John Wolcott: Thanks for your comments – I appreciate your informed perspective. I do understand (in broad outline) what Pix4Dmapper does in the default case. But the sophistication of that algorithm obviously requires highest quality inputs and careful attention to software parameters, otherwise it fails completely and ungracefully. So what I don’t understand is why the software doesn’t include a more robust fallback algorithm that will still successfully generate a usable (if not full-featured) product when overlap, contrast, and/or resolution are less than ideal, as they sometimes unavoidably are. As I think I mentioned elsewhere, the human eye could match up my thermal images well enough to generate a usable mosaic, given enough time and patience and knowing in advance the spatial arrangement of the images.
Like Reto, I also don’t understand why the software requires the thermal and RGB images to be processed separately rather than synergistically. If I were writing the algorithm (and I may have to at some level; see below), I would use the camera positions/orientations and elevation model derived from the RGB images to inform the mapping of the corresponding thermal images, presumably reducing the need for feature matching in the thermal image set. Even minor time synchronization issues mentioned elsewhere shouldn’t necessarily be complete showstoppers.
At the moment, I’m stuck with 360+ images per flight from a $7500 camera, taken from a several-day trip to another state, that look “good enough” when viewed individually – and have reasonable geolocation and orientation info – but for which I have been unable to find a software tool that will allow me to reliably produce even a crude thermal mosaic of a rectangular, relatively flat 40-hectare field (but I have admittedly not yet tried the settings recommended by Momtanu).
Greatly compounding my frustration is the fact that FLIR’s R_JPEG format is proprietary, so other software tools I have looked at don’t know how to utilize the files. I had not planned to learn to program C# in a dot-net/Windows environment so as to be able to use FLIR’s Atlas SDK, but right now that seems like my most promising option for doing the intended science with these images.
My impression is that this thread about the FLIR Duo Pro R camera is one of the longer threads in the forum. There’s clearly interest in, and high expectations for, the RGB/thermal combination and there’s clearly disappointment among at least a few users that it doesn’t “just work” with Pix4Dmapper out of the box – no messing with camera internal parameters, etc. I’m not sure how many of these units FLIR expects to sell once word gets out that there’s poor end-user software support for this unit, especially once you get away from the most common use cases.
You hit the main issue. The reason pix4d fails is it can’t calibrate camera’s (image orientations). It needs big, clear images to obtain exact parameters. Usually what fails is the focal length. That’s why with thermal a lot you see it calibrate all the cameras in the calibrate in the middle. In general terms thermal images are very “blurry” when looked at with photogrammetry in mind.
What I think you are asking (and I have as well with no luck) is to get the developers to produce a pure stitched (panoramic) image without photogrammetry to generate a point cloud for those that only need an ortho and not and 3d outputs. It’s obviously a small market for them and not cost effective.
Our best results with the xt2/duo pro has been to use 1 or 1/2 sized images and all prior. Sometime set calibration to (none) for internal but even then that doesn’t always work. Pick 4 “passes” and just run over and over with various settings to find the one that is the closest.
If you want to put 4 “passes” up on Dropbox or something I will look and see what I can do.
Followup: I have now finally had the opportunity to independently (outside of Pix4Dmapper) open and view the radiometric TIFF images and associated EXIF data, and I think I have found some possible additional reasons for the software’s inability to deal with our thermal imagery from the Duo Pro R:
There is pretty significant vignetting, with radiometric temperatures falling off by up to 2.5 degrees, on average, from the center to the lateral edges of the images. This was viewing a uniform flat target in a small windowless room at room temperature, so actual thermal variations of this magnitude are ruled out. Similar vignetting is also visible in the drone-acquired images, though somewhat harder to unambiguously separate from viewing angle effects. For some reason, the fall-off is not symmetric – it’s significantly worse on the right edge than the left. The pixel range of reasonably flat response seems to be between about 110 and 450 (out of 640, for a 45-degree viewing camera). I can imagine that when the thermal field is otherwise low in contrast, this vignetting confuses the matching algorithm in Pix4Dmapper, but I could be wrong.
The camera orientation values (yaw, pitch, roll) stored in the EXIF metadata seem pretty poor. I took several images on a tripod with the camera facing the horizon due north, east, south, and west (uncertainty < 5 degrees). The yaw values recorded for these images was typically about 30 degrees off (approximately 330, 60, 150, 240, respectively). This is at a location where magnetic declination is very small (< 2 degrees). In short, if Pix4Dmapper is trying to use the EXIF orientation data when matching images, it’s probably doing more harm than good.
@Grant, Pix4D does not use the orientation data if you are not using the accurate geolocation and orientation option in the calibration method. When using an RTK drone, we suggest using that option. It increases the processing speed. But for normal cases, when using standard or alternative calibration option, Pix4D calculates the orientation values from the images and uses that for processing. Pix4Dmapper calculates the optimized orientation during Step 1. Initial Processing.
You are right about the vignetting effect. Nice analysis!
Those who have or are considering the FLIR Duo Pro R might be interested in my experience with it: https://hail.aos.wisc.edu/wp/?p=2373 I originally posted this in December but didn’t make it public until now, because I wanted to give FLIR a chance to correct any misleading information (they didn’t respond to my message inviting them to do so.)
TL;DR version: If you want calibrated radiometric temperatures, watch out. If you want Pix4Dmapper to stitch together a thermal map for you, watch out.
Hi Grant,
I tried clicking your link, but it wasn’t active. I would be very interested in your experience with the flir. We bought one recently and it’s not friendly. Is there still a way to read about your experiences?
Thank you and stay safe.
These cookies are necessary for the website to function and cannot be switched off in our systems.
They are usually only set in response to actions made by you which amount to a request for services, such as setting your privacy preferences,
logging in, or filling in forms. These cookies do not store any personally identifiable information.
These cookies allow us to count visits and traffic sources so we can measure and improve the performance of our site.
They help us to know which pages are the most and least popular and see how visitors move around the site.
All information these cookies collect is aggregated and therefore anonymous.
If you do not allow these cookies we will not know when you have visited our site, and will not be able to monitor its performance.
These cookies may be set through our site by our advertising partner (Google).
They may be used by Google to build a profile of your interests and show you relevant adverts on other sites.
They do not directly store personal information but are based on uniquely identifying your browser and internet device.
If you do not allow these cookies, you will experience less targeted advertising.