Development of an Artificial Intelligence Deep Neural Network for the Identification of Individual Animals

Abstract

Monitoring of individual animals and their feeding habits can pose a challenge to farmers to prevent under- or overfeeding of individual animals. A system that can identify individual animals and allows feeding or medicating of a specific animal will increase productivity, profitability and animal condition. The objective of the research is the development of an automated system, making use of video and image processing techniques, as well as deep learning neural networks, to identify individual animals. Using Matlab^®, Python and Google Colab the principles of neural networks (NN) and Artificial Intelligence (AI) learning was tested and evaluated on a personal computer and will later be implemented on suitable portable hardware. Results will be verified and the suitability for such a project will be tested and judged by evaluating the accuracy of the recognized images. Initial tests indicated that an Artificial Neural Network (ANN) proved to be sufficient and could be run on compact hardware such as the NVIDIA Jetson Nano. A small amount of learning material could be used and proofed to be successful in identifying cattle heads on such a device.