摘要
荔枝的产后品质检测是进入市场前的一个重要工序,对提高荔枝商品化的处理水平、增强其市场竞争力以及提高其附加价值具有重要的意义。基于机器视觉的检测方法能在无损伤的前提下,快速准确的进行检测,具有良好的可靠性。为此,基于机器视觉技术,提出了一种荔枝品质快速自动检测的方法。利用深度学习中的SSD算法结合多视窗检测方法,对视觉获取的图像进行荔枝快速品质检测。建立了一套荔枝品质自动检测系统,以"桂味"荔枝作为实验对象,在模拟车间环境下进行多组实验,平均识别准确率为93.3%,其中成熟品、未成熟品和爆裂品的识别准确率分别为95.6%、93.2%和91.1%,平均识别耗时40 ms,可以实现产后荔枝的品质快速自动化检测。
Postpartum quality detection of litchi is an important process before entering the market. It is of great significance to improve the processing level of commercialization of litchi, enhance its market competitiveness and enhance its added value. The detection method based on machine vision can detect quickly and accurately without damage, and has good reliability. Therefore, based on machine vision technology, a fast automatic detection method for litchi quality was proposed. Using SSD algorithm in depth learning and multi-window detection method, litchi fast quality detection is carried out on the image acquired by vision. A set of automatic quality detection system for litchi was established. Taking "Guiwei" litchi as the experimental object, a series of experiments were carried out in a simulated workshop environment. The average recognition accuracy was 93. 3%. The recognition accuracy of mature, immature and explosive products were 95. 6%, 93. 2% and 91. 1% respectively, and the average recognition time was 40 ms. The rapid automatic quality detection of postpartum litchi could be realized.
作者
周伟亮
王红军
邹湘军
Zhou Weiliang;Wang Hongjun;Zou Xiangjun(College of Engineering,South China Agricultural University,Guangzhou,Guangdong,510642,China)
出处
《中国农机化学报》
北大核心
2020年第1期144-147,204,共5页
Journal of Chinese Agricultural Mechanization
基金
国家重点科技计划(2017YFD0700100)
广东省公益与能力建设项目(2016A010102013)
国家自然科学基金(51705365).
关键词
荔枝品质检测
机器视觉
SSD算法
多视窗检测
litchi quality detection
machine vision
SSD algorithm
multi-window detection
