基于树莓派的齿轮表面缺陷检测系统设计  

Design of Gear Surface Defect Detection System Based on Raspberry Pi

作　　者：李长齐[1] 王菡[1] 罗昊 LI Chang-qi;WANG Han;LUO Hao(School of Data Science,Baoshan University,Baoshan 678000,China;University Putra Malaysia,Kuala Lumpur 43400,Malaysia)

机构地区：[1]保山学院大数据学院,云南保山678000 [2]马来西亚博特拉大学,马来西亚吉隆坡43400

出　　处：《仪表技术与传感器》2022年第3期109-113,共5页Instrument Technique and Sensor

基　　金：云南省教育厅科学研究基金项目(2020J0699)。

摘　　要：针对齿轮缺陷人工检测错误率高、效率低的问题,研发了基于树莓派的齿轮表面缺陷检测系统。系统以树莓派、工业相机、光电传感器和剔除装置为硬件平台,搭载开发的系统检测软件,完成齿轮表面缺陷检测与剔除任务。通过CCD工业相机在线采集齿轮图像信息,通过提出的基于HSV空间的LBP特征加权融合提取方法提取图像纹理特征,通过VGG16算法对特征进行裂纹、脏污、断齿等缺陷的分类,根据分类信息由树莓派GPIO引脚控制剔除装置去除缺陷齿轮。经过实验验证,系统剔除准确率达到94.50%,平均速度为1 435 ms/次。其硬件设备体积较小、成本较低、便于维护与管理,易于安装与推广至生产线环境。Aiming at the problem of high error rate and low efficiency in manual detection of gear defects, a set of gear surface defect detection system based on Raspberry Pi was developed.The system used Raspberry Pi, industrial cameras, photoelectric sensors and rejection devices as the hardware platform, and was equipped with the developed system detection software to complete the task of detecting and removing gear surface defects.The gear image information was collected online by the CCD industrial camera, the image texture features were extracted by the proposed HSV space-based LBP feature weighted fusion extraction method, and the features were classified by the VGG16 algorithm for defects such as cracks, dirty, and broken teeth.According to the classification information, the Raspberry Pi GPIO pin controlled the rejection device to remove defective gears.After experimental verification, the system rejects accuracy rate of 95.3%,and the average speed is 1 435 ms/time.And because its hardware equipment is small in size, low in cost, easy to maintain and manage, and easy to install and promote to the production line environment.

关 键 词：缺陷检测 树莓派 VGG16 LBP特征 

分 类 号：TN919[电子电信—通信与信息系统]