基于图像分析的大米尺度检测方法

Method for Measuring Kernel Size Based on Rice Image Analysis

提出一种基于图像分析的大米实际尺度检测实用方法。在实物旁边放置直尺,用小波变换提取刻度,从而求出两像素点间的实际距离;再由二值化后的大米区域与边缘曲线求出大米的平面投影面积、长短轴与周长。根据数字图像的特点,对边缘曲线作适当的抛光处理,使周长的计算接近实际。实验结果表明,所得结果与实物尺寸基本相符,可在大米分类检测中使用。

A method for measuring the actual kernel size of rice grains by dint of image analysis is presented. A ruler is put beside the object, and the calibration is abstracted through wavelet transforming so that the actual distance between two pixels is taken. Then the binary plane projective area of the rice kernels, their long and short axes as well as circumference are obtained as per the rice zone and edge curves. According to the characteristics of a digital image, the edge curves are smoothed appropriately in order to make the calculation of the circumference closer to the reality. The experiment shows that the result is basically conform to the real size of the object, and this method can be used in the inspection for rice classification.