This package provides a simple way of processing and extracting camera images from a stored bag file. Many bags will store only the raw_image and camera_info, either because the creators want to avoid the overhead of running image_proc or to minimize the amount of redundant information stored in the bag. bag_to_images runs on a bag file, performs the same processing as an image_proc node, and saves the resulting images to disk as separate files. One of the main advantages of bag_to_images is that unlike setting up a full ROS graph to do the same processing there is no worry of dropped images.
Basic usage (dumps colorized rectified images from left wide stereo camera):
bag_to_images --bag [filename].bag
To see the various options available:
Dumping .png files:
bag_to_images --bag [filename].bag --extension png
To dump (rectified) images from the right narrow camera:
bag_to_images --bag [filename].bag --camera /narrow_stereo/right --color false
This package depends on the rosrecord C++ API, which is likely to change at some point. It also depends on the unreleased C++ API for image_proc, which is relatively stable.
Handling of camera_info
The camera model used for rectification is updated whenever a new camera_info message is encountered in the bag file. Until at least one such message has been encountered images will be dropped.
Possible future extensions
- A progress indicator would be useful and can be added when the rosrecord API provides some feedback about progress.
- Stereo processing might be useful, and options could be added to process and produce point cloud files and/or disparity maps