Closed stephen9999999 closed 1 year ago
hi @stephen9999999 ,
The black lines on the image represent the windows where a persistent landmark should be found. In other words, when you set the number of tiles in the parameters file, then the X library tries to distribute the landmarks more uniformly over the windows.
Regarding the bright images, you can play with the params epsilon_gap
and epsilon_base
in the yaml file to have a darker image. Refer to the paper Online Photometric Calibration of Automatic Gain Thermal Infrared Cameras
for more details.
Best, Vincenzo
Hi, Thanks for your suggestions. I have read the paper and tried to adjust the params but still cannot reproduce the results in the paper. With the PHOTOMETRIC_CALI ON, the contrast of images is lower than the raw images, and the noise is amplified. Therefore, the tracked feature number is less than the original images. Anything wrong? Looking forward to your suggestions. Thanks .
Hi,
This #5 should help. It can bring data race problems, but we tested with this piece of code that was missing in this release. Please let me know if it works for you.
Best, Vincenzo
Hi @stephen9999999, I will close this issue since there is not much activity here, and it should be solved. Please feel free to reopen it if you need to. Thanks. Best, Vincenzo
Hi, I turned the VERBOSE option to ON in the CMakeLists.txt, then the rviz can receive the trajectory topic and tracked feature topic. But, as shown in the following figure, the calibrated thermal images are much brighter than the result in the paper, how can I get the enhanced thermal images as shown in the paper? By the way, what's the black line in the calibrated images?
The calibrated image:![2022-11-09 15-02-42 的屏幕截图](https://user-images.githubusercontent.com/105863329/200772387-322fdbd8-6d2e-4184-bf2f-ef038c368908.png)
The final estimated trajectory is not bad:![2022-11-09 15-06-17 的屏幕截图](https://user-images.githubusercontent.com/105863329/200772627-8051763a-9c83-44d9-b0cc-24b64a67f751.png)