基于机载计算机的无人机智能巡检方案
2022,30(7):181-186
摘要：随着无人机飞行制造技术的不断发展，传统人力巡检在一些领域已逐渐被无人机巡检所取代，相关领域已涉及到森林虫害、电力巡检、工地安全、渔政执法等。针对目前无人机巡检中对巡检目标的实时检测、结果的实时传输以及巡检飞行的智能操控需求，设计了一种基于机载计算机的无人机智能巡检方案。通过机载计算机取代传统方案的移动图形工作站，并结合轻量化网络对检测模型进行轻量化改进，实现无人机检测端直接进行边缘计算。然后通过优化机载端与APP的传输方案以及设计云端远程巡检控制方案，实现无人机的机载端实时自动回传检测结果，并上传至云平台。实验表明，该方案无需移动端复杂操控，可实现远程下发巡检任务指令并实时接收检测结果，达到远端人员可快速精准响应的目的。
关键词：无人机;云平台;智能巡检;目标检测;轻量化
An Intelligent Inspection Scheme of UAV Based on Airborne Computer
Abstract：With the continuous development of UAV flight manufacturing technology, the traditional manual inspection has been gradually replaced by UAV inspection in some fields, and related fields have been involved in forest insect pests, electricity inspection, site safety, fishery law enforcement, etc. Aiming at the real-time detection of inspection targets, real-time transmission of results and intelligent control of inspection flight in UAV inspection, an intelligent inspection scheme of UAV based on airborne computer was designed. By replacing the traditional mobile graphics workstation with airborne computer, and combining with lightweight network, the detection model is improved to realize edge calculation directly at the detection end of UAV. Then, by optimizing the transmission scheme between the airborne end and APP and designing the cloud remote inspection control scheme, the airborne end of the UAV can automatically send back the detection results in real time and upload them to the cloud platform. Experimental results show that the scheme does not require complex operation on the mobile terminal, and can deliver inspection task instructions remotely and receive detection results in real time, so that the remote personnel can respond quickly and accurately.
Key words：UAV; Cloud platform; Intelligent inspection; Object detection; Light-weighting
收稿日期：2021-12-28
基金项目：上海市科委科技创新行动计划（项目编号：21511102605）
