thumbnail

Topic

Technologies and technical equipment for agriculture and food industry

Volume

Volume 59 / No.3 / 2019

Pages : 125-132

Metrics

Volume viewed 42 times

Volume downloaded 24 times

ESTIMATION OF LOSS RATE OF OATS CLEANING BASED ON WATERSHED SEGMENTATION

基于分水岭分割的燕麦清选损失率估计研究

DOI : https://doi.org/10.35633/inmateh-59-14

Authors

Yan Hongwen

College of Information Science and Engineering, Shanxi Agricultural University, Taigu

(*) Cui Qingliang

College of Information Science and Engineering, Shanxi Agricultural University, Taigu

Deng Xuefeng

College of Information Science and Engineering, Shanxi Agricultural University, Taigu

(*) Corresponding authors:

[email protected] |

Cui Qingliang

Abstract

This paper studied the loss rate of oats in the process of cleaning from the perspective of image processing. The sample was divided into group a that contained no impurities and group b that contained impurities. Otsu method was used to segment the oat kernels, with the recognition rate reaching 94.20%, and morphological opening was used for the openings appearing during the segmentation process for filling, while watershed segmentation algorithm was used for segmentation of adhesion area, with the recognition rate reaching 98.50%. For group b, the area method was used to identify and separate the impurities. Through statistical analysis, the area threshold was 600 pixels, and impurities could be removed without excessive segmentation. The estimated 5 g-sample loss rate in group a was 2.08%, which met requirements, so 5 g-sample was selected in group b, and it was calculated that the estimated loss rate of group b was 2.60%, The study showed that having good effect on image processing with less adhesion after cleaning, the algorithm could provide theoretical and methodological support for on-line monitoring of loss rate during oats cleaning.

Abstract in Chinese

本文从图像处理角度研究燕麦在清选过程中的损失率。将样品分为不含杂质的a组和含杂质的b组,使用Otsu方法分割燕麦籽粒,识别率达94.20%。并对分割过程中的孔洞使用形态学开运算填充,对粘连区域采用分水岭分割算法进行分割,识别率达98.50%。对b组采用面积法对杂质识别并分离,经统计分析,面积阈值为600个像素,可去除杂质且不会出现过度分割。a组中5g样品损失率估计值为2.08%,满足要求,故在b组选取5g的样品,计算b组损失率估计值为2.6%。研究表明,该算法对于清选后粘连较少的图像处理效果好,可对燕麦清选损失率的在线监测提供理论和方法支持

Indexed in

Clarivate Analytics.
 Emerging Sources Citation Index
Scopus/Elsevier
Google Scholar
Crossref
Road