基于手势的无人机飞行控制实验教学设计  

Teaching Design of Flight Control via Hand Gestures for UAVs

作　　者：魏晓敏 李艳阳 孙聪 张海宾 WEI Xiaomin;LI Yanyang;SUN Cong;ZHANG Haibin(School of Cyber Engineering,Xidian University,Xi'an 710071,China)

机构地区：[1]西安电子科技大学网络与信息安全学院,陕西西安710071

出　　处：《软件导刊》2024年第8期261-266,共6页Software Guide

基　　金：中央高校基本科研业务费专项资金项目(ZYTS23165,ZYTS24141);国家自然科学基金项目(62232013,62272366);陕西省重点研发计划项目(2023-YBGY-371)。

摘　　要：为解决新手在操纵无人机飞行的初始练习阶段容易造成丢失、坠落、碰撞等问题,利用AI技术设计一种基于手势的无人机飞行智能控制实验。首先通过机载或地面摄像头捕捉操纵者手势图像;其次使用加载于地面控制系统或机载设备的动态手势识别模型进一步处理图像,得到相关飞行控制指令;再次将飞行控制指令发送到无人机控制系统以达到操纵无人机飞行的目的。为了验证基于YOLOv5算法的动态手势识别模型性能,构建包含5个场景的2500余张图片的数据集进行验证。实验表明该方法不仅对手势识别具有较高的精确度,而且对新场景下的图像也具有更好的智能检测能力。实验设计涉及学科较为综合、功能扩展性好、实用性强,是控制系统教学、AI算法实践与嵌入式系统相互融合的创新实验项目。To solve the problems of loss,falling,collision and other issues that beginners may encounter during the initial practice stage of manipulating drone flight,a gesture based intelligent control experiment for drone flight is designed using AI technology.Firstly,capture the operator′s gesture image through onboard or ground cameras;Then,the dynamic gesture recognition model loaded into the ground control system or onboard equipment is used to further process the image and obtain relevant flight control instructions;Next,the flight control instructions will be sent to the drone control system to achieve the purpose of manipulating the drone′s flight.In order to verify the performance of the dynamic gesture recognition model based on YOLOv5 algorithm,a dataset of more than 2500 images containing 5 scenarios was constructed for validation.The experiment showed that this method not only has high accuracy for gesture recognition,but also has better intelligent detection ability for images in new scenarios.The experimental design involves a comprehensive discipline,good functional scalability,and strong practicality.It is an innovative experimental project that integrates control system teaching,AI algorithm practice,and embedded systems.

关 键 词：实验教学设计 飞行控制 手势识别 无人机 嵌入式系统 

分 类 号：G642[文化科学—高等教育学]