|
Detection of standing estrus behaviour of beef cow using video images and deep machine learning |
|---|---|
| รหัสดีโอไอ | |
| Creator | Peerayut Nilchuen |
| Title | Detection of standing estrus behaviour of beef cow using video images and deep machine learning |
| Contributor | Tipmon Yaigate, Pullop Tungtrakoolsub |
| Publisher | Maejo University |
| Publication Year | 2567 |
| Journal Title | Maejo International Journal of Science and Technology |
| Journal Vol. | 18 |
| Journal No. | 1 |
| Page no. | 46 |
| Keyword | standing estrus, beef cow, YOLOv8, internet-protocol camera |
| Website title | Maejo International Journal of Science and Technology |
| ISSN | 1905-7873 |
| Abstract | A deep machine learning model, YOLOv8, was developed to detect cow estrus behaviour using video images. The model was trained using Google-Colab on 130 original images from a mobile phone camera based on original images from various scenes. The performance was compared between two models with two different annotation patterns: 1) the mounting cow model (M), involving drawing bounding boxes around cows showing signs of mounting other cows, and 2) the standing cow model (S), involving drawing bounding boxes around cows showing signs of standing estrus and cows showing signs of mounting other cows. The final trained weight of each model was used with the Python program to test the estrus behaviour video data set, using for the first time daytime and night-time videos recorded with an internet-protocol camera installed inside the cow stable area. Results revealed superior performance of the M model, exhibiting higher precision, recall, and F1-score across the YOLOv8n, YOLOv8s and YOLOv8m models when compared to the S model. Notably, the inference speed of the models ranged from 36 to 48 frames per sec., meeting the crucial requirements for fast and accurate detection of cow estrus events through video surveillance by internet-protocol cameras. A computer vision program was developed to detect cow estrus by analysing instances where a cow remained still for over 3 sec. while others attempted mounting. The system, which uses an internet-protocol camera, sends data remotely and allows farmers to monitor their cows online. Real-time notifications are sent via the LINE platform, delivering messages and images to their mobile devices. |
