EEVblog Electronics Community Forum
Products => Thermal Imaging => Topic started by: tonyget on May 28, 2022, 04:06:11 am
-
I haven't read the report since it's not cheap. But according their conclusion，the InfiRay bolometric sensor per se is quite decent，however the image processing algorithm still leaves much to be desired.
https://piseo.fr/en/press-news/news-en/infiray-micro-iii-camera-core-performance-analysis/
PISÉO’s finding’s about InfiRay Micro III Camera
The Micro III camera module as tested by PISEO contains a strongly performing bolometric sensor, producing good quality images with low numbers of bad pixels. Unfortunately, the image quality is not enhanced well by the image processing firmware, which contains basic tone mapping functions but with errors on some of them. From a user’s perspective the software platform is not supportive enough to get the best out of the camera’s potential.
Compared to the Boson camera module from FLIR, which is a reference product in this field, the Micro III module is based on a higher quality sensor which should provide good image quality in many user cases. However, InfiRay’s algorithms for signal correction and image processing, which make the Boson camera’s approach interesting, largely underperform. The system engineer may be surprised by how difficult it is to reach the sensor’s potential.
The impression one gets from the analysis of the Micro III is that this module is better for internal InfiRay applications than for external ones. The multiple user interfaces and the approximation of the software tools suggests case-to-case improvements inspired by specific application contexts.
(https://i0.wp.com/piseo.fr/wp-content/uploads/2022/05/IRay-Camera-Micro-III-Performance-Analysis-2022-illu1.jpg)
-
An interesting report :-+
Historically the capture of a thermal scene by hardware was only part of the story. It was the performance of the image processing software that revealed how experienced an OEM was in the thermal imaging field. FLIR introduced the PM570 microbolometer based thermal imaging camera in 1997 and they did a decent job of the hardware but agreed that they were struggling to tame the noise present in the 1st Generation uncooled sensor FPA. They were used to using low noise cooled FPA’s prior to the PM570. Since the release of that camera, FLIR spent a lot of time and effort improving the signal to noise ratio of their microbolometer FPA’s and enhancing their image processing algorithms. The result was some decent uncooled thermal imagery. It has taken many years, and the acquisition of several specialist companies, to achieve FLIR’s current level of uncooled thermal imaging performance. With this in mind, I can forgive InfiRay for not creating the ‘perfect’ image processing software in these early days of Chinese decent microbolometer production.
It is not uncommon for manufacturers of consumer grade thermal imaging cameras to ‘tune’ their image processing software for the creation of what I call ‘pretty pictures’ at the expense of some detail or accuracy in the resulting image. This is an understandable approach as many consumers will judge a thermal camera on image noise content and how much it looks like a standard photograph. Thermographers may prefer a ‘more honest’ camera that produces highly accurate thermal scene information ‘warts and all’ For them, it is about the thermal data held in the image and not how pretty it is. They will not be putting it on a wall as a poster or displaying at an art exhibition. That said, over the years, the advances in microbolometer technology and image processing mean that a thermographers professional camera can produce beautiful, low noise, imagery whilst maintaining the accuracy of the data contained in the pixel values.
I can see that OEM’s targeting the consumer market will want their thermal cameras to stand out from the crowd when it comes to the images that they produce. We have previously seen naughty practices where higher resolution thermal images from FLIR cameras have been shown associated with budget cameras coming out of Asia to mislead the buyer. The budget cameras that we commonly see detailed on this forum have shown how much the technology has advanced in recent years, and the consumer grade camera sector is benefiting from this. The OEM’s can be prone to ‘over processing’ the images however. Image processing Experts on this forum quickly identify the heavy handed use of pixel interpolation, smoothing, edge enhancement and poor image equalisation in the produced images. The pictures look good to the untrained eye however and that is what gets the sales.
In the referenced report, I was interested to see the FLIR Boson being used as the ‘Reference’ for imagery. That is a totally understandable decision by PISEO but the Boson is an expensive underperforming core and a good example of how even the experts can make mistakes. For me, the FLIR TAU2 or one of the DRS 12um based imaging cores would be a decent reference point for good quality budget image production. Both of those are expensive cores, but then so is the Boson.
It would be interesting to see the whole PISEO report to determine exactly what characteristics of the thermal imagery are important to the test team. Is the testing biased towards fine detail, measurement accuracy, MRTD, or just pretty pictures, as might be the case in a photography magazine review ? In another PISEO thermal core test, they comment on the lack of a standard in the industry for testing and review of thermal imaging cores. They look to be trying very hard to create an accurate unbiased professional review of imaging cores and for that I applaud them. The test equipment for such testing is very expensive so I can understand that the reports produced are similarly expensive.
https://piseo.fr/en/press-news/news-en/seek-thermal-mosaic-core-s309sp-camera-analysis/
Fraser
-
The objectives of PISEO……
“ Currently, no international performance testing standards have been published. This situation leaves purchasers of such systems with supplied datasheet values that may not always be right or may only provide partial information. This is where Piséo fills the gap, with its independent evaluations and testing of thermal cameras.”
And
“ Establishing an international performance testing standard
In order to characterize the performance of the camera core, Piséo’s experts use their own test protocols, software, and lab equipment, which includes a calibrated HGH black body and climate chamber able to perform tests at ambient temperatures ranging from -20°C (-4°F) up to 100°C (212°F).
There is currently no international standard protocol for thermal camera performance testing available.
After performing our own characterization tests on many devices, analyzing manufacturers’ datasheets and discussing the reasons for the discrepancies between our test results and their published values, it became clear to us that the performance values published were all based on different test protocols. It is, therefore, quite impossible for the purchaser to use the datasheets to select a thermal camera or core based on comparable performance values between products. Therefore, we aim to publish such data based on a fully transparent and consistent testing protocol whatever the product tested.
By doing so, we provide the market with independent, unbiased test results that de facto establish an international performance testing standard for thermal cameras, cores, and sensors. This report contains for the first time a table highlighting the comparison of the main performance indicators between products. using Piséo’s own testing protocol. Manufacturers, integrators and end-users, if you also want to get your thermal camera tested, feel free to drop us a message!”
-
My above comments are well demonstrated by the PISEO sample image included in the OP’s post. If you look at the image from the Infiray core and compare it to that of the Boson….. which do you think your average consumer will prefer to buy ? I would suspect that the Infiray core would be chosen as it has greater noise reduction that makes the Boson appear a very noisy image and less ‘pretty’. The fact that the Infiray image processing is excessive and causes issues with the presented scene, from a purists point of view, will be lost on many end users of the core.
Fraser
-
I think InfiRay should combine bolometric sensor with AI chip，like what smartphone manufactures are doing with image CMOS sensor nowadays，so the image captured by the sensor can be processed by dedicated AI algorithm in realtime.
-
What would you hope to achieve by applying computational photography to a low resolution microbolometer based system ? Standard DSP can cope with processing the image data into a good thermal image. Smartphones etc tend to use computational photography to create effects or provide enhanced ezoom performance to emulate true optical zoom. Adding a computational photography image processor to a thermal camera system would be in line with the current fashion to include “AI” to anything that processes image data, but such could be a challenge to implement correctly for thermal imaging OEM’s not experienced with such.
-
The NISt work on thermal imagers for fire services may also be of interest...
https://www.nist.gov/el/fire-research-division-73300/firegov-fire-service/thermal-imaging (https://www.nist.gov/el/fire-research-division-73300/firegov-fire-service/thermal-imaging)
While it did result in some measurements for the NFPA standard, some of the ideas were not that practical / useful, or were too easily 'defeated'.
Bill
-
This Nist document is looks interesting …….
https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=902977
Fraser
-
A meeting of minds to discuss thermal imaging in the Fire Service……
https://nvlpubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication1040.pdf
Fraser
-
Another interesting paper on testing thermal imaging systems…..
https://www.govinfo.gov/content/pkg/GOVPUB-C13-9ba4abeded8fdb465c8ac189377ef0ce/pdf/GOVPUB-C13-9ba4abeded8fdb465c8ac189377ef0ce.pdf (https://www.govinfo.gov/content/pkg/GOVPUB-C13-9ba4abeded8fdb465c8ac189377ef0ce/pdf/GOVPUB-C13-9ba4abeded8fdb465c8ac189377ef0ce.pdf)
-
(https://s3.yolegroup.com/uploads/2022/07/IMG_FOUR_THERMAL_CAMERA_PERFORMANCE_ANALYSIS_ComparisonAssessment_PISEO_July2022-2048x1323.jpg)