Purpose: The purpose of this study is to develop a framework to quickly calibrate a near-infrared camera arbitrarily-set in the simulator or treatment room.
Methods: A calibration phantom was fabricated with a marker on the isocenter and the other six markers at 5-cm apart from the isocenter to the left, right, anterior, posterior, superior, and inferior direction. We used the Body Surface Tracker (BST) system comprising Kinect v2 camera (near-infrared camera; NIC) and Marker Tracker software which controlled and acquired the marker positions in the three-dimensional (3D) coordinate system aligned to the NIC. The calibration phantom was measured from ten different angles of the NIC. Each data was taken for 60 seconds with 30 frames per second. Data with shorter measurement times were mimicked by taking a part of the measured data. We developed a matrix to convert the NIC coordinate to the room coordinate with six parameters (three angles and three translation coordinates). The six parameters were then optimized to converge the converted room coordinates to the actual coordinates.The achieved spatial accuracy and random error were evaluated.
Results: The achieved spatial accuracy was 0.02 mm. There was no effect of the measurement time to the achieved spatial accuracy. Thus, 1-second measurement is considered to be long enough for the calibration. The random error for the actual respiratory monitoring was 0.46 mm for 1 frame. The random error was improved to 0.15 and 0.1 mm by averaging 10 and 20 frames, respectively.
Conclusion: We developed a quick calibration framework for arbitrarily-set NIC. This technique is considered to enable repetitive temporary setups of the NIC with no requirement of permanently occupying a space in the simulator or treatment room.