Sensor-agnostic detection using high-level feature extraction

Abstract

The systems and methods described herein disclose detecting objects in a vehicular environment using extracted high level features. As described here, a survey vehicle detects objects in the environment with a survey sensor. Vehicles can then use the extracted features from the survey sensor data to compare with extracted features from input sensor data. Thus, vehicles can benefit from different sensor data without employing said sensor. The systems and methods can include extracting one or more survey features from a survey data set, which can then be mapped to a feature-based map of an environment. An input data set can then be collected using an input sensor type with one or more input features being extracted from the input data set. The one or more input features can then be correlated to the one or more survey features to identify an object in the environment.