This week we’re featuring IGNITE Finalist Noel Khan from California. In this interview, Noel tells us about his experience in software engineering that led him to the geospatial industry and the creation of his Smart M.App, AiGIS: US Tax Maps for automatically georeferencing US Assessor maps.
Q: Where are you from?
A: I was born and raised in California, except for about 6 years of my early childhood spent in Colorado Springs. My folks are from back east (…the subcontinent) and are a nomadic bunch, so we got to experience life in most California counties along the southern coast. I love LA’s cool coastal breeze, south Orange County’s tree-lined meandering roads, Oceanside’s marine layer, and the diversity in this progressive state.
Q: Where did you attend school and what did you study? Why did you choose this subject?
A: There’s a long engineering tradition in my family. My grandpa was a civil engineer, my dad an electrical engineer, and after pursuing electrical engineering, I fell in love with software engineering because of easier debugging, shorter development times, and no development costs. I completed a BS in Computer Science from American Sentinel University and spent over a decade writing software in the civil engineering/geospatial domain for local government.
When DARPA held the Urban Challenge in my hometown and I witnessed autonomous vehicles engage in traffic with human drivers. I came to the obvious realization that all of my software was confined to a computer. Inspired, I wanted my software to interact with the real world so I gravitated back to electrical engineering along with the popularization of hobbyist microcontrollers and single board computers. Eventually, I enrolled in a Computational Intelligence and Robotics graduate program at De Montfort University, where my research work includes pattern recognition.
Q: Where do you currently work? What is your job? What are your responsibilities?
A: I’ve spent the majority of my time in the geospatial industry, but my day job is now in the healthcare industry where I build software to automate processes for a local management services organization. However, I still consult in the geospatial industry, because that’s where all the cool projects are. For example, I worked on the hierarchical adjustment of a digital basemap based on changes in GPS control points, wrote data migrations tools to transfer historical and active data from MGDM/MicroStation to Oracle Spatial/Oracle Workspace Manager, built a development framework on top of a vendor’s web based LIS so I could simplify making productivity tools for end users. I also wrote self-programming spatial data warehousing software that generated its own code to process and clean 20GB of spatial data to both Oracle and Esri validation standards.
More recently, I turned my pattern recognition research into a commercial batch-oriented system and started doing business as EightQueens – catering to county governments who want to georeference caches of raster maps. That system uses computer vision and artificial intelligence techniques to georeference 1,000 maps per day on a 12 node cluster. But I’m not a business person; my objective is only to demonstrate value to support the sale of the IP in order to fund subsequent research.
Q: What are some of your favorite things to do during your free time?
A: I love baby and daddy time.
Q: Where did your interest in geospatial/mapping/information services come from?
A: The choice of industry didn’t matter much to me as a young programmer, as I couldn’t believe I was getting paid to do what I loved. I stumbled into this industry, but was pulled in by the scale and complexity of the projects.
Q: How did you hear about IGNITE?
A: I heard about IGNITE through a mass email.
Q: Why are you excited about this competition?
A: There’s always opportunity in new things. This is a new marketplace that isn’t yet saturated with everyone else’s apps, which means it’s easier for people to find your apps and there’s a greater opportunity to establish a reputation. Also, since Hexagon Goespatial recently launched various online forums, there’s a smaller volume of posts to go through to get up to speed. None of that may be true a year or two from now, so I strongly encourage folks to get onboard now.
Q: What is the problem your Smart M.App addresses?
A: Data integration, but further up the fishbone the problem is more fundamentally the labor cost to manually georeference raster maps. In Orange County, for example, the historical average cost is $10 per page. My batch oriented system automatically georeferences maps at $1 per page. For clients that need very high-precision results, maps can be fine-tuned in China for an additional $1 per page (pass through cost), while manual georeferencing from scratch is about $3 per page. That pricing demonstrates the cost savings of even rough georeferencing results and suggests the lions share of the cost comes from the global search – finding corresponding vicinities. It’s cheaper to fine tune.
Q: What is your Smart M.App called?
A: AiGIS is what I call the batch oriented system, but since I plan to create a franchise of Smart M.Apps that are tailored to different map types, I suppose I’ll call this one “AiGIS – US Tax Maps.” There’s no one “algorithm that solves all problems,” so you have to make custom algorithms for a narrowly defined set of problems if you want good results.
Q: What IGNITE Theme does it support?
A: Infrastructure/Civil Engineering.
Q: What does it do?
A: This Smart M.App automatically georeferences US tax maps to a GIS model in under a second on the desktop and about 1 minute online. The batch oriented system uses a different algorithm and takes 10 minutes per problem, but produces better results. Here, I’ve purposely struck a different tradeoff between speed and precision to make this kind of processing viable online. That’s not to say the results are necessarily worse, but that the Smart M.App algorithm may successfully solve a narrower set of maps than the more generic but slower algorithm.
Q: What do you hope to accomplish with your Smart M.App?
A: A motivating philosophy in my work is to “bring high technology down to the level of the common user,” which isn’t a new idea because it’s what eventually happens anyway. I want to accelerate the transfer of high technology from academia to ubiquitous personal use. For example, my team published a paper on automatic georeferencing just this past December and since then we redesigned the approach to reduce computation from 10 minutes on the desktop to 1 minute per problem online while maintaining acceptable results. That’s the technique used in this Smart M.App.
Secondly, there are many opportunities for data fusion and pattern recognition (especially in 3D) that will benefit the geospatial, remote sensing, and robotics communities. If this Smart M.App manages to inspire more people to work on that problem, we all benefit from that collective effort.