基于优化矩不变特征的鲜切菜在线分级技术

    Online grading method for fresh-cut vegetables based on optimal invariant moment

    • 摘要: 为了准确地从复杂背景中分割鲜切菜,该文将矩不变特征的平移、旋转不变性与几何特征优秀的尺度表达能力融合,提出了一种基于优化矩不变特征模型的鲜切菜在线分级方法。首先提取标准鲜切菜的几何特征和矩不变特征,并利用其统计信息建立数学模型,然后提取待分级的鲜切菜的几何特征和矩不变特征,最后计算待分级的鲜切菜特征与标准鲜切菜特征模型的相似度,完成鲜切菜的在线分级。以鲜切土豆丁为例,采集11幅图片共679个土豆丁样本,分类的准确率达到99.40%,处理一帧耗时0.108 s。试验结果表明,该文方法能准确地将鲜切菜分级,实时性和鲁棒性高,为鲜切菜在线分级提供参考。

       

      Abstract: In order to extract accurately fresh-cut vegetables from complex background of the image, an online classification method of fresh-cut vegetables was studied, which was based on the optimized moment invariant features model and integrates the two aspects of the moment invariant features?(translation and rotation) and the excellent scale expressing ability of the geometrical features. First, geometrical and moment invariant features of the standard fresh-cut vegetables was got, and a mathematical model was set up with statistics,; then two types features of fresh-cut vegetables to be graded was extracted; finally, similarity between the features of the sample and that of the standard fresh-cut vegetables so as to accomplish the online classification was made out. Take example of the fresh-cut pieces of potatoes, 11 images with 679 pieces of potatoes was disposed, and accuracy of the classification up to 99.40% , while it takes only 0.108 seconds to work on one image. The experiment shows that the method can classify exactly ,and have good real-time performance and robust result, and it can provide reference for the online classification of fresh-cut vegetables.

       

    /

    返回文章
    返回