Feature extraction and separation in airborne laser scanner terrain integrity monitors

This paper describes the methodology and algorithms used in an implementation of a downward-looking airborne laser scanner (ALS) based terrain and feature integrity monitor. Using a high accuracy and high resolution ALS sensor, the described integrity monitor can first separate features from the terrain and then use the extracted feature data to detect and observe systematic and blunder errors in a terrain feature database. Two applications are envisioned for such a system - the first is to check the quality and update a terrain feature database and the second is as a real-time monitor on all aircraft which have systems that use terrain feature databases. Real-time terrain database integrity monitors have been studied at the Ohio University Avionics Engineering Center (AEC) for nearly 10 years. The feature integrity monitor is different from previous research performed at Ohio University in that it extracts specific features, such as buildings, roads, and towers, and performs a consistency check between these objects and a stored feature database. To perform the consistency check between the feature database and the ALS data a four-part building extraction algorithm is used. Once the high-frequency building shapes are extracted, they can be compared to the onboard feature database to determine changes and errors. The four-part building extraction algorithm described in this paper has the following characteristics: it works on non-uniformly spaced point-cloud data, it is designed for integrity monitoring rather than complete scene reconstruction, and the automatic feature extraction is not dependent on (but may use if available) a-priori feature shape information. This paper outlines the four-part building extraction algorithm and provides insight into its operation by applying the algorithms to ALS data collected on NASA's DC-8 Airborne Laboratory over Reno, Nevada in 2003.