and night in complex terrains.Using thermal imaging cameras,this paper suggests using the vehicle and pedestrian detection method.By improving the YOLOv5 algorithm,add multi-head attention mechanism,the model′s feature differentiation ability and robustness of pedestrian vehicles can be improved.Experimental verification in border scenarios demonstrates the significant practicality and effectiveness of this robot detection technology.
作者
张鑫
ZHANG Xin(Shanghai Yuanqing Information Technology Co.,LTD.Shanghai 201702)
出处
《长江信息通信》
2025年第3期50-52,共3页
Changjiang Information & Communications
