Systems and methods for detecting traffic violations using mobile detection devices

1. A method for detecting a traffic violation, comprising:

capturing a first video of a vehicle and a restricted road area using one or more video image sensors of a first edge device;

identifying, using one or more processors of the first edge device, the vehicle and the restricted road area from frames of the first video by applying functions from a computer vision library to the first video and passing at least some of the frames of the first video to one or more deep learning models running on the first edge device;

bounding, using the one or more processors of the first edge device, the vehicle and the restricted road area from the frames of the first video in a plurality of first bounding boxes;

detecting, at the first edge device, a first potential traffic violation based in part on overlap of the plurality of first bounding boxes and transmitting data obtained from the first video to a server;

capturing a second video of the vehicle and the restricted road area using one or more video image sensors of a second edge device;

identifying, using one or more processors of the second edge device, the vehicle and the restricted road area from frames of the second video by applying functions from a computer vision library to the second video and passing at least some of the frames of the second video to one or more deep learning models running on the second edge device;

bounding, using the one or more processors of the second edge device, the vehicle and the restricted road area from the frames of the second video in a plurality of second bounding boxes;

detecting, at the second edge device, a second potential traffic violation based in part on overlap of the plurality of second bounding boxes and transmitting data obtained from the second video to the server; and

determining, at the server, that a traffic violation has occurred based on the data received from the first edge device and the second edge device.

2. The method of claim 1 , wherein the restricted road area is a restricted lane of a roadway.

3. The method of claim 1 , wherein bounding the vehicle and the restricted road area in the plurality of first bounding boxes further comprises passing at least some of the frames of the first video to the one or more deep learning models running on the first edge device to bound the vehicle and the restricted road area in the plurality of first bounding boxes, and wherein bounding the vehicle and the restricted road area in the plurality of second bounding boxes further comprises passing at least some of the frames of the second video to the one or more deep learning models running on the second edge device to bound the vehicle and the restricted road area in the plurality of second bounding boxes.

4. The method of claim 1 , wherein the one or more deep learning models each comprise a convolutional neural network trained for object detection and wherein the one or more deep learning models are trained in part from videos captured using the first edge device and the second edge device.

5. The method of claim 1 , wherein the one or more deep learning models comprises a convolutional neural network trained for object detection and wherein the one or more deep learning models are trained in part from videos captured using edge devices other than the first edge device and the second edge device.

6. The method of claim 1 , wherein the method further comprises:

bounding the vehicle from the frames of the first video using a first vehicular bounding box;

bounding the restricted road area from the frames of the first video using a first road bounding box;

detect the first potential traffic violation by detecting an overlap of the first vehicular bounding box with the first road bounding box;

bounding the vehicle from the frames of the second video using a second vehicular bounding box;

bounding the restricted road area from the frames of the second video using a second road bounding box; and

detect the second potential traffic violation by detecting an overlap of the second vehicular bounding box with the second road bounding box.

7. The method of claim 1 , further comprising generating, using the one or more server processors, a simulation of the traffic violation using a reasoning engine running on the server by utilizing a game engine based on data and video frames obtained from at least one of the first video and the second video.

8. The method of claim 1 , wherein the first edge device is coupled to a first carrier vehicle, wherein the first video is captured using the one or more video image sensors of the first edge device while the first carrier vehicle is in motion, and wherein the second edge device is coupled to a second carrier vehicle, wherein the second video is captured using the one or more video image sensors of the second edge device while the second carrier vehicle is in motion.

9. A system for detecting a traffic violation, comprising:

a first edge device comprising one or more first edge device processors and one or more first video image sensors,

wherein the first video image sensors are configured to capture a first video of a vehicle and a restricted road area;

wherein the one or more first edge device processors are programmed to:

identify the vehicle and the restricted road area from frames of the first video by applying functions from a computer vision library to the first video and passing at least some of the frames of the first video to one or more deep learning models running on the first edge device;

bound the vehicle and the restricted road area from the frames of the first video in a plurality of first bounding boxes; and

detect a first potential traffic violation based in part on overlap of the plurality of first bounding boxes and transmit data obtained from the first video to a server; and

a second edge device comprising one or more second edge device processors and one or more second video image sensors,

wherein the second video image sensors are configured to capture a second video of the vehicle and the restricted road area;

wherein the one or more second edge device processors are programmed to:

identify the vehicle and the restricted road area from frames of the second video by applying functions from a computer vision library to the second video and passing at least some of the frames of the second video to one or more deep learning models running on the second edge device;

bound the vehicle and the restricted road area from the frames of the second video in a plurality of second bounding boxes; and

detect a second potential traffic violation based in part on overlap of the plurality of second bounding boxes and transmit data obtained from the second video to the server; and

wherein the server comprises one or more server processors programmed to determine that a traffic violation has occurred based on the data received from the first edge device and the second edge device.

10. The system of claim 9 , wherein the restricted road area is a restricted lane of a roadway.

11. The system of claim 9 , wherein the one or more first edge device processors are further programmed to bound the vehicle and the restricted road area in the plurality of first bounding boxes by passing at least some of the frames of the first video to the one or more deep learning models running on the first edge device to bound the vehicle and the restricted road area in the plurality of first bounding boxes, and wherein the one or more second edge device processors are further programmed to bound the vehicle and the restricted road area in the plurality of second bounding boxes by passing at least some of the frames of the second video to the one or more deep learning models running on the second edge device to bound the vehicle and the restricted road area in the plurality of second bounding boxes.

12. The system of claim 9 , wherein the one or more first edge device processors are further programmed to:

bound the vehicle from the frames of the first video using a first vehicular bounding box;

bound the restricted road area from the frames of the first video using a first road bounding box; and

detect the first potential traffic violation by detecting an overlap of the first vehicular bounding box with the first road bounding box;

wherein the one or more second edge device processors are further programmed to:

bound the vehicle from the frames of the second video using a second vehicular bounding box;

bound the restricted road area from the frames of the second video using a second road bounding box; and

detect the second potential traffic violation by detecting an overlap of the second vehicular bounding box with the second road bounding box.

13. The system of claim 9 , wherein the server comprises a reasoning engine running on the server, wherein the reasoning engine generates a simulation of the traffic violation utilizing a game engine based on data and video frames obtained from the first video and the second video.

14. The system of claim 9 , wherein the first edge device is coupled to a first carrier vehicle, wherein the first video is captured using the one or more first video image sensors of the first edge device while the first carrier vehicle is in motion, and wherein the second edge device is coupled to a second carrier vehicle, wherein the second video is captured using the one or more second video image sensors of the second edge device while the second carrier vehicle is in motion.

15. The system of claim 9 , wherein the one or more deep learning models comprise a convolutional neural network.

16. A device for detecting a potential traffic violation, comprising:

one or more video image sensors configured to capture a video of a vehicle and a restricted road area; and

one or more processors programmed to executed instructions to:

identify the vehicle and the restricted road area from frames of the video by applying a plurality of functions from a computer vision library to the frames and passing the frames to one or more deep learning models running on the device to identify the vehicle and the restricted road area;

bound the vehicle in the frames with a vehicular bounding box and bound the restricted road area in the frames with a road bounding box; and

detect that a potential traffic violation has occurred based in part on overlap of the vehicular bounding box with the road bounding box.

17. The device of claim 16 , wherein the restricted road area is a restricted lane of a roadway.

18. The device of claim 16 , wherein the device is coupled to a carrier vehicle, and wherein the video is captured using the one or more video image sensors of the device when the carrier vehicle is in motion, and wherein the one or more deep learning models are running on the device when the carrier vehicle is in motion.

19. The device of claim 16 , wherein the one or more processors are further programmed to pass the frames of the video to the one or more deep learning models to bound the vehicle in the frames with the vehicular bounding box and bound the restricted road area in the frames with the road bounding box.

20. The device of claim 16 , wherein the one or more deep learning models comprise a convolutional neural network.