2019/07/11 by William P. Andrew, Andrew, William, Colin Greatwood +3 · 2 citations
Agricultural and Biological Sciences · Biochemistry, Genetics and Molecular Biology · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Food Supply Chain Traceability #Identification and Quantification in Food #Milk Quality and Mastitis in Dairy Cows #Robotics (cs.RO)
paper · pdf · doi:10.48550/arxiv.1907.05310
openalex publication_date 2019/07/11 · openalex created_date 2022/07/20 · openalex updated_date 2026/07/28
This paper describes a computationally-enhanced M100 UAV platform with an\nonboard deep learning inference system for integrated computer vision and\nnavigation able to autonomously find and visually identify by coat pattern\nindividual Holstein Friesian cattle in freely moving herds. We propose an\napproach that utilises three deep convolutional neural network architectures\nrunning live onboard the aircraft; that is, a YoloV2-based species detector, a\ndual-stream CNN delivering exploratory agency and an InceptionV3-based\nbiometric LRCN for individual animal identification. We evaluate the\nperformance of each of the components offline, and also online via real-world\nfield tests comprising 146.7 minutes of autonomous low altitude flight in a\nfarm environment over a dispersed herd of 17 heifer dairy cows. We report\nerror-free identification performance on this online experiment. The presented\nproof-of-concept system is the first of its kind and a successful step towards\nautonomous biometric identification of individual animals from the air in open\npasture environments for tag-less AI support in farming and ecology.\n