基于机器视觉技术的小米品质检测方法研究

Research on Detection Method of Millet Quality Based on Machine Vision Technology

为实现小米品质的快速鉴评,以“豫谷18”新米和陈米为研究对象,利用工业相机采集出样本的RGB图像,通过图像处理裁剪出同样大小感兴趣区域(ROI)图像块。由于新米与陈米在颜色方面差异较大,为更好提取颜色特征,将RGB图像转换到HSV颜色空间并提取HSV颜色空间下ROI区域的色调(H)和饱和度(S)对应的中心矩特征。根据特征值的分布图,筛选出更易区分的特征,组成特征向量作为分类器的输入,建立基于支持向量机(SVM)的识别模型。通过留一校验方法进行模型的训练测试,结果表明,当选择饱和度(S)中心矩特征作为特征向量输入模型时,识别率达到95%,且耗时较少,可以应用于小米的品质检测。

In order to study the rapid detection and identification of millet quality, taking the latest millet and aging millet of the “Yugu 18” as the research object, the RGB image of each sample is captured by an industrial camera, and the same size region of interest (ROI) image block is cropped by image processing. Since the latest millet and aging millet differ greatly in color, in order to extract the color features better, the RGB image is converted to the HSV color space and the central moment corresponding to the hue (H) and saturation (S) of the ROI region under the HSV color space is extracted. In terms of the eigenvalue distribution map, feature vectors with distinct features are selected as input of classifier, and a detection model based on support vector machine (SVM) is established. Training and testing the model by leave-one-out cross-validation method, the results show that when the center moment feature of saturation (S) is selected as the input model of eigenvector, the recognition rate can reach 95% and it takes less time, which can be applied to millet quality detection.