Deep learning-based weight estimation using a fast-reconstructed mesh model from the point cloud of a pig
- Authors
- Kwon, Kiyoun; Park, Ahram; Lee, Hyunoh; Mun, Duhwan
- Issue Date
- Jul-2023
- Publisher
- ELSEVIER SCI LTD
- Keywords
- 3D reconstruction; Deep learning; Real-time weight estimation; Pigs; Point cloud
- Citation
- COMPUTERS AND ELECTRONICS IN AGRICULTURE, v.210
- Journal Title
- COMPUTERS AND ELECTRONICS IN AGRICULTURE
- Volume
- 210
- URI
- https://scholarworks.bwise.kr/kumoh/handle/2020.sw.kumoh/26384
- DOI
- 10.1016/j.compag.2023.107903
- ISSN
- 0168-1699
1872-7107
- Abstract
- In the livestock industry, securing the quality of the livestock is as important as reducing the rearing costs. Therefore, the body type and weight of livestock should be measured periodically during the entire rearing period to maintain a history of changes in body shape and weight. This study proposed the method of estimating the weight of a pig in real-time using mesh reconstruction and deep learning. The proposed method is divided into two stages, that is, generating training data by mesh reconstruction from point clouds of pigs and developing a deep neural network (DNN) to estimate weight by utilizing the training data. After acquiring 1022 point clouds for 70 pigs with the measurement equipment, the pig weight estimation experiment proceeded according to the proposed method. As a result, the mesh model could be rapidly generated within 1 s. Additionally, 48 types of measurements were determined from the mesh model, and the weight was estimated using a fully connected DNN. The results showed very high accuracy with an error of 4.89 kg (2.11% error with respect to pig weight) for the test data set.
- Files in This Item
- There are no files associated with this item.
- Appears in
Collections - School of Industrial Engineering > 1. Journal Articles
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.