Deep Association for Sensor Fusion

A system comprising:

• at least one processor configured to:
• determine, using first sensor data obtained by a first sensor of a vehicle, a first plurality of tracks, each track of the first plurality of tracks related to a different first object of a first plurality of objects;
• determine, using second sensor data obtained by a second sensor of the vehicle, a second plurality of tracks, each track of the second plurality of tracks related to a different second object of a second plurality of objects;
• generate, based on the first plurality of tracks and the second plurality of tracks, an association score matrix indicative of a probability that each first object in the first plurality of objects and each second object in the second plurality of objects are a same object;
• apply a threshold to each element of the association score matrix;
• generate, based on applying the threshold to the association score matrix, an assignment matrix that is indicative of each first object from the first plurality of objects being associated with a second object from the second plurality of objects;
• and
• output, based on the assignment matrix and to an automotive system of the vehicle, information identifying that first objects from the first plurality of objects and second objects from the second plurality of objects, respectively, are same objects.

A computer-readable storage media comprising instructions that, when executed, configure a processor to:

• determine, using first sensor data obtained by a first sensor of a vehicle, a first plurality of tracks, each track of the first plurality of tracks related to a different first object of a first plurality of objects;
• determine, using second sensor data obtained by a second sensor of the vehicle, a second plurality of tracks, each track of the second plurality of tracks related to a different second object of a second plurality of objects;
• generate, based on the first plurality of tracks and the second plurality of tracks, an association score matrix indicative of a probability that each first object in the first plurality of objects and each second object in the second plurality of objects are a same object;
• apply a threshold to each element of the association score matrix;
• generate, based on applying the threshold to the association score matrix, an assignment matrix that is indicative of each first object from the first plurality of objects being associated with a second object from the second plurality of objects;
• and
• output, based on the assignment matrix and to an automotive system of the vehicle, information identifying that first objects from the first plurality of objects and second objects from the second plurality of objects, respectively, are same objects.