Choosing a Thermal Camera for Bat Surveys

Introduction

Bat inventory surveys are evolving. The Bat Conservation Trust (BCT) now requires night vision cameras for professional studies in the UK, and ecology firms across Europe are adopting thermal imaging for greater accuracy and reproducibility. But choosing the right model among hundreds of options is far from simple.

This guide covers the four criteria that actually matter for automated bat detection — and cuts through the specifications that don't.

Four numbers determine whether your camera will produce usable data: frame rate, NETD, FOV, and resolution.

Frame Rate: Above 25 Hz

Bats exit roosts quickly — typically 20 to 50 km/h during emergence. A camera running below 25 Hz will miss frames between positions, producing gaps in the trajectory that make tracking unreliable or impossible.

TrackBats requires a minimum of 25 Hz. 30 Hz is the practical standard; 50 Hz gives a comfortable margin at longer distances or higher flight speeds.

Thermal Sensitivity (NETD): Below 30 mK

NETD (Noise Equivalent Temperature Difference) measures the smallest temperature difference a sensor can distinguish from thermal noise. A bat in flight has a body temperature of 35–40°C. Against a stone wall at 15–20°C, the contrast is large; against a warm background or in humid conditions, it narrows.

A camera with NETD above 80 mK will struggle to separate bats from thermal clutter at any meaningful distance. Below 30 mK is the target for reliable detection; below 20 mK gives additional margin in difficult conditions.

Field of View: Adapted to Distance

The field of view determines how much of the scene a camera captures and, critically, how large individual bats appear in the frame. Too wide an angle and bats appear as single pixels at typical survey distances. Too narrow and the camera must be perfectly aligned to catch bats passing the exit.

The right FOV depends on the distance between camera and exit. Use the formula: scene width at distance D = 2 × D × tan(FOV/2). From that, the number of pixels per bat = (bat length / scene width) × horizontal resolution.

Resolution: Enough Pixels Per Bat

TrackBats needs at least 6–8 pixels across the length of a bat for reliable detection. Below that threshold, the shape is indistinguishable from noise and the model cannot consistently separate real bats from false positives.

Resolution interacts directly with FOV and distance. A 320×240 camera with a narrow FOV can outperform a 640×512 camera with a wide one, if the narrow FOV puts more pixels on the bat.

Do You Need a Radiometric Camera?

The Traditional Assumption

Traditional bat survey methods using thermal imaging were built around radiometric cameras — devices that record absolute temperature values for every pixel, not just a visible image. Radiometric cameras allow post-processing of temperature data, measurement of exact thermal signatures, and export of raw sensor data. They are also significantly more expensive, often costing several thousand euros more than equivalent non-radiometric models.

The assumption was that temperature data was essential to reliably detect bats against complex backgrounds. For manual review workflows, where an ecologist is making subjective judgements frame by frame, this made sense: the temperature overlay helped confirm that a detected object was a warm-bodied animal rather than a thermal artefact.

What AI-Based Detection Changes

Modern AI detection models like the one powering TrackBats are trained directly on thermal video frames — standard MP4 or video files from non-radiometric cameras. The model learns to recognise the visual signature of a bat in flight: its shape, the contrast against the background, and the consistency of movement across frames. It does not need access to temperature values to make that distinction reliably.

In practice, this means that a non-radiometric thermal camera producing a standard video output is fully compatible with automated bat detection software. The spatial and temporal resolution of the image matter far more than whether the camera exports raw temperature data.

When Radiometric Still Matters

Radiometric cameras remain useful for specific research applications: measuring body temperature, studying thermoregulation, or producing calibrated thermal maps of a roost structure. For standard legal emergence surveys — counting bats, mapping exits, characterising roost use — they add cost without improving the automated count result.

Cameras Used by Ecologists

No commercial partnerships. These models have been tested by TrackBats users and shown consistent performance across real survey conditions.

Useful Resources