DEPENDENCIES

1. CMake >= 2.8.3
2. OpenCV >= 3.0
3. ROS/tf

STRUCTURE ORGANIZATION

• libcbdetect: C++ implmentation for chessboard detection from this repo
• align_thermal_to_rgb: align thermal image to RGB image (in this case, Intel RealSense D435) with the help of chessboard
  • include
    • cali
      • align_thermal_to_rgb.h: class definition for align_thermal_to_rgb
  • src
    • helper.hpp: some helper functions
    • align_thermal_to_rgb.cpp: class implementation for align_thermal_to_rgb
    • get_extrinsic.cpp: intrinsic/ extrinsic calibration for thermal camera, user provideds intrinsic of rgb and thermal cameras, rgb, depth and thermal images, and an interger for mean filter kernel size, then it will generate a file with thermal camera intrinsic, distortion coefficients and rgb-to-thermal extrinsic, named in camera_model.txt
    • example.cpp: align_thermal_to_rgb usage example
  • bin
    • get_extrinsic -> src/get_extrinsic.cpp: get thermal intrinsic and rgb-to-thermal extrinsic parameters with images with chessboard inside
      • ./get_extrinsic [path_to_rgb_intrisic_file] [path_to_thermal_intrinsic_file] [path_to_rgb_image] [path_to_depth_image] [path_to_thermal_image] [kernel_size]
    • example -> src/example.cpp: test calibrated camera model from get_extrinsic result
      • ./example [path_to_camera_model_file] [path_to_rgb_intrisic_file] [path_to_rgb_image] [path_to_depth_image] [path_to_thermal_image]
  • lib
    • libalign_thermal_to_rgb.so: library for align_thermal_to_rgb, link this library in your own source codes
  • result: some concrete alignment result
  • example_input: example input for get_extrinsic execution
  • CMakeLists.txt

BUILD

Since the align_thermal_to_rgb should do chessboard detection, so we have to build the libcbdetect first

1. cd && git clone https://github.com/sean85914/cali_thermal.git
2. cd cali_thermal/libcbdetect
3. mkdir devel && cd devel && cmake ../ && make
   (Note: If you want to show more detail information, use cmake ../ -DCMAKE_BUILD_TYPE=Debug then make) then we can build the align_thermal_to_rgb
4. cd ~/cali_thermal/align_thermal_to_rgb
5. mkdir devel && cd devel && cmake ../ && make && cd ../bin
6. Then, if you have data for testing (i.e., chessboard in the image), run
   ./get_extrinsic [path_to_rgb_intrisic_file] [path_to_thermal_intrinsic_file] [path_to_rgb_image] [path_to_depth_image] [path_to_thermal_image] 7
   Then you will get the output file camera_model.txt, an aligned thermal image and a combined one
7. With this file, you can use align_thermal_to_rgb class, for example, ./example [path_to_output_txt] [path_to_rgb_intrinsic] [path_to_rgb_image] [path_to_depth_image] [path_to_thermal_image]

RESULT

Sample output:
intrinsic: [606.526, 625.874, 283.076, 247.947]

distortion: [0.144775, -32.6391, 0.0255662, 0.0092998, 274.744]

quaternion: [-0.0151522, 0.0415791, -0.0190183, 0.998839]

origin: [-0.0579184, 0.0172561, 0.119621]