Design of Pig Inventory and Abnormality Monitoring System Based on Livestock Internet of Things Orbital Inspection Robot

Abstract

Pig inventory is an important task in intensive asset management of pig farms. Accurate quantity can realize precise feeding, help reduce farming cost and improve farming efficiency. The realization of pig inventory through computer vision algorithms has become the main development trend of intelligent pig inventory methods [1]. At present, pig inventory mostly adopts manual inventory, which is costly, inefficient and prone to errors. Some of the existing large pig farms use camera-based pig inventory, but they are based on fixed cameras, and cannot complete the inventory of the entire pig farm. Meanwhile, sometimes pigs will jump out of the pen and can't eat, which is very dangerous especially for piglets. Design a track inspection robot carrying a camera to complete the inspection of pig farms, according to the pen information on the pig data collection, this project collected 2000 valid pictures to form a dataset, through the YOLOv5 target detection algorithm to train it, learn a training model, in order to improve the accuracy of inventory, enhancement processing of the photos, through the deep learning model training to obtain the optimal model and deployed, and finally Data correction is performed by Kalman filtering method to improve the inventory accuracy, and this project is carried out on-site experiments in pig farms of Zhengzhou Xinrong Agricultural and Animal Husbandry Information Co. The experimental results show that the overall accuracy of pig inventory is 95.6%. Compared with manual inventory, this method has low cost and good application prospect. When pigs are found in the channel or non-pen labeling, it is an abnormal situation, and the relevant personnel will be notified in time to deal with it, so as to ensure the safety of farming.