Moving into the Galleria dell'Accademia

Photographs and text by Marc Levoy
January 25, 1999

At 8:30am on Monday, January 25, an air suspension truck rolls up the door of the Palazzo Bargagli-Petrucci, home of the Digital Michelangelo Project's Florence laboratory. Our laser scanner and about half of our computers are packed in crates and boxes, loaded into the truck and...

...driven across town to the Galleria dell'Accademia, which is closed today, as it is every Monday. The crates are unloaded and...

...rolled into the museum.

Our first statue will be the St. Matthew. The glass barricades surrounding it have already been removed in preparation for our arrival. We set up our computers on either side of the statue, we erect the Cyberware gantry, and we immediately begin scanning. Visible in this shot is, from left to right, Kari Pulli, Matt Ginzton, Jelena Jovanovic, Szymon Rusinkiewicz, and Brian Curless.

Brian Curless runs the gantry through its motions. The software that combines multiple range images was Brian's PhD thesis at Stanford University. Brian is now an Assistant Professor at the University of Washington.

Here is a closeup of the scanner head acquiring range data. As the head nods up and down, the red laser line sweeps across the statue. Each sweep returns thousands of 3D points spaced 0.29mm apart. After a sweep is finished, the head turns to the left or right and performs another sweep. After 10 minutes of nodding and turning, the scanner has captured whatever surfaces fall within a curved "shell" 50cm wide x 75cm high x 15cm thick. Click here to see our first shell. Because the shell is thin, the scanner head must translate back along its arm and acquire more shells, each at a different distance from the statue.

Once the scanner has finished acquiring geometry, it revisits the same part of the statue acquiring color. For this we use a high-resolution digital camera and a calibrated light source, which you can see illuminating the statue. To compensate for ambient illumination, we actually acquire two sets of color images, one with the light source switched on, and one with it switched off. When we have finished acquiring color, we move the arm up or down on its vertical truss or roll the gantry into another position and repeat the process. After some practice, we found that each statue (except the David) took us about 6 days to scan.

Meanwhile, in another part of the museum, our Cyra scanner is acquiring an approximate (1cm) model of the David to help us plan our high-resolution scan of the statue in a few weeks. (Click here to see photographs, taken three weeks later, showing us scanning the David.) We're also using the Cyra scanner to build a virtual model of the museum's floor, walls, and ceiling.

webmaster@graphics.stanford.firenze.it

	At 8:30am on Monday, January 25, an air suspension truck rolls up the door of the Palazzo Bargagli-Petrucci, home of the Digital Michelangelo Project's Florence laboratory. Our laser scanner and about half of our computers are packed in crates and boxes, loaded into the truck and...
	...driven across town to the Galleria dell'Accademia, which is closed today, as it is every Monday. The crates are unloaded and...
	...rolled into the museum.
	Our first statue will be the St. Matthew. The glass barricades surrounding it have already been removed in preparation for our arrival. We set up our computers on either side of the statue, we erect the Cyberware gantry, and we immediately begin scanning. Visible in this shot is, from left to right, Kari Pulli, Matt Ginzton, Jelena Jovanovic, Szymon Rusinkiewicz, and Brian Curless.
	Brian Curless runs the gantry through its motions. The software that combines multiple range images was Brian's PhD thesis at Stanford University. Brian is now an Assistant Professor at the University of Washington.
	Here is a closeup of the scanner head acquiring range data. As the head nods up and down, the red laser line sweeps across the statue. Each sweep returns thousands of 3D points spaced 0.29mm apart. After a sweep is finished, the head turns to the left or right and performs another sweep. After 10 minutes of nodding and turning, the scanner has captured whatever surfaces fall within a curved "shell" 50cm wide x 75cm high x 15cm thick. Click here to see our first shell. Because the shell is thin, the scanner head must translate back along its arm and acquire more shells, each at a different distance from the statue.
	Once the scanner has finished acquiring geometry, it revisits the same part of the statue acquiring color. For this we use a high-resolution digital camera and a calibrated light source, which you can see illuminating the statue. To compensate for ambient illumination, we actually acquire two sets of color images, one with the light source switched on, and one with it switched off. When we have finished acquiring color, we move the arm up or down on its vertical truss or roll the gantry into another position and repeat the process. After some practice, we found that each statue (except the David) took us about 6 days to scan.
	Meanwhile, in another part of the museum, our Cyra scanner is acquiring an approximate (1cm) model of the David to help us plan our high-resolution scan of the statue in a few weeks. (Click here to see photographs, taken three weeks later, showing us scanning the David.) We're also using the Cyra scanner to build a virtual model of the museum's floor, walls, and ceiling.