基于机器视觉的种子品种实时检测系统研究

Research on Real Time Identification of Seed Variety by Machine Vision Technology

机器视觉技术的快速低成本实时检测杂交水稻种子品种新系统包括一组可以自动上料、光照箱、图像采集卡、CCD摄像头和自动下料等硬件以及图像处理及品种识别软件组成。系统的实现过程为:通过上料斗自动装入种子,种子由单片机控制的输送装置进入光照箱,每隔2秒停止一次,CCD摄像头采集图像,图像被读入内存,通过软件系统处理种子的图像,并提取品种的特征参数。通过此系统识别100粒种子的时间为5秒,品种的识别率分别达到了:先农5号为99.99%,金优桂为99.93%,优166为98.89%,先农3号为78.82%,中优463为86.65%。

A new, quick and low-cost method and a system for inspecting the variety of hybrid rice seeds was developed. This automatic inspect system includes automatic uploading device, light source camber, CCD camera and conveying device, hardware and software for image collecting and processing variety identification. Hybrid rice seeds was loaded by upload device and conveyed, controlled by single-chip device, which is stopped every other two seconds to collect image. The image was extracted its feature by the software and varieties was identified by BP neural network. The identifying rates of five rice varieties are： No.5 ‘Xiannong＇ of 99.99%, ‘Jinyougui＇ of 99.93%, ‘You 166＇ of 98.89%, No.3 ‘Xiannong＇ of 78.82%and ‘Medium you 463＇ of 86.65%.