Automated Nest Monitoring

The Problem

Monitoring of reproductive biology of birds over many seasons can help reveal impacts of climate change, habit degradation and other major stresses on bird populations. However tracking detailed brood metrics such as when nesting occurs, number of eggs laid, how many eggs hatch, and how many hatchlings survive requires frequent monitoring and significant time commitment. In many regions nest boxes are regularly placed hundreds of meters apart and lines of boxes can extend over many kilometers. The resources required to monitor many nests limit the number of visits that can be made and can reduce the accuracy of the various key metrics such as egg laying dates.

Our Proposal

The plan is to develop a system for automating the monitoring of bird nests using computer vision and machine learning at the nest site and reporting the results over long range radio. Three different camera and computer options will be evaluated (OpenMV H7, ESP-eye, Raspberry Pi Camera) for cost, computational capability and power usage. The extracted data will then be sent over a lora mesh network to LoraWAN gateway.

We Assume that...

Assume its possible to perform required image analysis on inexpensive computers such as Raspberry Pi.

That an economical solution can be arrived at that is lower than the cost in time and fuel to visit nest boxes.

Constraints to Overcome

Automating nest monitoring using computer vision with reporting over LoRa will provide finer scale information on the brood as it develops while reducing monitoring effort and fuel expenditure. New more capable consumer electronics are becoming more accessible. Edge computing with the latest OpenMV cameras with onboard deep learning may allow for extracting the metrics currently done by site visit such as identifying species, brood stage and counting number of eggs or nestlings. To train deep learning models requires large set of classified images which will be a major limitation constraint.

Current Work

Acquire and test different cameras for placement in nest boxes. OpenMV H7, ESP eye, Picamera. Capture as many pictures of different species and stages of brood development. Develop machine learning project to identify species, brood stage, and counting eggs, hatchlings. Optimize power management of the electronics system.

Current Needs

Resources: Cameras, radios, LoRaWAN Gateway Skills: Machine learning model development, electronics