摘 要:文章以实现地质勘探无人化和岩石分类智能化为目标,设计出用于地质勘探现场的具有岩石分类功能的智能导航小车系统。该系统分为小车模块和岩石分类模块两部分,小车模块利用ROS 平台进行开发,得到勘探区域环境数据后,利用SLAM 算法生成栅格地图并进行路径规划来获得全局最优路径,最后小车自动导航至该地点并获得该区域的岩石图片。岩石分类模块采用深度学习技术对岩石图片进行分类,基于Softmax 分类器和ResNet101 神经网络模型,建立岩石分类模型。
关键词:岩石分类;SLAM 算法;自动导航;ResNet 模型;Softmax
DOI:10.19850/j.cnki.2096-4706.2023.08.022
基金项目:智能化机器人课程体系改革方案研究(201901059032);面向机器人相关课程的智能化探索及实训方法研究(JYX19062)
中图分类号:TP242 文献标识码:A 文章编号:2096-4706(2023)08-0088-05
Application Research of ResNet Model in Intelligent Rock Survey Car
YU Feigen, LIU Ke
(School of Computer Science, South-Central Minzu University, Wuhan 430074, China)
Abstract: Aiming at realizing unmanned geological exploration and intelligent rock classification, this paper designs an intelligent navigation car system with rock classification function for the situation of geological exploration. The system is divided into two parts of the car module and the rock classification module. The car module is developed by using the ROS platform. After obtaining the environmental data of the exploration area, the SLAM algorithm is used to generate a grid map and carry out path planning to obtain the global optimal path. Finally, the car automatically navigates to the location and obtains the rock pictures of the area. The rock classification module uses deep learning technology to classify rock pictures, and establishes a rock classification model based on Softmax classifier and ResNet101 neural network model.
Keywords: rock classification; SLAM algorithm; automatic navigation; ResNet model; Softmax
参考文献:
[1] 张野,李明超,韩帅. 基于岩石图像深度学习的岩性自动识别与分类方法 [J]. 岩石学报,2018,34(2):333-342.
[ 2 ] S U N Y,WA N G X ,TA N G X . D e e p l e a r n i n g f a c e representation by joint identification-verification [J].Advances in Neural Information Processing Systems 27 (NIPS 2014) .Montreal:MIT,2014:1988-1996.
[3] 李阳. 基于ROS 系统移动机器人SLAM 算法研究与实现[D]. 西安:西安工程大学,2018.
[4] RIBEIRO D ,MATEUS A,MIRALDO P ,et al. A realtime deep learning pedestrian detector for robot navigation [C]//
IEEE International Conference on Autonomous Robot Systems and Competitions.Coimbra:IEEE,2017:165-171.
[5] 田宇. 基于机器人操作系统的智能移动小车目标识别研究[J]. 软件工程,2021,24(8):11-15.
[6] 梁双,倪晓昌,董娇玲,等. 基于激光雷达的SLAM 算法综述 [J]. 信息与电脑:理论版,2022,34(3):59-61.
[7] 詹润哲,姜飞. 基于ROS 与深度学习的移动机器人目标识别系统 [J]. 电子测试,2018(15):70-71+64.
[8] 冯震,潘锦涛,陈相贺. 基于深度学习的岩石样本分类研究 [J]. 信息与电脑:理论版,2021,33(9):71-73.
[9] 葛昊. 基于深度学习的图像分类方法研究 [D]. 成都:电子科技大学,2020.
[10] 万磊,佟鑫,盛明伟,等.Softmax 分类器深度学习图像分类方法应用综述 [J]. 导航与控制,2019,18(6):1-9+47.
作者简介:喻飞根(1999—),男,汉族,湖南岳阳人,本科在读,研究方向:机器人与计算机视觉;
通信作者:刘科(1979—),男,汉族,湖北荆州人,讲师,博士,研究方向:智能机器人技术。