Design of Tan Sheep intelligent feeding navigation system based on RTK and deep learning

An driverless feeding cart is designed for the roblems of high cost and low efficiency of manual feeding in m gxia Tan sheep breeding process. In this paper, GPS-BeiDou RTK navigation technology is used. The feeding cart mobile station receives GPS satellite signals and reference station signals, and locates the current position of the feeding cart with centimeter -level accuracy through real-time carrier phase difference technology, while combining deep learning YOLO vision algorithm to assist RTK navigation. By training a large number of relevant obstacle datasets from Tan sheep farms, the vision system achieves the effect of being able to recognize specific catepries of obstacles, thus achieving obstacle avoidance of the feed cart. In addition, if the RTK navigates the feed cart in an area where the GPS positioning signal is unstable, deep learning can be used to navigate through visual label localization combined with multi -sensor fusion techniques.