We use OpenCV to read images and send individual points. Compiling using cmake and running ./test <path/to/image> should result in the image being displayed in an OpenCV window. This will confirm that OpenCV has been installed correctly.

extract_transforms.py will extract the transforms of the camera that was used to collect the point clouds and output it in a form easy to read from C++. This is to make it convenient to transform point coordinates into the global frame so that they can be temporally aggregated.

This module reads in a list of times images were collected, and searches the log of robot states to find the latest state that occurred before each image collection time. It then uses forward kinematics to compute the transform of the camera in the world frame at this time.

The output file has 12 columns of space-separated numbers. The first 9 columns are a flattened (column-major) rotation matrix, the last 3 are the position of the camera. Row i corresponds to image i.

The intrinsics of the camera are stored in intrinsics.txt as space-separated numbers in the format fx fy cx cy.