张少杰　艾宇　张冰　孙轩　张铭　张天顺　宋国平

摘要：仓库中醇化的烟叶因含有水分，在微生物作用下会发生霉变。为降低烟叶霉变检测的劳动强度，提高烟叶检测效率，我们基于机器视觉研发了一款搭载FLIR C3红外热成像检测仪的烟仓机器人。本研究主要介绍了烟仓机器人软硬件控制系统的设计，并对机器人行进路径的识别与控制进行研究。烟仓机器人的硬件控制系统采用32位ARM处理器STM32F103ZET6，利用OV2640传感器进行图像的采集与传输;软件控制系统基于Microsoft Visual Studi0 2010开发环境，调用OpenCV等图像处理函数，实现图像分割、边缘检测及路径识别与规划。路径识别采用基于机器视觉的方法，利用高斯滤波去除图像中的噪音、干扰，经阈值化处理后采用Canny算子进行图像边缘检测处理，通过Hough变换，提取出车道的边缘线及车道中心线，根据车道中心线的角度偏差和像素偏差制定控制策略，从而实现烟仓机器人的自控行走，实际运行试验结果表明，该系统能够很好地实现路径识别和运动控制。本研究可为无线通讯和路径识别控制类机器人的相关设计研兖提供参考。

关键词：烟仓机器人;机器视觉;路径识别;STM32处理器;WIFI通讯

Path Recognition and Control of TobaccoWarehouse Robot Based on Machine VisionZhang Shaojiel ， Ai Yul ， Zhang Bingl ， Sun Xuanl ， Zhang Mingl ， Zhang Tianshunl ， Song Guoping2

Abstract The alcoholized tobacco leaves in the warehouse could mildew due to the presence of water.In order to reduce the labor intensity of tobacco leaf mold detection and improve the detection efficiency， wedeveloped a cigarette warehouse robot equipped with FLIR C3 infrared thermal imaging detector based on ma-chine vision. In this paper，the design of the software and hardware control system of the smoke bin robot wasintroduced， and the identification and control of the robot travel path were studied. The hardware control sys-tem adopted 32 - bit ARM processor STM32F103ZET6 ， and used OV2640 sensor to collect and transmit ima-ges. The software control system was developed based on the Microsoft Visual Studi0 2010 development envi-ronment， and called the image processing functions such as OpenCV to realize image segmentation， edge de-tection and path recognition and planning. Path recognition adopted a machine vision - based method， whichused Gaussian filtering to remove noise and interference in the image. After thresholding， the Canny operatorwas used to detect the image edge， and the edge line and center line of the lane were extracted by Houghtransform. According to the angular deviation of the lane centerline and the pixel deviation to develop a controlstrategy， the smoke bin robot achieved the self - controlled walking. The actual operation test results showedthat the system could realize path recognition and motion control. This study could provide reference for the re-lated design research of wireless communication and path recognition control robots.

Keywords Tobacco warehouse robot; Machine vision; Path identification; STM32 processor; WiFicommunication

煙叶霉变是全球烟草行业普遍存在的问题[1]。随着我国仓储烟叶的总量和规模不断增加，每年因烟叶霉变造成的经济损失达70多亿[2]。而且引起霉变的微生物所分泌的真菌毒素严重威胁人类健康。所以研究如何快速检测和预防仓储烟叶霉变具有重要意义。……