On Tuesday, May 22, from 12:00–1:00 pm please join us for the next Spatial Technology Lunch in the Center for Spatial Studies (Phelps Hall 3512). This semi-regular series, hosted by spatial@ucsb, aims to promote discussion and interaction within the university’s spatial technology community. Please RSVP to Crystal Bae (cbae@spatial.ucsb.edu) by Sunday, May 20. Pizza and drinks will be provided.

How well can a $750 DIY LiDAR scanner scan?

Jorge Chen

Chen Tech Lunch
Abstract: Laser scanners provide a fast, convenient, and accurate way to take distance measurements of the surrounding environment. They operate by calculating the time it takes for a light beam to travel to a distant object and back using a process called light detection and ranging, or LiDAR, that, when repeated numerous times, forms a “point cloud” of (x,y,z) coordinates. Until very recently, only large enterprise users with big budgets could afford this type of technology, with the cost of most LiDAR scanners running well over $100K. However, the trickling down of LiDAR to consumer products has resulted in a new class of relatively cheap sensors that can now be found in robotic vacuum cleaners, drones, autonomous vehicles, and maybe even in upcoming smartphones.

This presentation looks at the performance of the Scanse 3D panoramic LiDAR scanner, one of the first panoramic scanners designed for consumer use. At an incredible price of $750, this camera-sized do-it-yourself scanner uses a $150 LiDAR sensor attached to two orthogonally rotating servos that are controlled by open source software on a Raspberry Pi — all powered by an off-the-shelf cell phone charger. Performance assessment involved comparing measurements of a conference room taken with the Scanse 3D and a professional Trimble scanner. Results showed the Scanse consistently overestimated room dimensions by 15 cm, although at the local level of a flat surface it showed sub-centimeter accuracy, with high standard deviation and sub-centimeter precision. This latter result indicated systematic drift, which can be seen in a plot of the point cloud. Perhaps more interesting than the results, though, were the challenges faced in aligning the noisy and wavy Scanse data with the highly accurate and precise Trimble data. These were addressed using extended Gaussian image analysis, histogram analysis, and the iterative closest point process, all of which will be covered during the presentation.

Jorge Chen is currently a postdoctoral researcher in the Department of Geography at UCSB.