Project for loading the raw data, including images, transforms and camera parameters.

The file helper_functions.h contains the following methods:

• parseDataFolder - returns 4 vectors containing the 3x RGB images and 1x Depth images. The first RGB vector corresponds to Xtion images, the 2nd one to XS1 images and the 3rd to XS2 images.
• visualize_data - debug method for viewing the images in opencv
• initializeCamParamsFromFiles - loads the camera parameters for the 3 cameras. Each set of camera parameters is a double[6] array containing (k1 and k2 are the 2nd and 4th order radial distortion coefficients):
  • fx
  • fy
  • cx
  • cy
  • k1
  • k2
• loadPosesFromFile - returns a vector of Eigen::Matrix4f transforms, 1st one from Xtion -> XS1 and the 2nd one from Xtion -> XS2.
• createCloudFromImages - creates a point cloud from an RGB image, a depth image and camera parameters.

This is compiled as a lib called xs_data_reader.

NOTE XS1 is on the left, Xtion is in the middle and XS2 is on the right (in terms of the actual physical layout).

The executable xs_data_reader_main is meant to serve as an example on how to use the xs_data_reader library for loading the raw data. It also contains the method testTransformsROS which publishes the loaded camera poses so that they can be viewed in TF (with e.g. rviz).

The executable xs_data_publisher loads the data and publishes it in ros. It can be used for debugging how well the camera transforms and paramters fit, by visualizing in rviz the projected Xtion point cloud in the two camera images. The topics published are:

• /xs1/image and /xs1/camera_info
• /xs2/image and /xs2/camera_info
• /xtion_cloud

By subscribing to /xtion_cloud and adding an object of type Camera in rviz subscribed to one of the xs image topics, you can visualize the projection of the point cloud in the camera image.