Using the Minolta Vivid 700 Scanner

These are instructions for students in 15-869, Image-Based Modeling and Rendering, for using the Minolta Vivid 700 scanner, particularly to acquire shape and color data for Assignment 2.

This scanner belongs to Martial Hebert's lab. Robotics graduate student Owen Carmichael showed me how to use it. They have generously allowed us to use their scanner. Please be considerate, defer to regular lab users, take good care of their equipment, and make sure their lab is locked when you leave.

The instructions I've written here are a summary of what Owen advised me and what I've learned from experimenting with the scanner a bit.

The first time you scan, this procedure might take you 15-30 minutes. Once you have the routine down, it could be done in under 5 minutes. You might want to print this page and take it with you.


The scanner is in Building D, room 147, one floor below ground level. The phone number in that room is 8-7931. To get in the room, you can borrow a key from me. Phone me at 8-7899 or come by my office if you want to borrow the key or need the password to log in.

The Minolta Vivid 700 sits next to the SGI O2 computer (blue rounded thing) in the far corner of the far room. Here are some of Minolta's web pages on the device, listing some specs on range, resolution, etc. There should be a binder containing Minolta's documentation for the scanner hardware and software somewhere near the O2, but it's not particularly helpful for beginners.


From the menus, pull down remote and then one shot. I'll abbreviate this remote / one shot.

Now you're in a mode where you can capture new color images and range data. The laser doesn't come on until you hit the release button. You can move the Vivid around a bit (but keep in mind that it's cabled to the O2) to point it at the lawn animals on the table, or out into the room if you want to scan your face. Prop it on binders to aim it up or down. The laser shoots out of the top window, and the camera for color and range is the lens at bottom.

Owen advises us to close our eyes when scanning our faces, or if standing within the laser swath, while the Vivid is scanning. The documentation says that only gazing into the laser beam at close range would cause damage, but to be on the safe side, we advise that you close your eyes.

Hitting release results in a beep, and a brief initial horizontal laser pulse to come out in the middle of the scanning area. The machine uses this to adapt the laser intensity for the scan, which follows a split second later, and takes only .6 sec. The range is about .6m to 2.5m, so you won't be able to get depths across the room. Close or white objects require lower laser intensity, while far or black objects require higher intensity. The room lights affect the color image, but don't affect the laser much. You can adjust the zoom if you want a narrower field of view. Clicking on the zoom slider bar (without changing the zoom level) will cause a new color image to be grabbed, without the laser. Sometimes doing this improves the focus of the color image.

The right window shows the range data, with white for near, dark gray for far, and black for unknown depth. It can take some trial and error to get the correct depth range. If the center of the depth range is too large, nearby objects (e.g. the tip of your nose) can get cut off. If the depth center is too small, distant objects will be cut off. Sometimes it helps to place a piece of cardboard or other background surface just behind the object being scanned. By default, the depth range is determined by that initial laser pulse. Moving the background in will apparently reduce the average depth (reducing the depth center).

Another way to adjust the depth range is by hitting the option and manual buttons, and adjusting the distance slider. The laser slider there adjusts laser intensity. The automatic intensity setting is usually fine, but I found it helpful to turn up the laser intensity in order to acquire depths for a piece of cardboard placed behind my head, for example. Setting this too high can cause distant objects to be erroneously sensed as nearby.

Save Data

Once you have a scan you like, hit convert, type in the basename for this scan, e.g. paul1, and reduce rate 1/1 (don't reduce), fill holes? no (I'm not sure I trust their algorithm), save camera data? no. Then hit OK.

Your data, displayed as a red grid and dots will be displayed in four views. If you want, you can spin it around a bit with view / orbit and view it with various options, but most importantly, save your data with:

At this point you can file / quit the program, if you're confident you're done, or if you want to scan again, I suggest you edit / delete / elements to clean out the workspace, otherwise multiple scans will appear in 3-D space. To scan again, return to Scanning above.

Move Your Data Files

These iv files take up 3 to 5 MB each. You should transfer your files to a private place and then delete them on meatball or move them to /tmp. For example, if you have a CS account, klog userid and mv yourfile.iv /afs/cs/project/classes-ph/869/students/userid/asst2 (assuming that dir exists). If you have an Andrew ID, the procedure is similar. If that doesn't work (I have seen it fail, for unknown reasons), you could try ftp.

The Open Inventor file you created can be viewed with the ivview command on SGI machines (note that on our large files, it can take this program 30 seconds to display a picture). Here is some documentation on the Open Inventor file format. Files created by the Vivid software, following the procedure above, will look something like this data/example.iv. The sections of interest are Texture2 (a 400x400 pixel 24-bit color picture), TextureCoordinate2 (pixel (x,y) for each of the points to follow), and Coordinate3 ((x,y,z) for the shape). The number of points in the last two sets will be equal, and it will be 200x200 or less (fewer if many points had unknown depth). Software will be provided to parse these files and convert it into a form useful for Assignment 2, so you needn't understand this file format.


Cleaning Up

What to Scan

I suggest you do a simple scan first to familiarize yourself with the machine. For assignment 2, you only need one scan, unless you're going to explore some of the extra credit options. If you have the time/energy, you could try to scan something else. A few tips: the rangefinder doesn't work as well on specular surfaces, hair, beards, or glasses as it does on diffuse surfaces. You might want to bring in some cloth to use as a backdrop. Objects between 10cm and 100cm or so in size are probably best, since the laser has limited range.

Interfacing to the Vivid Directly

Scott Thayer has source code and documentation for controlling the Vivid directly and acquiring data in a lower level format (an alternative to the "vivid" program). See his email.

Paul Heckbert, 6 Oct 99