Alternative TitleWeight measurement techniques for dairy cows under free posture conditions using mathematical models and image analysis
Note (General)Body weight is an important indicator for monitoring the growth and health conditions of cattle. It is also frequently used as an indicator for evaluating livestock productivity and breeding selection. However, weighing is not widely carried out due to the need for expensive equipment, the labor intensive and the risks involved in measurement. Therefore, more effort-saving and simple weight estimation techniques need to be developed. If weight measurement technology can be developed using images taken of cows, expensive measuring equipment will no longer be necessary. In addition, the need for contact with the cow's body will be eliminated, which will lead to reduced labor and safety in the measurement process. Thus, this paper aims to develop a new weight estimation technique for dairy cows under free-posture conditions using images, and therefore, a photography technique for weight measurement was developed (Study I), and a cow weight estimation model considering free posture and growth stage was constructed (Study II). In Study I, we proposed a technique for photographing cows raised under grazing pasture and free barn conditions. By using a simple gimbal and a long stick, the camera could always be pointed vertically to the top of the cow. This photographic equipment is lightweight, making it portable to the free barn and pasture, and can acquire upper body images in the standing and lying positions. Therefore, it is expected to be applied to weight estimation technology from images. In addition, this imaging device is inexpensive and easy to make. Trainable Weka Segmentation (TWS), a machine learning technique, was shown to be effective in extracting cow body regions from images in grassland. This technique can extract body regions of different cows in free-posture, using a training model created from only one individual. In Study II, dairy cows at different life stages were photographed from above in both lying and standing positions. Several body-shape indices were measured based on the captured images, and relationships with body weights were analyzed. From these relationships, we created models to estimate body weights using shape indices based on all of the data for each life stage of multiparous cows and heifers of more/less than 250 kg. In both standing and lying positions, area measurements of the upper portion of the body had the highest correlations with body weight. In the model using all of the data, predicted body weights varied most for cattle that were multiparous, and were underestimated for low-body weight cattle. In the models of the standing and lying positions created for each life stage, body weight was estimated with high accuracy, except in the case of multiparous cattle. No bias was shown between body weights in the standing and lying positions. These results imply that cow body weights can be estimated using photographs of standing and lying cattle. These results indicate that the image-based dairy cow weight measurement technique presented in this study has advantages not found in existing weight measurement techniques, such as: 1) measurement can be done in the presence of cows in pastures or barns, 2) measurement can be done under free posture conditions of standing or lying cows, 3) the labor and risk involved in gathering, leading, and holding cows in posture during measurement can be eliminated, and 4) stress on cows during measurement is reduced.
新大院博(農)第234号
Collection (particular)国立国会図書館デジタルコレクション > デジタル化資料 > 博士論文
Date Accepted (W3CDTF)2023-06-02T22:06:20+09:00
Data Provider (Database)国立国会図書館 : 国立国会図書館デジタルコレクション