The primary objective of this experiment is to establish a controlled testing environment for the collection of raw data. This data will be utilized to evaluate the depth accuracy achieved via the LuminaBone photometric approach.
Experimental Setup
A specialized camera holder is configured with four endoscope cameras. One unit functions as the primary imaging sensor, while the remaining three serve as LED light sources to simulate varied lighting conditions through sequential activation.
The procedure utilizes a spine sawbone model. At each designated pose, the three LEDs are toggled sequentially to capture a series of images before moving to the subsequent pose for the next data group. We will also capture one photo that has all LED lights up; this data will be used to calculate the accuracy based on the algorithm that the CMU paper uses.
The ground truth for this evaluation is derived from a high-fidelity 3D spine model.
To ensure optimal image quality and minimize ambient interference, the data collection must be conducted within a strictly controlled black-box environment.
Software Development
- Capture endoscope images. OpenCV’s example code can be used to capture the image.
- Turn on/off LED.