RIT will permanently shut down the RIT Wiki (Confluence wiki) on February 2, 2024. Review the Wiki Shut Down Overview for information on alternative tools, exporting content from the wiki, and other FAQs.
To find the best possible configuration of cameras, specific characteristics of the array were varied. These variations were done in the model for efficiency and so the hardware of the physical array could remain in tact. The characteristics were varied in increments based off of the prototype array so that the results could be compared. Each configuration has a video of it being run through the program.
Cameras
This configuration uses Raspberry Pi cameras in place of the Point Grey Chameleons that are in the prototype. The Raspberry Pi cameras were chosen because they are inexpensive and compatible with the array's hardware. The file is labelled RaspberryPiCams.ogv and is shown below. The Point Grey model is also shown below for comparison.
Camera Number
These configurations have a fixed width and a changing number of cameras. The files are labelled by the number of cameras in the array. There are videos for configurations with 2, 3, 4, 6, 8, 10, 12, and 24 cameras. The videos use the targets from the static validation and focus on each target's depth to show how well the cameras can see through occlusions. It has been concluded that using between 4 and 12 cameras has the best effect.
Spacing
These configurations have a fixed number of cameras and changing width. The files are labelled by the distance between each camera. There are videos for configurations with cameras 5cm, 25cm, 35cm, 45cm, 55cm, 65cm, and 85cm apart. It has been concluded that cameras spaced between 25cm and 55cm apart work have the best effect.
Geometry
This configuration tests the functionality of an array with curved geometry. Each camera's angle increases by 10° from cameras 1-6. For comparison, there is an array with linear geometry. It has been concluded that using a linear array has the best effect.
Rows
These configurations use multiple rows of cameras in different locations. The files are named by the locations of the additional rows. There is a video for a configuration with a row 30cm above the original and another with rows 30cm both above and below the original. For comparison, there is a video with a single row for comparison. It has been concluded that a single row of cameras has the best effect.
Crowd Tests
To further determine which configurations are the most effective, dynamic tests were run in which the figure moving was obscured by several static figures so that each camera would have entirely different perspectives. This test was run for the arrays with varied camera numbers that were within the ranges previously determined. The files are labelled as they are for sections 2 and 3, but with "Crowd" before the title (ex. "Crowd4Cams.ogv" is the video for this test using the array with 4 cameras). Results have not been concluded.