Back to Projects List
Registration of Intel RealSense depth camera image to CT + marker based tracking
Key Investigators
- Colton Barr (Queen’s University, Canada)
- Andras Lasso (Queen’s University, Canada)
- Steve Pieper (Isomics)
Project Description
The goal of this project is to develop a module in Slicer for registering [Intel RealSense][realsense] depth data to a CT scan of a patient’s
head and maintain this
tion using an optical marker. The module will use facial surface anatomy to perform initial
registration between the RealSense and a CT scan. An optical marker will be rigidly fixed to the patient’s head within the
camera’s field of view and used to preserve registration when the patient’s face is no longer visible.
Objective
Objective A: Perform patient registration using segmented facial surface anatomy and depth images streamed from the Intel RealSense
Objective B: Maintain patient registration after face is no longer visible using an optical marker tracked by the Intel RealSense RGB camera
Approach and Plan
- Acquire depth images from the RealSense in Slicer and use them to generate a point cloud
- Determine how to stream these depth images and calculate the point cloud in real time
- Use a captured point cloud to verify the accuracy of the registration to a model
- Use Slicer to track an optical marker defined rigidly relative to the real time point cloud
- Perform registration on the point cloud with the marker and maintain the registration without visibility of the facial anatomy
Progress and Next Steps
- Created Slicer module for generating a point cloud from a depth image
- Demonstrated streaming of real-time point clouds from depth images captured by the RealSense
- Investigated accuracy of Model Registration module for registering captured point clouds to dense models
- Explored RealSense parameters to find settings best suited to close range facial scanning
Illustrations
User interface of DepthImageToPointCloud module
Example of point cloud generated from depth image
Demonstration of streamed point cloud generated using RealSense depth data
Examples of acquired surface mesh registered to model
Basic workflow of the project
Background and References