1. REAL-TIME OPEN FIELD CATTLE MONITORING BY DRONE: A 3D VISUALIZATION APPROACH.
- Author
-
Fei Yang, Ningbo Zhu, Shuonan Pei, and Cheng, Cheng
- Subjects
DRONE aircraft ,AGRICULTURAL industries ,CONVOLUTIONAL neural networks ,IMAGE processing ,REMOTE sensing ,VISUALIZATION - Abstract
Monitoring a large herd across an open field is challenging in the agricultural industry but is essential for the welfare of cattle. With the advancement of Unmanned Aerial Vehicle (UAV) technology, drones are now commonly used for surveillance. In this work, we apply UAV technology and drone-captured videos to monitor cattle in open pastures. We use cow headcount as a use case. Although cattle headcount in a confined indoor environment has been studied extensively, our contribution lies in developing a framework that can localize the cows in the video and track their movements on a 3D canvas in real-time. By using a 3D visualization approach, we expect to resolve many of the occlusion issues by guiding the drone operator to navigate the drone to discover important information. We use a pre-trained Mask R-CNN classification model to detect and track cows in the video. We then use Matplotlib 3D to create the 3D canvas and display the relative cow positions. Our real-time 3D cow visualization framework allows tracking herds remotely, saving time and labor for on-site manual herding, as well as providing better global monitoring of the herd. The complete implementation can be found in our publicly available GitHub link upon request. [ABSTRACT FROM AUTHOR]
- Published
- 2021