If you’ve ever seen a Google Street View car rolling down your block, you know what a mobile mapping system is. These moving platforms are equipped with sensors that observe the surrounding environment, collecting data to create digital duplicates of the real world. You can imagine how much data is streaming from these multiple mobile mapping systems operating around the world. At Purdue, we’re using algorithms and artificial intelligence/deep learning to fuse this data to create a detailed digital picture of the planet.
Mobile mapping systems can be terrestrial, airborne or spaceborne. They roam the streets and skies to capture datasets that, in total, build digital replicas of the real world for applications in agriculture, transportation, infrastructure inspection and multiple other fields. Many data acquisition systems, sensors and platforms must be integrated to ensure the accuracy of this “digital twin.”
A good example involves highway work zones. We’ve all had to slow down as construction crews rebuild road surfaces. Monitoring traffic and maintaining roadway capacity in these zones is a challenging task for transportation agencies. There are set requirements for minimum lane widths, merging tapers and work zone geometry to ensure worker safety and traffic flow. But it often is not feasible, or is too unsafe, for inspectors to physically check these geometric features amid the bustle of freeway traffic.
Our Digital Photogrammetry Research Group has been building and using wheel-based mobile mapping systems equipped with LiDAR (Light Detection and Ranging) and imaging sensors to create an accurate geometric characterization of the work zone. This representation is integrated with connected-vehicle speed data to evaluate the impact of work zone geometry on traffic operations. Our research is the first of its kind and has been received well by other researchers and the Indiana Department of Transportation.
The data integration task for this application alone is formidable — as you imagine larger and larger areas and uses, the challenge is gargantuan. Mobile mapping systems are ubiquitous today, so we are inundated with a flow of geospatial data that contains a wide range of complementary characteristics pertaining to the mapped environment. To transform this raw data into reliable information, the interrelated data flood from multiple sensors, platforms and dates, each with differing timestamps, must be synchronized and calibrated precisely — for example, by adjusting for variables like temporal and spatial offsets to account for the position and orientation of the mapping platforms.
We have developed algorithms that integrate and model the acquired geospatial data and then extract features, match them with other data, classify and interpret them, and quantify observed changes. We also are in the process of developing special-purpose, artificial intelligence-inspired deep learning algorithms, turning them loose to crunch the tons of raw data and unearth patterns to derive more accurate mapping results.
This helps mobile mapping systems “fill in the blanks” in such areas as agriculture — for instance, gathering data to forecast crop yields by recording multi-modal remote sensing data over fields at a much finer scale. This allows us to gain deeper insight into such predictors including plant stress, chlorophyll content, and nitrogen levels.
Mobile mapping systems also chart shoreline erosion, recreate the dynamics of colliding vehicles at a crash scene, provide digital data for archeological documentation, and measure spatial and rotational displacements in infrastructure to detect rust, corrosion and fatigue. They’re even having an environmental impact: Google Street View cars equipped with air-quality sensors are helping the Environmental Defense Fund map street-level air quality in Oakland and Houston — collecting granular data at the point where people actually breathe the air.
Humans have been drawing and representing our world in maps for thousands of years; one of the earliest maps dates from the late 7th millennium BC, more than 8,000 years ago. Many advances over the years have increased the realism and rendering of the world in maps. Mobile mapping systems, one of the latest breakthroughs, are crucial to our digital reconstruction of the real world, helping us gain the detail and accuracy we need so we can confidently use digital models to innovate in a variety of fields.
We would like to recognize the contributions of the DOE ARPA-E TERRA Program and Joint Transportation Research Program (Purdue University), as well as the collaboration with Purdue University faculty members in Lyles School of Civil Engineering, Electrical and Computer Engineering, and College of Agriculture.
Thomas A. Page Professor of Civil Engineering
Lyles School of Engineering
Associate Director of the Joint Transportation Research Program
Co-Director of the Civil Engineering Center for Applications of UAS for a Sustainable Environment (SE-CAUSE)
Purdue University College of Engineering