基于机器视觉的马铃薯图像特征参数获取方法n

Method for Obtaining Characteristic Parametersof Potato Image Based on Machine Vision

目前马铃薯的分级方法主要还是以人工为主,而人工分级普遍存在不稳定性,费时成本高,依赖人的主观经验等.近些年来,机器视觉技术在农业领域取得了广泛的应用,尤其是农产品分级领域取得了较多的成果.机器视觉技术具有无损害、经济性好、精确率高等优点.对于马铃薯分类,可以从外观尺寸上判断马铃薯的形状类型和重量范围.人工分级时,通过观察不同马铃薯样本的不同横截面大小,就可以主观上判断出马铃薯的形状类型,大致评估质量范围.基于人工分级的特点,通过机器视觉进行分级,只需获得不同横截面的的投影,通过提取图像特征参数的方式来对这些横截面进行定量分析,就可以模拟人工分级的模式对马铃薯进行基于机器视觉的自动分级.本文设计了一套机器视觉系统,分别获取马铃薯平放时的图片与竖放时的图片,以获取最大横截面与最小横截面的投影.采集图片并对得到的照片进行图像处理,得到只有目标区域和背景的二值化图像.从一副图像中获得9个特征,然后通过100个马铃薯样本得到了基于这些特征的数据集.利用软件Unscramble对这些数据集进行分析,以验证方法的合理性.

Potato is the fourth most worldwide crop in the world after wheat, corn and rice. Classification of potatoes is thought to producegreater economic benefits. The current grading methods mainly relies on artificial methods which is not stable and expensive. Inrecent years, machine vision technologies have been widely used in the field of agriculture, especially achieved great results in thefield of agricultural products classification. Machine vision could classify different varieties of agricultural products based on color,volume, quality and other characteristics. For the same variety of potatoes, the shape type and weight range could be graded by theexternal dimensions. For this purpose, a set of machine vision system was designed, including CCD camera, image acquisitioncard, computer system, LED light source, two mirrors placed into V type, as well as MATLAB software for image processing andUnscramble software for data analysis. Photos were processed into binary images with MATLAB software. Each number of pixelsof target area can be counted as its area. The minimum external rectangle of each target area was obtained for calculating its lengthand width. The area, length and width were regarded as the image characteristic parameters of the target area. A binary imagecontains three target areas, for which a photo can be obtained nine characteristic parameters. Data sets based on thesecharacteristics were obtained from 100 potato samples. These data sets were analyzed to verify the correctness of the method withUnscramble software.