Systems and methods for collision detection and classification

A system (105) for detecting a vehicle collision, comprising:

• at least one processor coupled to a non-transitory memory, the at least one processor configured to:
• detect a vehicle event based on inertial sensor data and speed from at least one sensor in a housing inside a cabin of a vehicle;
• process the inertial sensor data, the speed, and vehicle class data of the vehicle with an initial classifier to generate an indication that the detected vehicle event is a collision event or a non-collision event, wherein the indication comprises an initial collision classification confidence value;
• in response to generating an indication that the detected vehicle event is a collision event, compare (310) the initial collision classification confidence value with a predefined confidence threshold;
• in response to a determination that the initial collision classification confidence value is below a predefined confidence threshold, determine (320) whether a location of the vehicle associated with the detected event is within a distance of a non-collision cluster point, wherein the non-collision cluster point is within a region identified by executing a clustering technique using a plurality of vehicle events;
• in response to a determination that the location of the vehicle is not within a distance of a non-collision cluster point, classify an event subclass of the collision event or the non-collision event based on the inertial sensor data, the speed, and the vehicle class data of the vehicle;
• update a size of the region in response to detecting further non-collision events within or proximate to the region;
• and
• generate a notification based on the event subclass.

A non-transitory computer-readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the steps of the method of:

• receiving, by at least one processor, vehicle class data of a vehicle (110), and inertial sensor data and speed from at least one sensor (135) in a housing inside a cabin of the vehicle (110);
• detecting (302), by the at least one processor, a vehicle event based on the inertial sensor data and the speed;
• processing (304), by the at least one processor, the inertial sensor data, the speed, and vehicle class data of the vehicle with an initial classifier to generate an indication that the detected vehicle event is a collision event or a non-collision event, wherein the indication comprises an initial collision classification confidence value;
• in response to generating an indication that the detected vehicle event is a collision event, comparing (310), by the at least one processor, the initial collision classification confidence value with a predefined confidence threshold;
• in response to a determination that the initial collision classification confidence value is below the predefined confidence threshold, determining (320) by the at least one processor whether a location of the vehicle associated with the detected event is within a distance of a non-collision cluster point, wherein the non-collision cluster point is within a region identified by executing a clustering technique using a plurality of vehicle events;
• determining, by the at least one processor, whether the vehicle event corresponds to a non-collision event based on the location of the vehicle being within the region;
• in response to a determination that the location of the vehicle (110) is not within a distance of a non-collision cluster point, classifying (314), by the at least one processor, an event subclass of the collision event based on the inertial sensor data, the speed, and the vehicle class data of the vehicle;
• generating, by the at least one processor, a notification based on the event subclass;
• and
• suppressing, by the at least one processor, the notification if the vehicle event is determined to correspond to the non-collision event based on the location of the vehicle being within the region.