Method identifying presence of occluded objects by the light emitted by said objects, implemented in autonomous vehicles, comprising: obtaining sensor data 504; obtaining predetermined environment data 506, which comprises positions of known light sources; building a three-dimensional scene of the environment 503 from the sensor and environment data; comparing known light sources in the environment to detected light sources to determine an unexpected light source 510 thereby the light emitting object; and controlling a trajectory of the autonomous vehicle 512. Sensor data may be obtained from cameras, light-intensity sensors and LIDAR. Determination of unexpected light in the environment may be dependent on whether the sensor data meets a criterion. Scene data may include: vehicle location; time of day; and weather conditions. Determining the unexpected light source may comprise creating a differential scene between light intensities of the environment and the sensor data; an unexpected light intensity indicating an unexpected light source. Location and identity of the hidden light emitting object may be determined from the position of the unexpected light, through reverse ray tracing from the unexpected light source. Object types may include: vehicles; cars; motorcycles; pedestrians; and bicycles. A trajectory of the hidden object may be determined.
Light-based object localization
2023-09-20
Patent
Electronic Resource
English