Sheep Identification using IOT and Machine Learning

Background

Kauricone commissioned AUT students to use our technology to identify sheep in a paddock with 3 objectives

1. Take images of a flock of sheep at regular intervals to identify the number of sheep in the paddock
2. Using Tensorflow, identify each individiual sheep in the flock
3. Monitor the change in the grass level, and notify the farmer when it is time to move the flock

Hardware 

1. Kauricone IOT Server, 4GB, 128GB eMMC Memory, ARM 6 Core Processor
2. Camera

Software

1. Ubuntu 18.04
2. Python V3
3. Tensorflow Light V1.14
4. MySQL Database

Outcomes

 

IOT and Machine Learning to Identify Progency

Kauricone collected images from a farm with the intention of identifying Ewes to Lambs to easily identify Progeny for selective breeding

Using IOT and Machine Learning, we can accurately identify Ewes and Lambs to over 99%. Images can be collected at regular intervals to provide growth rates, eating patterns, or ailments

 

 

 

IOT, Machine Learning, and Drone to Identify Sheep numbers, and Pasture

The images below are collected from a Drone, and are too high to identify each sheep accuratly. We expect lower level images to have a better result. However using drone images on a regular basis, it would be easy to Dry Matter, and Sheep Numbers in paddocks both near and far

Machine learning also looks for exceptions to the norm. Stray dogs, pests, and unwanted visitors can raise alerts for the farmer