基于线性拟合模型的苹果大小分级方法

Apple Size Grading Method Based on Linear Fitting Model

针对苹果大小分级图像处理过程中苹果直径像素数与实际直径映射不准确的问题,设计了一种基于线性拟合模型的苹果大小分级方法。将图像色彩表达方式由RGB(red green blue,红、绿、蓝)颜色空间转换到HSV(hue saturation value,色调、饱和度、明度)颜色空间,利用H、S、V三项阈值从图像中分割出苹果区域作为苹果最大横切面,提取苹果横切面轮廓并过滤图像中噪声轮廓。用最小外接圆法求苹果轮廓最小外接圆,其直径作为苹果直径。按苹果实际直径大小的分布规律从总体样本中选取数据作为建模集,计算建模集中苹果直径像素数,用最小二乘法将直径像素数与实际直径拟合为线性模型。将直径像素数输入模型,计算得到的苹果直径数值用于大小分级。用75组数据测试模型,得到92%的总体分级准确率及0.56 mm的平均测量误差。该方法分级准确率高,直径测量误差小,能满足苹果产后分级需求。

Aiming at solving the problem that pixel number of diameter was inaccurate corresponding to real diameter in image processing of apple size grading,an apple size grading method based on the linear fitting model was designed.Images were converted from the RGB color space to the HSV color space.The apple region was segmented from the image as the largest cross section with the values of H,S and V.The apple contour was extracted and the noise contours were filtered out.The minimum circumscribed circle of apple contour was calculated by the minimum circumscribed circle method,and its diameter was taken as apple diameter.According to the distribution law of the real diameter of apples,the data was selected from the overall sample as the modeling set.The pixel number of the apple contour diameter in the modeling set was calculated,and the number of diameter pixels and the real diameter were fitted to the linear model by the least squares method.The diameter pixel number was input into the model,and the calculated apple diameter value was used for apple size grading.Seventy-five sets of data were tested to verify the accuracy of the proposed method,the grading accuracy was 92%and the average measurement error was 0.56 mm.The method had high classification accuracy and small diameter measurement error,which could meet the needs of apple postnatal grading.