EEVblog Electronics Community Forum
Products => Thermal Imaging => Topic started by: Davidoo on November 02, 2021, 01:09:58 pm
-
Hi, I'm looking for an infrared camera that can is better at biological measurements. But not overly expensive. I have little knowledge about how these cameras work, but I seem to see a trend that the cameras that have a high temperature measurement range, say -30C to 300C, has higher measurements accuracy from around 1 to 5+-C, while cameras with lower temperature measurement ranges, say 30C-45C often have lower measurements accuracy 0.4 to 1+-C.
Also with regards to the accuracy. Is it such that all the individual pixels will be relative to any one given accuracy fault, so say one pixel is at -2C accuracy, does that mean that all of the pixels will be at -2C accuracy such that the overall contrast/difference between adjacent pixels in the image taken from a biological organism will still be visible. Or can one pixel be -2C off and another +2C, meaning that it could greatly skew the contrast/difference in heat coming from different places on the object. How common is accuracy issues?
Do some cameras come with the ability to set lower temperature measurement ranges and hence decrease the accuracy measurements?
Some of the cameras with wider temperature measurement ranges seem to not display the temperature difference between a window of 3-5C very well because the color changes too slightly; too similar color between the temperature change. Is it common for cameras to be able to change the color scheme such that it mostly colors temperatures with different colors within a user set range within the wider temperature measurement range? Example: camera with -30 to 300C range, the color gradient in the image will change equally within that range, but you set it to only show colors within say 30C to 40C, everything else being black, so that you get the contours of the color gradient within that temperature range, making it easier to see the temperature difference.
Sorry, this is a lot of questions. And I appreciate any feedback. If you have a link to a good, but not too technical overview of how the cameras work (I already found some basic videos, but still confused) that would be great as well. Thank you! Any clarification needed, just ask.
-
I will answer more fully later, when I have time (if someone else does not do so first)
1. The quality of the thermal camera design and parts used has much to do with real world measurement accuracy and repeatability of measurement.
2. The Microbolometer array is biased for the intended use and some cameras have different ranges to cope with different measurement ranges. These ranges apply differing biases to the microbolometer array and this does effect sensitivity and image noise levels. Higher temperature ranges are normally lower sensitivity and have higher noise content.
3. The Microbolometer array produces an analogue signal that is passed to an Analogue to Digital Converter. The ADC bit depth dictates the granularity of data that is passed to the video processor section of a camera. You can commonly have 8, 12 or 14 bit ADC’s in thermal cameras. The ADC bit depth does not change with different ranges and temperature spans.The collected pixel data has the available bit depth but the video processing stages can choose which analogue display units (bits representing image data) to display on the cameras screen or saved to media. The accuracy of the raw data remains constant but the ability to visually interpret differences in the image is effected by the span of ADU’s that are presented to the display system and Colour Palette chosen. The display systems bit depth will also dictate how much data granularity there is in the presented image. This is why it is common to use computer analysis of captured images to pull out details that may not be seen on the cameras display.
4. Calibration of a thermal camera at the factory can be expensive due to labour costs. The cheaper the camera, the more likely that costs were cut at the calibration stages of production. That is just a fact of life.
5.Microbolometer based thermal cameras have temperature sensors that monitor the microbolometer array die temperature, ambient temperature and FFC flag temperature. Some also monitor the lens assembly temperature. These temperature sensors are used by the system to maintain temperature measurement accuracy but the quality of such a temperature compensation system varies from camera to camera and cheaper cameras tend to have less money spent to maintain measurement accuracy over a wide ambient temperature range. The camera ‘self-calibrates’ it’s measurement feature using the FFC flag as a known temperature reference.
That is enough for now :)
Fraser
-
Regarding pixel output temperature accuracy and individual pixel drift issues…….
The microbolometer array contains many pixels and some drift in these individual thermally sensitive resistors is inevitable. To counter the issue of pixel drift, the Flat Field Correction process is used. This correction system places a flag (flat plate) in front of the microbolometer and equalises all pixels to produce a nice flat field with minimal difference between individual pixels. The temperature of the FFC flag is known thanks to a temperature sensor in its vicinity. The camera uses this known temperature to set the measurement offset to account for errors in the measurement system. This means that every time the FFC event occurs, the camera creates a fresh flat field correction and effectively calibrates all its pixels outputs to give the correct temperature (within the limits of its factory calibration).
Thermal imaging cameras that are designed for use imaging biological specimens may be ‘fine tuned’ to produce high accuracy measurements over a limited temperature range. This is achieved through good temperature stability, by design and excellent calibration at the factory. The Temperature stability is important for obvious reasons but the calibration process is enhanced to include many reference points for the calibration table within the temperature range of interest. More calibration points normally mean greater accuracy of a measurement within the calibrated range. Many generic use thermal cameras will use a 2 point calibration to lock the generic response calibration curve in place for ADC count to Temperature conversion. A specialist thermal camera may have many more calibration points in a mere fraction of the generic cameras measurement range. For example….. A Generic thermal camera covering -10C to +120C has 2 calibration points, one at +20C and the other at +100C. A specialist medical camera covering +30 to +50C has a calibration point at every Degree C so has 20 calibration points. The specialist camera system will also incorporate a temperature stabilised microbolometer and carefully calibrated reference for temperature measurements to avoid measurement errors. Such specialist thermal cameras are expensive due to fine engineering plus the time it takes to create the many calibration points in their calibration tables.
A specialist thermal camera still has a number of uncertainties in its measurements however.
1. Accuracy of the cameras calibration at the time of the measurement.
2. Real Accuracy of set Emissivity for the selected target material
3. Humidity of the atmosphere between the camera and target
4. Distance between camera and target
Many high end thermal camera manufacturers state a measurement tolerance of +/-2C, or +/-2%, whichever is greater. Their cameras often provide far greater accuracy of measurement than this, down to points if a Degree, but they recognise the effect of uncertainties and reflect this in their honest specifications.
Where highly accurate temperature measurements are required, it is common to position a thermal reference source within the measurement scene to act as a known temperature reference for measurement quality checks and any corrections needed. This is why the fever detection thermal cameras use such thermal references as it helped to correct errors in the measurement system that would otherwise have made such thermal measurements unusable for the intended purpose.
Hope this helps
Fraser
-
As a side note, some thermal camera manufacturers are adding a warning to their cameras that states that they are NOT suitable for fever detection. This is because some people and organisations mistakenly thought that detecting fever in a human using any old thermal camera is easy. It is not.
Fraser
-
A little comment on automatic vs manual temperature span modes on thermal cameras…….
A thermal camera operating in fully automatic image adjustment mode will normally identify the highest and lowest temperatures in a scene and set the cameras temperature span to these maximum and minimum limits. This is great if you want to see every temperature that appears in the scene but it also means that the colour palette is applied to the whole temperature span so differentiation between low Delta T areas of the scene may become challenging to see on the cameras display.
If a thermal camera offers a manual mode for image adjustment, the user can normally set the minimum and maximum temperatures that are shown on the display. If the camera has its displayed temperature span reduced and the full colour palette is applied to that reduced span, it can be easier to differentiate low Delta T areas of the displayed scene.
It is also worth experimenting with different colour palettes as some are better at showing small Delta T areas of a scene than others. There are high contrast palettes like Rainbow HC that are intended to give a high contrast image to assist in low Delta T scene work.
A colour palette will have a number of shades or colour levels that it applies to the scene. Avoid palettes that have reduced level counts and choose those with the maximum available number of shades or colour levels. I personally work in monochrome and Iron palettes most of the time and use a Rainbow HC palette in specific scenarios.
Fraser
-
Do some cameras come with the ability to set lower temperature measurement ranges and hence decrease the accuracy measurements?
You can change the range, but it will have ne effect on accuracy. You would need to actually change the detector setup and do a full factory level recalibration. Even then while you would reduce image noise (sampling of pixel to pixel, pixel over time) the accuracy is mainly down to the factory calibration accuracy, flag temperature accuracy and lens temperature effects.
Some of the cameras with wider temperature measurement ranges seem to not display the temperature difference between a window of 3-5C very well because the color changes too slightly;
If closing up the range too much you are likely to enhance noise, and visually chroma noise is a lot more objectionable than luminance noise (eg a greyscale image). Having a reduced noise camera (by reduced range or better parts) will help here.
As a quick example from versions of fire camera (same sensor) with several ranges built in...
Range 1 (to 100°C) MDTD = 40mK
Range 1 (to 150°C) MDTD = 60mK
Range 2 (to 450°C) MDTD = 100mK
Range 3 (to 1000°C) MDTD = 300mK
MDTD reading is 'can I see a large object at this temperature difference' so includes eye-brain effects. As a single shot frame standard deviation these would be a bit worse.
The accuracy at 25°C is the same at the instant of flag calibration, but range 3 would drift more than the range 1's up to the next flag operation.