English

中国农机化学报

中国农机化学报 ›› 2024, Vol. 45 ›› Issue (9): 104-110.DOI: 10.13733/j.jcam.issn.2095-5553.2024.09.016

• 农产品加工工程 • 上一篇    下一篇

种子破损率快速检测方法研究

杨浩勇 1,王超柱 1,刘浩义 2,关心桐 3,刘璎瑛 2,丁永前 2   

  1. (1.江苏省农业机械试验鉴定站,南京市,210017;2.南京农业大学人工智能学院,南京市,210031; 3.南京农业大学工学院,南京市,210031)
  • 出版日期:2024-09-15 发布日期:2024-09-02
  • 基金资助:
    江苏省现代农机装备与技术示范推广项目(NJ2021—33)

Research on a rapid detection method for seed breakage rate

Yang Haoyong1,Wang Chaozhu1,Liu Haoyi2,Guan Xintong3,Liu Yingying2,Ding Yongqian2   

  1. (1. Jiangsu Agricultural Machinery Testing and Appraisal Station,Nanjing,210017,China; 2. College of Artificial Intelligence,Nanjing Agricultural University,Nanjing,210031,China; 3. College of Engineering,Nanjing Agricultural University,Nanjing,210031,China) 
  • Online:2024-09-15 Published:2024-09-02

摘要:

旋耕播种机推广鉴定时,需要人工检测种子的破损率。为提高检测效率,以小麦种子为例,对种子破损率快速检测方法进行研究。设计种子破损率自动检测平台,可一次性采集 50 g小麦种子图像,基于图像处理技术和机器学习方法,提取小麦种子图像 13个形状特征和 8个纹理特征,建立基于特征的种子破损识别模型;识别的破损种子图像与种子质量的关系,建立基于图像的破损种子质量预测模型,按照鉴定大纲要求实现小麦种子破损率的快速检测。对江苏省 “A888”“丹阳 001”“泰州 014” 4个品种的小麦种子破损率进行试验测试, 4个品种小麦种子破损“无锡 004”每个品种 3次取样测定。结果表明:率自动检测的平均相对误差分别为 0. 08%、0. 07%、0. 06%、0. 08%,检测的相对均方根误差为 0. 08%,检测平均时长为 5. 216 s。该研究实现小麦种子破损率自动、快速检测,节省农机鉴定时间,推动农机鉴定过程的标准化和智能化。

关键词: 小麦种子, 旋耕播种机, 图像处理, 破损率, 机器学习

Abstract:

When the rotary tillage seeder is popularized and appraised,it is necessary to measure the damage rate of seeds manually. In order to improve the detection efficiency,this paper takes wheat seeds as an example to study the rapid detection method of seed breakage rate. The automatic detection platform of seed breakage rate was designed,which could collect 50 g wheat seed images at one time. Based on image processing technology and machine learning methods,13 shape features and 8 texture features of wheat seed images were extracted,and a feature.based seed damage recognition model was established. The relationship between the identified damaged seed images and seed quality was studied,and an image.based broken seed quality prediction model was established,and the rapid detection of wheat seed damage rate was realized according to the requirements of the identification outline. In this study,the seed breakage rate of four wheat varieties as “Danyang 001”“A888”“Taizhou 014”and“Wuxi 004”in Jiangsu Province was experimentally tested,and each variety was sampled and measured for three times. The experimental results showed that the average relative errors of the automatic detection of seed breakage rate of the four wheat varieties were 0. 08%,0. 07%,0. 06% and 0. 08%,respectively,and the relative root mean square error of the detection was 0. 08%,and the average detection time was 5. 216 s. In this study,the automatic and rapid detection of wheat seed damage rate was realized,which saved the time of agricultural machinery identification and promoted the standardization and intelligence of agricultural machinery identification process.

Key words: wheat seed, rotary tillage seeder, image processing, breakage rate, machine learning

中图分类号: