基于北斗导航的地质勘探小车  

Geological Exploration Vehicle Based on Beidou Navigation

作　　者：张笑颜 杨安琪 凡中华 孙迎澳 吴俊[1] ZHANG Xiaoyan;YANG Anqi;FAN Zhonghua;SUN Yingao;WU Jun(School of Electronic Science and Engineering,Southeast University,Nanjing 210096,China)

机构地区：[1]东南大学电子科学与工程学院,江苏南京210096

出　　处：《电子器件》2020年第6期1411-1416,共6页Chinese Journal of Electron Devices

摘　　要：岩石岩性的识别与分类对于地质勘探至关重要,目前岩性识别多基于人工判别方法,需要一定的专业背景和丰富的判别经验,受限于环境、天气和人力的综合作用。采用机器视觉技术训练识别模型进行分类是一条新的途径,可以大大提高效率和自动化程度。利用树莓派和intel Movidius算力棒平台,设计出了一款基于北斗导航系统和YOLOv3-tiny神经网络岩石识别模型的地质勘探小车,该小车由定位系统、图像采集系统和识别分类系统三部分组成,可以在野外自由行进并拍摄岩石图像,并将识别结果实时传送终端。该智能小车在测试图集的识别正确率高于80%,视频流测试结果实现了高于80%的敏感性和大于89.5%的特异性,准确度达到88.3%。The identification and classification of rock lithology is essential for geological analysis.At present,lithology identification is mostly based on manual methods,which requires professional background,discrimination experience,and is limited by the combined effects.Using machine vision technology to train model for classification is a new approach,which can greatly improve the efficiency and automation.A geological exploration cart has been designed based on Beidou navigation system and yolov3 tiny neural network model by exploiting raspberry pie and Neural Compute Stick.The cart consists of three parts:positioning system,image acquisition system and classification system.It can travel freely and take images,and the identification results are transmitted to the terminal in real time.The recognition accuracy in the test atlas is higher than 80%.The sensitivity and specificity of the video stream test results are more than 80%and 89.5%respectively,and the accuracy is 88.3%.

关 键 词：北斗导航系统 岩石识别 机器视觉 神经网络 

分 类 号：TH763[机械工程—仪器科学与技术]