Espartacus,
A fair point. Image quality is only part of the story when dealing with a fully radiometric thermal camera.
When it comes to sales, many buyers will be swayed by “pretty pictures” that contain good detail and low noise. Clever video processing techniques do make thermal images look much cleaner, but sometimes the images are “over-processed” which can cause a loss of fine detail or mask small anomalies. It is a delicate balance for a manufacturer to create a thermal camera that produces “pretty pictures” whilst also delivering the required image detail and measurement accuracy. Thankfully a lot of the clever image processing is applied after the stage in the system where pixel measurements occur.
The Infiray Tiny1 and Guide TIMO 256 thermal imaging “building block” cores have proven to be capable of delivering both decent thermal imagery and very good measurement accuracy. As these cores, or their siblings, are to be found in many 256x192 pixel thermal cameras, measurement accuracy is not normally an issue. Some USB Thermal Dongle cameras have been designed for imaging performance, rather than temperature measurement and these either provide less accurate measurements, or do not include any measurement capability in their APP. The common “+/-2 Celsius or +/-2%” specification is normally met by the vast majority of modern thermal cameras claiming such in their specification. With “building block” type imaging cores, this is easier for OEM’s to achieve as much of the hard work has been done by the experienced and skilled Thermal Imaging Core manufacturer. The OEM just needs to provide an embedded computer/SoC host to display the cores data and any desired camera functionality and overlays.
Experienced thermal camera manufacturers may take a different path when designing a thermal camera as they have the knowledge and skills to integrate a bare microbolometer into a System. Such a ‘ground-up’ design contains all the required and fine-tuned stages of image processing and measurement calibration. With an experienced camera manufacturer this can lead to a high performance product that is very accurate and performs very well indeed. It is the harder development path to take though and there is the risk of issues with a new design.
The Thermal Master THOR camera has been mentioned. This is a case that challenged both me and Thermal Master in terms of discovering what was going on inside the cameras measurement system. When a manufacturer decides to try a new approach to Radiometric thermal imaging, issues can arise and need to be addressed by the design team. I have never been told what caused the unusual measurement behaviour in the original THOR cameras, but Thermal Master appear confident that they have now addressed the issues with new firmware, and even released a new platform revision. I believe that Thermal Master were trying to improve Radiometric accuracy using ‘Smart’ algorithms, and something went wrong. We may never know the details of why the THOR cameras suffered from measurement accuracy issues.
Thankfully it would be uncommon to find a decent modern fully Radiometric thermal imaging camera with poor temperature measurement accuracy these days. There really is no excuse for such in a general use Radiometric thermal camera. Some specialist application cameras prioritise imaging performance, and high frame rate, over temperature measurement accuracy. These normally reflect this in their specifications or lack the measurement function to avoid misleading the user. Wildlife spotting scopes and some USB dongles are examples of such systems.
I have seen less than stellar measurement accuracy and stability in very cheap thermal cameras that are built down to a price, but such is almost to be expected as these devices use the cheapest available components and minimalist designs. The manufacturer should reflect this in the specifications. Some do and some do not. I tested a prototype camera that used the SEEK THERMAL QVGA Mosaic imaging core and was shocked to see measurement errors of around 10 Celsius and very poor measurement stability. The warm-up time was also excessive. It transpired that the designer of the electronics package had never worked with SEEK THERMAL imaging cores before and was having difficulty understanding the English Language core integration documentation. Mistakes were made but I highlighted them before the camera was released and they were addressed and corrected by a new, more experienced, design engineer. There is always a chance that a designer will make a mistake or a production issue occurs, but thankfully such appears rare these days.
It is my experience that consumer grade Radiometric thermal imaging cameras normally meet the measurement accuracy specification stated in their specifications. Some claim the common “+/-2C or +/-2%” accuracy, whilst others state “+/-3C or +/-3%” as they are being very honest about what you are getting for your money. I have seen some cameras stating a measurement accuracy of “+/-5C or +/-5%” but there is normally a good reason for this. Either they are “bargain basement” models, or they are specialist cameras intended for mainly non Radiometric tasks, such as wildlife observation or SAR duties.
The better quality Radiometric thermal cameras still state an accuracy of “+/-2C or +/-2%” in their specifications, but I regularly see them providing far better accuracy than this specification. They are professional designs that acknowledge the “Industry standard” measurement accuracy figures, but they are designed to provide significantly better measurement accuracy, stability and repeatability. That is what you pay the higher price for when buying an industrial fully Radiometric thermal imaging camera. My FLIR E40 (upgraded to E60) normally provides measurements within +/-0.3C of that set on a calibrated Blackbody Thermal Reference. That E40 was built in 2013 and has not been calibrated since. You get what you pay for with most Industrial thermal imaging cameras. The question is….do people actually need such excellent accuracy for many generic tasks where other, non camera related, factors will potentially adversely affect the measurement accuracy anyway ? An example would be guessing the Emissivity value of the target surface. There is much opportunity for measurement errors to creep in due to an incorrect emissivity setting.
Fraser
Home
Help
Search
Login
Register
