基于机器视觉的咖啡生豆快速检测与分级研究

Research on rapid detection and grading of green coffee beans based on machine vision

中文关键词:  机器视觉  图像处理  主成分分析  智能分级

英文关键词:machine vision  image processing  principal component analysis  intelligent grading

中文摘要:

      目的 建立基于机器视觉的咖啡生豆快速检测分级方法, 实现咖啡生豆的快速无损等级评估。方法 本研究提取咖啡生豆图像预处理后的形状和颜色共13种特征, 首先通过特征分布图分析与主成分分析找到13类缺陷豆的显著特征, 并确定其显著特征值范围来判定缺陷豆, 然后对正常豆按粒度大小判断其等级, 最后在MATLAB App Designer平台上设计了咖啡生豆快速检测与分级界面。结果 利用机器视觉技术能很好地识别各个类别咖啡生豆, 检测一级咖啡生豆准确率为94.77%、二级为93.20%、三级为95.85%, 13类缺陷豆平均准确率为82.25%, 咖啡生豆检测平均准确率达到91.52%, 检测300 g咖啡生豆总用时25.3156 s。结论 本方法提高了咖啡生豆分级的智能化水平, 分级过程平稳且快速, 为今后咖啡生豆在线检测分选提供了技术支持。

英文摘要:

      Objective To establish a rapid detection and grading method for green coffee beans based on machine vision, and achieve quick and non-destructive quality assessment of green coffee beans. Methods This study extracted 13 kinds of distinct features from pre-processed images of coffee beans, encompassing both shape and color. Firstly, significant features for 13 categories of defective beans through feature distribution ana-lysis and principal component analysis were identified. The range of significant feature values were determined to classify defective beans. Subsequently, normal beans were graded based on their size. Finally, user interface for rapid detection and grading of coffee beans were designed by using the MATLAB App Designer platform. Results Machine vision technology could be used to identify various types of green coffee beans well, and the detection accuracy of first-level green coffee beans was 94.77%, that of second-level coffee was 93.20%, and that of third-level coffee waws 95.85%, and that of 13 types of defective beans was 82.25%, and that of green coffee beans was 91.52%, and the total time taken to detect 300 g of green coffee beans was 25.3156 s. Conclusion The intelligent level of the classification of raw coffee beans is improved, and the classification process is stable and fast, providing technical support for the online detection and sorting of raw coffee beans in the future.