Development of a Method for Pig Weight Measurement by Image Analysis.

9.提出了一种基于图像分析的猪体重估测方法。从猪的指定投影面积与其活重的关系出发进行分析。323头猪的特定投影图像面积（不包括耳和尾图像面积）与体重呈指数关系。试验猪由长白猪、杜洛克猪、大白色猪和汉普夏猪共12个杂交种组成，体重4.2 ~ 114.2 kg。这种关系取决于一种类型的两个杂交。回归方程中的差异主要是由树高差异引起的，而不是由体长或平均宽差异引起的。对特定投影面积、身高与体重的多元回归方程进行因次分析，发现最接近三维体积。因此，如果能间接测量猪的身高，用指定的投影面积和高度的多元回归方程估计猪的体重，其精度约为 *5%。尝试用灯向猪投射光线，用图像分析法测量猪肩处的阴影长度，利用阴影长度与光线入射角的几何关系估计猪的身高。本试验选用7头体重在47.6 ~ 107 kg之间的猪，与直接测量猪高的方法进行比较。估计5头猪的身高，平均误差为 * 4.6%。然而，两只猪的误差超过 *10%。这就是为什么阴影在腹部而不是在。肩膀被测量。因此，利用光投影和图像分析技术间接估计猪的身高。该方法可以快速估计猪的体重。

9. A method was developed to estimate pig weight with image analysis. Analysis proceeded from the relationship between specified projected area of a pig from above and its live weight. Specified projected image area, which lacked ears and tail image areas, had an exponential relationship with weight for 323 pigs. The experimental pigs, weighing from 4.2 kg to 114.2 kg, consisted of 12 kinds of two-crossbreeds of Landrace, Duroc, Large White, and Hampshire. The relationship depended on a type of two-crossbreed. Height difference was the major reason of the difference in the regression equation rather than length or mean width differences. Dimensional analysis in the multiple regression equation of specified projected area and height with weight conducted that it was most close to 3 dimension, meaning volume. Hence if pig height could be measured indirectly, it will be achieved to estimate pig weight with about *5% accuracy using the multiple regression equation of specified projected area and height.Projecting light on pig with a lamp and measuring shadow length at shoulder of the pig with image analysis were attempted to estimate pig height using a geometrical relation between shadow length with light incident angle. Seven pigs, weighing from 47.6 kg to 107 kg, were experimented to compare with direct measurement of pig height. Heights for five pigs were estimated with a mean error of*4.6%. However more than *10% error were obtained for two pigs. It was the reason that shadow at belly not at. shoulder was measured. Thus, an indirect method was conducted to estimate pig height using light projection and image analysis technique. The method allows for rapid estimation of pig weight.