As users continue to expand and evolve their security systems, many have begun integrating thermal cameras to improve their overall situational awareness at the perimeter. As more visible camera manufacturers are introducing thermal offerings, it is important to understand what differentiates an entry level camera from a higher performing one.
Those familiar with visible cameras understand that resolution and available pixels determine how well analytics will work for the necessary application. While resolution is also important with thermal cameras, it is crucial to understand how thermal sensitivity or noise equivalent temperature difference (NETD), impacts the performance of the camera.
Thermal sensitivity is another way of referring to NETD, which is the smallest temperature difference you can see when using a thermal device. The value is measured in milliKelvin (mK) and is essentially a noise rating of the system.
Ideally, you want your thermal camera to have the highest thermal sensitivity – or lowest NETD – as possible. The lower the NETD, the better the sensor can register small temperature differences; whereas the higher the mK, the more likely a camera will deliver inaccurate temperature readings that can lead to missing important scene details. There are other variables that can help ensure you capture as many scene details as possible, including the f-stop number; but for the purposes of this piece, we will focus on NETD.
The below scale, expressed using milliKelvin (mK = 1/1000th of a Kelvin degree), can be used to determine the quality of a thermal detector:
The figures demonstrate two images captured in the rain, with an ambient temperature of ~55 degrees Fahrenheit. In figure 2, the camera with a NETD of 35 mK provides a much crisper image with sharper scene details.
So, what does this mean for you, as you research the right thermal security camera for your application? It is common to find cameras from different manufacturers with a wide range in price. From the snippets being advertised, it will often show that the cameras all have a 640×480 thermal resolution, but the price varies by several thousands of dollars. Why is that? Typically, the reason for the variation in price is due to a difference in NETD value. Usually, the camera with the lower NETD will have a higher price and as a result, it will perform better in challenging environmental conditions.
A thermal camera with a <30 mK is going to be able to distinguish between very small temperature differences in the field of view, compared to a camera with a <60 mK. If both cameras have a resolution of 640×480, then they both have 307,200 available pixels to create an image; but the camera with the 30 mK will produce a significantly crisper image, compared to that with <60 mK.
A thermal camera creates an image by assigning a shade of gray to each temperature change in the field of view. Since the lower mK is able to see smaller changes in temperature, the image will produce more shades of gray, ultimately providing a clearer image. This will generate better performance during inclement weather conditions, along with better detection and classification, when paired with analytics.
Conversely, a camera with a higher NETD value, such as <60 mK, while acceptable, will produce a grainier image, have a lower detection range and most likely trigger more false alarms when paired with analytics. As an additional benefit, lower NETD cameras can detect changes in temperature more quickly and with greater accuracy, leading to increased efficiency in identifying potential problems and managing resources.
For security applications, thermal cameras are usually paired with an analytic. Most manufacturers provide an onboard solution; but some customers elect to use a server-based option. Regardless of whether the analytics are edge or server based, most will be either motion or neural network-based platforms. Both platforms have strengths and weaknesses; however, when choosing the right thermal camera, it is important to understand how thermal sensitivity will impact the performance of the analytics.
In general, motion-based analytic platforms provide object classification initiated by motion. With this system, the analytics must be calibrated to the camera’s field of view to determine target sizes in the foreground and background. When the camera sees motion, it then works to classify the object moving, based on target size and pixel change. The camera will usually generate an alarm if the object in motion is classified as the desired target and breaks the defined rules set.Â
With a motion-based analytic, the camera resolution is most important for initial detection. By having more pixels available, the camera can detect and classify objects at farther distances. However, having a higher thermal camera sensitivity or lower NETD value will provide more stable performance, even in inclement weather conditions. For example, a <30 mK NETD value will maintain a more stable image during rain than a camera with a <50 mK value with the same resolution (refer to Figures 1 and 2). This will allow for more accurate and consistent performance.Â
As an alternative to motion-based analytics, an analytic platform based on neural networks can be closely compared to the way a human brain works. It uses neural networking algorithms to recognize, classify and locate objects in a single image. Neural networks are trained by a deep learning process and require a huge dataset but also a specific processor capable of running the analytics. Instead of classifying an object based on pixel change Âand movement, they analyze an image using a model that was trained on hundreds of thousands of samples to produce a more accurate classification, resulting in a lower rate of false alarms.
When looking for the right thermal camera to use with a neural network-based analytic, your best option will be one with the lowest mK. Especially when image analysis is being done at the software level, details and clarity will be important for optimal performance. The lower NETD value will not only help maintain optimal performance through inclement weather, but it will also help reduce false alarms.
Chances are, if you are looking for a thermal camera, you are looking for a device that will provide something that a visible camera cannot. If you are going to make the investment, don’t sell yourself short by purchasing a product that isn’t going to provide higher accuracy, better image quality, improved object recognition and increased efficiency, even in the toughest weather conditions.Â
Thermal cameras coupled with analytics help to provide a virtual barrier to all types of perimeters. If you are making the investment in the technology, make sure you purchase a camera that will provide results in snow, rain, fog and darkness – not just clear sunny days.Â
Matt Strautman is the Director of Global Business Development for Security at Teledyne FLIR. With over a decade of experience in the security industry, Matt has been on all sides of the business, from being a security integrator, a manufacturers’ rep, to working for the manufacturer. His background has provided him with the knowledge to support customers at all levels on complex solutions.
This article was originally published in the July edition of Security Journal Americas. To read your FREE digital edition, click here.