基于图像分析的大米形状识别

提出了一种基于计算机图像分析的大米米胚脱落与颗粒破碎识别新方法。首先,将边缘曲线变换为极坐标形式,采用椭圆模板定位目标,获得一组与目标平移、旋转和尺度无关的形状描述数据;再运用小波变换提取奇异点及特征参数。取100颗样品进行验证,结果表明,该方法用于目标定位与识别时的鲁棒性强,可信度高。对米胚脱落、米粒破碎的判定正确率分别为98.9%、100%。

Rice Figure Identification Based on an Image Analysis

A new method for identifying embryo break-off and kernel fragmentation of rice was put forward. Firstly, the edge curve was transformed into a polar coordinate form, and a group of data for describing the figure was obtained through positioning the objects by an ellipse template. The data were independent from shift, rotation and scale. Then the singular points and compute character parameter were detected by wavelet transformation. The experiment results of 100 samples showed that the method could be used for positioning the objects with a strong robustness and high reliability. The identification correctness of rice embryo break-off and kernel frag- mentation was 98.9% and 100% respectively.