Camera and radar-based threat monitoring for micromobility platforms
Embodiments described herein provide systems and methods for threat detection for micromobility platforms based on image and range data. An example method generally includes capturing image data from a first sensor and corresponding ranging data from a second sensor communicatively coupled with a processing system, all mounted on a micromobility platform; identifying, based on an object detection machine learning model and the image data, one or more objects to monitor; predicting, based on processing the image data and the corresponding ranging data through a machine learning model, future motion of the identified one or more objects; determining that a threatening object from the identified one or more objects is a collision threat to the micromobility platform; and recording subsequent image data from the first sensor based on the determination until the machine learning model determines that the threatening object is no longer a collision threat has elapsed.
1 . A processor-implemented method, comprising:
capturing image data from a first sensor and corresponding ranging data from a second sensor communicatively coupled with a processing system, the first sensor, the second sensor, and the processing system being mounted on a micromobility platform;
identifying, based on an object detection machine learning model and the image data, one or more objects to monitor;
predicting, based on processing the image data and the corresponding ranging data through a machine learning model, future motion of the identified one or more objects;
determining, based on the predicted future motion of the identified one or more objects and a type of the identified one or more objects, that a threatening object from the identified one or more objects is a collision threat to the micromobility platform, wherein determining that the threating object is a collision threat comprises:
comparing a predicted future location of the threatening object, determined based on the predicted future motion of the threatening object, to a predicted future location of the micromobility platform,
determining that the threatening object is a collision threat when the predicted future location of the threatening object is within a threshold location of the predicted future location of the micromobility platform, and
the threshold location comprises a specified distance from the micromobility platform and a lateral range in which a vector points, the vector starting at a prior location associated with the threatening object in the image data and ending at a current location associated with the threatening object in the image data; and
recording subsequent image data from the first sensor based on the determination until a threshold amount of time after which the machine learning model determines that the threatening object is no longer a collision threat has elapsed.
2 . The method of claim 1 , wherein the captured image data is captured at a lower resolution and periodicity than the subsequent image data.
3 . The method of claim 1 , wherein the object detection machine learning model is trained to classify objects in the image data into different types of vehicles and disregard objects that are not likely to be a collision threat to the micromobility platform.
4 . The method of claim 1 , wherein the defined point comprises a vanishing point in the image data in which lines converge.
5 . The method of claim 1 , further comprising transmitting, to a display device communicatively coupled with the processing system, information about the threatening object for display on the display device.
6 . The method of claim 5 , wherein the information comprises an indication of a type of the threatening object, a velocity delta between the threatening object and the micromobility platform, and a predicted path of the threatening object.
7 . The method of claim 5 , wherein the display device comprises a smartphone.
8 . The method of claim 5 , wherein the display device comprises an augmented reality display.
9 . The method of claim 1 , wherein the predicted future motion of the threatening object comprises a predicted path and velocity of the threatening object.
10 . The method of claim 1 , further comprising activating one or more signaling lights displaying an alert to the threatening object based on determining that the threatening object is a collision threat.
11 . The method of claim 1 , wherein the first sensor comprises an imaging sensor and the second sensor comprises a millimeter wave radar.
12 . The method of claim 1 , further comprising:
determining a tilt angle of the micromobility platform relative to a reference plane; and
determining a relative position of the threatening object and the micromobility platform based on the tilt angle.
13 . A system for detecting collision threats on a micromobility platform, comprising:
an imaging device mounted on a micromobility platform;
a ranging device mounted on the micromobility platform; and
a processing system communicatively coupled with the imaging device and the ranging device and having a memory with executable instructions stored thereon and one or more processors configured to execute the executable instructions to cause the processing system to:
capture image data from the imaging device and corresponding ranging data from the ranging device;
identify, based on an object detection machine learning model and the image data, one or more objects to monitor;
predict, based on processing the image data and the corresponding ranging data through a machine learning model, future motion of the identified one or more objects;
determine, based on the predicted future motion of the identified one or more objects and a type of the identified one or more objects, that a threatening object from the identified one or more objects is a collision threat to the micromobility platform, wherein to determine that the threating object is a collision threat, the one or more processors are configured to cause the processing system to:
compare a predicted future location of the threatening object, determined based on the predicted future motion of the threatening object, to a predicted future location of the micromobility platform,
determine that the threatening object is a collision threat when the predicted future location of the threatening object is within a threshold location of the predicted future location of the micromobility platform, and
the threshold location comprises a specified distance from the micromobility platform and a lateral range in which a vector points, the vector starting at a prior location associated with the threatening object in the image data and ending at a current location associated with the threatening object in the image data; and
record subsequent image data from the imaging device based on the determination until a threshold amount of time after which the machine learning model determines that the threatening object is no longer a collision threat has elapsed.
14 . The system of claim 13 , wherein the captured image data is captured at a lower resolution and periodicity than the subsequent image data.
15 . The system of claim 13 , wherein the one or more processors are further configured to cause the processing system to transmit, to a display device communicatively coupled with the processing system, information about the threatening object for display on the display device.
16 . The system of claim 13 , wherein the predicted future motion of the threatening object comprises a predicted path and velocity of the threatening object.
17 . A non-transitory computer-readable medium having executable instructions stored thereon which, when executed by one or more processors, causes the one or more processors to perform an operation comprising:
capturing image data from a first sensor and corresponding ranging data from a second sensor communicatively coupled with a processing system, the first sensor, the second sensor, and the processing system being mounted on a micromobility platform;
identifying, based on an object detection machine learning model and the image data, one or more objects to monitor;
predicting, based on processing the image data and the corresponding ranging data through a machine learning model, future motion of the identified one or more objects;
determining, based on the predicted future motion of the identified one or more objects and a type of the identified one or more objects, that a threatening object from the identified one or more objects is a collision threat to the micromobility platform, wherein determining that the threating object is a collision threat comprises:
comparing a predicted future location of the threatening object, determined based on the predicted future motion of the threatening object, to a predicted future location of the micromobility platform,
determining that the threatening object is a collision threat when the predicted future location of the threatening object is within a threshold location of the predicted future location of the micromobility platform, and
the threshold location comprises a specified distance from the micromobility platform and a lateral range in which a vector points, the vector starting at a prior location associated with the threatening object in the image data and ending at a current location associated with the threatening object in the image data; and
recording subsequent image data from the first sensor based on the determination until a threshold amount of time after which the machine learning model determines that the threatening object is no longer a collision threat has elapsed.