I agree with the comments made. A conventional NIR sensitive sensor element makes for a poor thermal imaging device, especially when uncooled. The optics of the system would also need to be carefully crafted to provide best transmission. It is true that a target at a temperature of around 400C does emit enough energy to be seen on a sensitive NIR camera. Why can it see it ? Well if you take the temperature of the target higher it will start to radiate in the human eye sensitivity range and appear dull red, followed by bright red, then yellow and eventually almost white, if hot enough. The hotter the target the closer the radiated energy moves down the spectrum towards NIR and visible light wavelengths until it may be seen by the camera sensor.
For medical use it is normal to require a decent sensitivity around the temperature of a living human so approx 36C. Such a sensitivity is easily achievable with a microbolometer and so this has become the standard sensor of choice, or a thermopile sensor for non contact IR thermometers.
The microbolometer is now the most affordable FPA for thermal imaging thanks to the recent developments in the industry. Sadly a CCD or CMOS visible light/NIR sensor just does not cut the mustard for imaging at 36C, even if cooled.
A possibility that I have considered looking into is thermally reactive liquid crystal films. These change colour in response to heat falling upon their surface. A ZnSe lens can be used to focus thermal energy from a target onto the thermally reactive film to create a crude thermal image. The film will respond to the heat of a hand so theoretically could be used to image a 36C target. Ambient temperature would effect the film however so maybe it would have to be mounted in a Peltier element cooled chamber ? The image produced on the thermally reactive film could be imaged using a common visible light camera via a lens system or even, theoretically via direct contact imaging. Direct contact imaging would usually require a large sized sensor plus a fibre optic light channel plate to reduce light contamination across pixels. I would use a simple lens system adjust ed for very close focus.
Such a system might be relatively cheap to produce but sadly it is very crude and is unlikely to produce very detailed images or calibrated temperature readings. The microbolometer leaves it dead in the water, hence why no one has developed such a product commercially.
If you wish to experiment I think Edmunds Optics USA can supply thermally reactive liquid crystal fluid that may be used to coat a suitable thin plastic film. The non LC coated side of the film is sprayed matt black for better responsiveness to the image projected from the lens.
Fraser