I had the pleasure of spending a few days using a DotProduct DPI-7 imager this week. This wasn’t my first time playing with the DPI-7, as I used it a bit when it first showed up at SPAR International (2013, I think). I wasn’t too impressed then, but the guys at DotProduct have made a lot of improvements in the intervening time.
The DPI-7 kit is comprised of a sensor array (made by Primesense), an Android powered tablet (in this case an HP Slate 7”), and DotProduct’s software (Phi.3D). The hardware is mounted on an adjustable handle and runs off of the battery in the tablet. All post processing is performed onboard the tablet using the Phi.3D software.
There are actually two possible configurations: the DPI-7, and the DPI-7sr, which uses a shorter range/higher accuracy sensor. The sr model is designed with a 0-3 ft. (0-0.9 m) range, as compared with the standard model’s range of 2-10 ft. (0.6-3.05 m).
How the software works
Scanning is fairly straight forward, as the screen displays a black & white image of the scan area in real time as you scan. A color gradient is overlaid to provide feedback about the scan data. As an area is captured, the scene turns yellow then green to indicate good coverage and stitching of the adjoining images. The screen also displays a vertical progress bar that indicates how full the RAM storage is on the tablet. This is important, as all scans are saved to RAM until the session is complete. Once the session is saved and optimized, it is stored to the tablet’s main memory so that the RAM is freed up for more scanning.
The DPI-7 can register a new session using targets, or by using a coordinate system to append the session to an existing session’s data. One of the problems with the older implementation of the append function, as I remember it, was that you had to load the previously recorded session. If it filled your RAM, you couldn’t do much about it. Fortunately, I was able to use a unit that had what I think was a beta version of Phi.3D, which unloads the session to which you are appending once the unit is able to orient itself to the existing data. This gives users the ability to continuously append and store multiple files that are all in the same coordinate system. This way you can import all of the files (sessions), and no registration is needed as they are all in the same UCS.
Most of the time that I tried this it worked pretty well. I did find that it worked better if I always appended from the same spot as opposed to daisy chaining the files. Any errors I did find were correctable in an application that allows cloud-to-cloud registration of non-gridded point clouds (like Cloud Compare). Given the improvement in technique over the 2-3 days that I had it, I do not doubt that my lack of skill was to blame on more than one occasion!
Capture of a Harley Davidson with the DPI-7
There are a couple of things to note about the DPI-7. Most of the complaints I had were hardware related, but that’s the part of the puzzle that DotProduct is leaving to others. While that may seem like a bad thing, the reality is that that multiple large companies are already improving tablet RAM and processing power, or developing better sensor resolution/light condition/accuracy, and they are doing it faster than a company of DotProducts’ size could ever hope to match. DotProducts’ business plan is to be a software company, not a hardware manufacturer. They are actively involved with multiple hardware manufacturers to embed their Phi-3D software in various platforms alongside their own DPI-7. These facts lead me to believe that the rate of improvement for the DotProduct line will be much faster than with manufacturers that build all of their own hardware.
Who and what is the DPI-7 for?
It’s definitely not for large areas. I scanned part of my office and it took about as long as a Z+F 5010 for the same area. However, when I used it to cut in areas of high interest, like the electrical breaker box or under-sink areas. it was a wonderful addition. Not only do you get the data as a point cloud, you can also export all of the still images, which allowed me to do things like read the labels on the breaker box. I think the idea is to use the device to punch in areas of change or where higher detail is needed within a larger dataset.
While not publically available, DotProduct does have tools that allow users to take point clouds from other scanners, convert them into a DP format and then use them in the Append tool with the DPI-7. I expect this to be an embedded function as soon as they can make it as idiot proof as the rest of the Phi-3D GUI.
Of course, you could also use it for documenting areas and assets even when there is not a larger dataset. Now that I’ve sent the unit back, I wish I had used it for insurance recordation at my house… However, with a retail price that’s right at $5000 USD, it may still be a little pricey for hobbyist use. That being said, I would argue that it is the most capable scanner that I’ve ever used that was anywhere near $5K.
Lastly, the compression used by the DPI-7 is also quite impressive. DP formatted files tended to be 8-12 MB and the same data in a RCS format was 80-115 MB. There is a plugin available that allows ReCap to read the native DP format, which also saves a bit of time if you are taking the data into Autodesk products. Another surprising perk was that the data is compatible with EdgeWise software, which typically requires all data to be statically collected. In fact, the data used to demonstrate the ducting tools during last month’s ClearEdge3D webinar was collected by a DPI-7 (DPI-7 data at 46:35).