Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Unmanned aerial vehicle inspection method and system based on unattended operation
The invention relates to an unmanned aerial vehicle inspection method and system based on unattended operation, and the method comprises the steps: constructing a data set of road diseases shot by an unmanned aerial vehicle camera, and carrying out the marking of the diseases; a road disease detection model RPMF-Net model based on regional dynamic perception and multi-feature information fusion is constructed, improvement is performed on the basis of single-stage detection YOLOX-S, and the RPMF-Net model comprises a backbone network Backbone, an LEAM module and a DCFM fusion module; the output of the SPP module in the backbone network and the outputs of the last two RDP-CSPLayerX modules are connected with an LEAM module, the output of the LEAM module is connected with a DCFM fusion module, the output of the DCFM fusion module passes through three parallel Decoupled Head modules, the position, category and confidence information of a predicted target in a target detection task is output respectively, and a final detection result is obtained. The inspection method for road disease recognition is more convenient and more accurate in recognition.
本发明为一种基于无人值守的无人机巡检方法及系统，所述巡检方法包括：构建无人机相机拍摄的道路病害的数据集，并进行病害标注；构建基于区域动态感知和多特征信息融合的道路病害检测模型RPMF‑Net模型，以单阶段检测YOLOX‑S为基础进行改进，所述RPMF‑Net模型包括主干网络Backbone、LEAM模块和DCFM融合模块；主干网络中SPP模块的输出和最后两个RDP‑CSPLayer_X模块的输出连接LEAM模块，LEAM模块的输出连接DCFM融合模块，DCFM融合模块的输出经过三个并行的Decoupled Head模块，分别输出目标检测任务中的预测的目标的位置、类别和置信度信息，获得最终检测结果。为一种更加便捷、识别更加精准的道路病害识别的巡检方法。
Unmanned aerial vehicle inspection method and system based on unattended operation
The invention relates to an unmanned aerial vehicle inspection method and system based on unattended operation, and the method comprises the steps: constructing a data set of road diseases shot by an unmanned aerial vehicle camera, and carrying out the marking of the diseases; a road disease detection model RPMF-Net model based on regional dynamic perception and multi-feature information fusion is constructed, improvement is performed on the basis of single-stage detection YOLOX-S, and the RPMF-Net model comprises a backbone network Backbone, an LEAM module and a DCFM fusion module; the output of the SPP module in the backbone network and the outputs of the last two RDP-CSPLayerX modules are connected with an LEAM module, the output of the LEAM module is connected with a DCFM fusion module, the output of the DCFM fusion module passes through three parallel Decoupled Head modules, the position, category and confidence information of a predicted target in a target detection task is output respectively, and a final detection result is obtained. The inspection method for road disease recognition is more convenient and more accurate in recognition.
本发明为一种基于无人值守的无人机巡检方法及系统，所述巡检方法包括：构建无人机相机拍摄的道路病害的数据集，并进行病害标注；构建基于区域动态感知和多特征信息融合的道路病害检测模型RPMF‑Net模型，以单阶段检测YOLOX‑S为基础进行改进，所述RPMF‑Net模型包括主干网络Backbone、LEAM模块和DCFM融合模块；主干网络中SPP模块的输出和最后两个RDP‑CSPLayer_X模块的输出连接LEAM模块，LEAM模块的输出连接DCFM融合模块，DCFM融合模块的输出经过三个并行的Decoupled Head模块，分别输出目标检测任务中的预测的目标的位置、类别和置信度信息，获得最终检测结果。为一种更加便捷、识别更加精准的道路病害识别的巡检方法。
Unmanned aerial vehicle inspection method and system based on unattended operation
一种基于无人值守的无人机巡检方法及系统
ZHANG JUNFEI (Autor:in) / CHENG HUISHENG (Autor:in) / YAN HAOHUI (Autor:in) / CHEN JUNLIN (Autor:in) / ZHANG LEI (Autor:in) / WANG LING (Autor:in)
23.07.2024
Patent
Elektronische Ressource
Chinesisch
| Europäisches Patentamt | 2022
| Europäisches Patentamt | 2022
| Europäisches Patentamt | 2024
| Europäisches Patentamt | 2024
| Europäisches Patentamt | 2021