摘 要:路径规划技术具有巨大的科研价值和广阔的应用前景,其在高科技领域的应用包括无人机的避障飞行、无人机飞行过程中的最优路径选择。计算机视觉技术可以对图片或视频进行处理,获取对应场景的多维信息,经过进一步的处理得到更加有效的信息。文章探析现有的基于深度神经网络的计算机视觉算法的原理、优缺点,对现有的计算机视觉算法进行优化,并将计算机视觉问题的解决方法应用到强化学习控制算法中,提高无人机路径规划的精度和效率。
关键词:深度神经网络;Faster R-CNN;YOLOv5 目标检测
DOI:10.19850/j.cnki.2096-4706.2022.013.028
基金项目:咸阳职业技术学院 2020 年度大学生科技创新研究项目(2020XS03)
中图分类号:TP18 文献标识码:A 文章编号:2096-4706(2022)13-0113-04
Application Research of Computer Vision Technology Based on Deep Neural Network in Path Planning
HU Wenjie, LEI Bingqiang
(Xianyang Vocational Technical College, Xianyang 712000, China)
Abstract: Path planning technology has great scientific research value and broad application prospects. Its applications in hightech fields include obstacle avoidance flight of UAV and optimal path selection in the flight process of UAV. Computer vision technology can process pictures or videos, obtain the multi-dimensional information of the corresponding scene, and get more effective information after further processing. This paper analyzes the principle, advantages and disadvantages of the existing computer vision algorithm based on deep neural network, optimizes the existing computer vision algorithm, and applies the solution of computer vision problems to the reinforcement learning control algorithm to improve the accuracy and efficiency of UAV path planning.
Keywords: deep neural network; Faster R-CNN; YOLOv5 object detection
参考文献:
[1] 江磊,崔艳荣 .基于 YOLOv5 的小目标检测 [J]. 电脑知识与技术,2021,17(26):131-133.
[2] 杨晓玲,蔡雅雯 .基于 yolov5s 的行人检测系统及实现 [J].电脑与信息技术,2022,30(1):28-30.
[3] 马琳琳,马建新,韩佳芳,等 .基于 YOLOv5s 目标检测算法的研究 [J]. 电脑知识与技术,2021,17(23):100-103.
[4] MACHADO M C,BELLEMARE M G,TALVITIE E,et al. Revisiting the arcade learning environment: Evaluation protocols and open problems for general agents [J].Journal of Artificial Intelligence Research,2018,61:523-562.
[5] LI Y. Deep reinforcement learning: An overview [J/OL].arXiv:1806.08894v1 [cs.LG].[2022-05-26].https://arxiv.org/pdf/1806.08894.pdf.
作者简介:胡文杰(1987—),女,汉族,陕西咸阳人,助教,硕士,研究方向:人工智能、计算机视觉、计算机软件专业的教学、科研工作。