Development and implementation of a training dataset to ensure clear boundary value of body condition score classification of dairy cows in automatic system

文献类型： 外文期刊

第一作者： Tao, Ya

作者： Tao, Ya;Li, Feng;Sun, Yukun

作者机构：

关键词： Body condition score; Ultrasonographic backfat thickness; Repeatability; Image model

期刊名称：LIVESTOCK SCIENCE （ 影响因子：1.929； 五年影响因子：2.279 ）

ISSN： 1871-1413

年卷期： 2022 年 259 卷

页码：

收录情况： SCI

摘要： Recent studies have focused on the automatic body condition score (BCS) system to completely eradicate labor input and increase identification accuracy, but the effects of the problems of subjectivity and discreteness on automatic systems were ignored. The current study aimed to evaluate the repeatability of regressive BCS in the use of different operators and compare regressive BCS datasets with manual BCS datasets to test the effects on automatic model classification. Four assessors with same ultrasonic detector were used to evaluate BCS and determine BFT in a pen. The Kappa coefficient (K-w) and agreement between assessors, across all assessors, and for the average of the assessors within 0.25 units of the exact score were used to evaluate the repeatability. Subsequently, for the establishment of automatic models, the small DenseNet was composed of a depth channel to analyze the top-view images of 524 lactating cows. The results showed that the implementation of regressive BCS increases agreement to a perfect level (K-w= 0.93). Importantly, the use of average scores was not expected to significantly reduce the dispersibility, with an R-2 of 0.89 relative to manual BCS and a maximal R-2 of 0.87. Additionally, relative to the manual training dataset of regressive BCS, a significant increase of 0.58 in accuracy was found for the model classification. In conclusion, the measurement of ultrasound BFT by using a regressive method should contribute to the development of an automatic BCS system that can be used in place of the average assessors' BCS, which can increase the repeatability when evaluating the body fat reserve and improve the accuracy of the visual model.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序