基于计算机视觉的大米整精米率检测

Whole-Rice Ratio Detection Based on Computer Vision

提出了一种在线的基于计算机视觉的大米整精米率检测的新方法。采用最大方差阈值分割对大米图像进行处理,再对分割结果进行形态学操作,实现连接着的大米的分离,最后对分离后的大米二值图像进行面积和周长特征的提取。根据米粒周长像素数目的大小采取不同识别模式,当大米周长的像素数目大于某一固定值(先验值)时,即该种大米是长粒型,采取周长识别模式;短粒型大米则采取面积识别模式。通过使用这一能够智能选择识别模式的检测方法,能够大大提高整精米率的检测效率。

A method was developed for real - time determining whole - rice ratio based on computer vision. The maximum variance threshold segmentation methods was adopted to process rice image and to conduct morphological operations on the segmentation result in order to separate connected rice and to finally obtain area features and perim- eter features after we got the separated rice binary image. According to the number of pixels of rice perimeter,we de- eided whether recognition mode, perimeter recognition mode or area recognition mode should be adopted. When the number of pixels of rice perimeter was more than a fixed value, in other words, we considered rice was long - grain shape and took the perimeter recognition mode, or otherwise, the area recognition mode should be taken for short - grain rice. This kind of detection method,which could smartly choose recognition mode, would greatly improve detection efficiency of whole -rice ratio