Award

LSA Autonomy purposes a ground vehicle mounted LIDAR-based methodology that employs advanced sensor processing algorithms to reliably detect partially obscured objects, such as roadside hazards, beneath light and medium vegetation. Our approach leverages both sensor domain and point cloud analysis. By employing our proven sensor domain analysis techniques, we are able to accurately pre-classify laser returns at the sensor's full data rate. Following the sensor domain analysis, a local world model is constructed from a temporal history of pre-classified LIDAR points. The local world model significantly improves the likelihood of penetrating obscuring foliage and accumulating detections from objects beneath the foliage. Objects that are partially obscured by clutter are identified as potential concealed hazards and reported to the user for further assessment. We leverage our existing LIDAR sensor processing framework to accelerate laser point classification and object recognition. The result is a robust and reliable detection algorithm for partially obscured roadside hazards.