Thermal camera

[Mohabon]

Hello,

i have to purchase a thermal camera to install on UAV for mapping. I am thinking to purchase a FLIR A35 camera ( see link http://www.imc-store.com.au/FLIR_A35_p/63209-0101.htm ), which is said to work quite ok on UAV.

2 Questions:

1) Does anyone have some feedback or reccomendations with this camera?

2) Is it possible to process these cameras pictures with photoscan PRO? Was this processing tested by any user in the forum and what is the feedback? I am concerned because automatic point matching in PS Pro could be really difficutl with thermal pictures, so any idea is welcome.

Regards,

- M -

[RalfH]

As the spatial resolution of the camera is quite low, I would suggest pairing it with a regular digital camera on the UAV, use those images for 3D model generation and camera positioning, and then bring in the TIR images with the now known camera coordinates. Also, I would ask the TIR camera manufacturer regarding technical details relevant for deploying it on a moving platform (exposure times, rolling shutter?) before buying.

[George]

Hi
interesting topic indeed.
Ralf you offer to capture simultaneously RGB and thermo images and then just substitute rgb with thermal (after the RGB aligned). Provided that the thermal and RGB were oriented exactly the same point.
Alexey, did you have an experience with thermal images?
George

[RalfH]

George,

yes, ideally the optical and the thermal cameras should point exactly in the same direction. As this will be very difficult to achieve, an alternative may be to measure the differences in orientation of the two cameras. If they are both securely mounted on a common carrier structure, it should theoretically be possible to use those measurements to calculate the orientation angles of the thermal camera from those of the optical camera. Would be a quite advanced project though.

[George]

yeah! but still the problem of post-alignment processing persists ...
Would actually Agisoft intepret thermal images correctly?
That isthe question, I think.

[Alexey Pasumansky]

Hello George,

The problems with thermal cameras are caused mostly by very low image resolution and slow rolling shutter speed.

[George]

Alexey, could you specify please the issue of the rolling shutter?
Seems like the problems appears more often exactly with different spectral cameras.

[RalfH]

The "rolling shutter" problem with many video, cell phone and thermal cameras is due to the fact that the sensor is read line by line (they usually don't even have a mechanical shutter). If that happens and the camera moves during exposure, you get image "smear" effects and distortions. The camera model of PhotoScan can not correct for such non-radial distortions. You should ask the camera manufacturer how fast the camera can capture a full image.

[Alexey Pasumansky]

Hello George,

For example, TetraCam ADC Lite rolling shutter speed is about 130 mSec. While their Mini-MCA camera has several times shorter shutter speed.

[RalfH]

Hello Alexey,

does that time refer to capuring a whole image or to a single sensor line? I assume whole image?

[Alexey Pasumansky]

Hello RalfH,

Yes, that's time for the whole image. For ADC camera the resolution is 2048x1536.

[RalfH]

OK, that's nothing that could work on a fast-moving platform, and for a slowly moving airborne platform you'd need a good gyro stabilisation to keep the camera pointed in the same direction.

[George]

Does that simply mean that Pscan a bit limited when post-processing the images captured by the cameras of that type?
You guys often refer to some kind of distorsions caused by the shutter speed. However, that issue is not clear to me. What are these cameras then meant for I wonder if they deliver poor imagery/data (distorted)? Or what are the limitations of their use?
Could you also clarify please what should be then the shutter speed (range in Sec or mSec) or what kind of shutter should camera posess to be able to post-process at Pscan?

[George]

OK, that's nothing that could work on a fast-moving platform, and for a slowly moving airborne platform you'd need a good gyro stabilisation to keep the camera pointed in the same direction.

However, Pscan is able to process a bit oblique images from cameras of some types perfectly orthorectifying them ...

[RalfH]

George,

photographs are affected by different types of blurring and distortions. Blurring happens, when there is relative movement between camera and object while the picture is taken. Camera lenses are never perfect, but most distortions caused by lenses are radial distortions, simply put, the image is increasingly distorted with increasing distance from the image centre. PhotoScan models radial distortions very well. A rolling shutter scenario causes non-radial distortions: because the sensor is read line by line, this can be a linear stretching/compression of the image (if relative camera movmenent is orthogonal to the sensor lines), shearing/smearing of the image causing a parallelogram-type distortion of the image (if relative movement is parallel to the sensor line) or a combination of both. PhotoScan (or any comparable software) can not deal with this type of distortion. The way this can be dealt with is to either record all movement (in particular rotations) of the camera using a IMU and correct the rolling shutter problem in post-processing, or to use a gyro-stabilised camera rig which makes sure that rotation of the camera about any axis is so slow that the rolling shutter problems do not occur. The same goes for airborne line-scanner sensors. A simple gyro stabilising system can be built for a few hundred Euros (3 axes gyro, some electronics and servos).

With an optical camera on a slowly or moderately moving airborne platform, exposure times between 1/1000 and 1/4000 should usually be OK (unless you have strong vibrations). Movement as such is usually not the biggest problem, but rotation.