An Algorithm for Automated Feature Extraction from Laser Ranging Data

An algorithm is presented which extracts features from a time series of Flash LADAR images without a-priori knowledge of the surrounding environment. Flash LADAR utilizes laser ranging to acquire 3D information in the object space, which may be utilized to reconstruct the sensor environment. The extracted features are tested to identify motion artifacts and separate entities into static and non-static states. The separation of static and non-static features is essential to support mobile mapping systems in a variety of applications. In addition, static features can be utilized to provide a positional fix owing to the availability of range data, enabling navigation in environments where GPS is challenged or intermittent. The algorithm is linear based, robust to sensor noise, and utilizes image data to drive a series of heuristics on an autonomous basis. The algorithm can be utilized to extract static features without requiring manual input or adjustment. Location error from extracted features is examined on the basis of sequential frames.