Video-based system for automated detection of double parking violations
A method for detecting a double-parked vehicle includes identifying a parking region in video data received from an image capture device monitoring the parking region. The method includes defining an enforcement region at least partially surrounding the parking region. The method includes detecting a stationary vehicle in the enforcement region. The method includes determining the occurrence of an event relative to the stationary vehicle. In response to the determined occurrence of the event, the method includes classifying the stationary vehicle as being one of double parked and not double parked.
1. A computer-implemented method for detecting a double-parked vehicle, the method comprising:
at a server computer, executing instructions for:
identifying a parking region in video data received from an image capture device monitoring the parking lane;
defining an enforcement traffic lane at least partially surrounding the parking lane;
detecting a stationary candidate double-parked vehicle in the enforcement lane, the detecting including:
detecting a presence of vehicles in a difference image formed between a selected frame and a background estimate,
generating a binary image using the difference image, and
determining if each detected vehicle is moving or stationary using the binary image and a current frame, wherein the current frame is different from the selected frame;
in response to the detected vehicle being stationary, performing a normalized cross-correlation between a first frame where the vehicle was detected and subsequent frames to determine a duration the detected vehicle remains stationary, the determining including:
in response to the cross-correlation meeting or exceeding a predetermined threshold for a given frame, determining that the detected vehicle has not moved away from a location where it was originally detected and is a candidate violator, and
calculating the duration based on a number of frames that the detected vehicle remained stationary;
in response to the duration meeting a predetermined time threshold, determining whether hazard lights are operating on the candidate violator; and,
in response to the hazard lights operating on the candidate violator, classifying the detected vehicle as being double-parked.
2. The method of claim 1 , wherein the determining whether the hazard lights are operating includes:
identifying a hazard light region in the sequence of frames surrounding one of a front light and rear light; and,
determining pixel colors in the hazard light region in the sequence of frames.
3. The method of claim 1 further comprising:
in response to the hazard lights operating on the candidate violator, determining if the candidate violator is one of an emergency vehicle and a commercial vehicle; and
in response to the candidate violator being one of the emergency vehicle and the commercial vehicle, classifying the detected vehicle as being double-parked.
4. The method of claim 1 , wherein the determining the occurrence of the event includes:
detecting a second stationary vehicle in the parking lane located within proximity to the candidate violator in the enforcement lane.
5. The method of claim 1 , wherein the detecting the vehicle is performed using one of background estimation and subtraction, tracking, temporal difference, optical flow, and an initialization process.
6. The method of claim 1 , wherein the detecting the vehicle includes:
determining a number of frames the detected vehicle is located in a same position;
comparing the number of frames to a threshold; and,
in response to the number of frames meeting or exceeding the threshold, classifying the detected vehicle as the candidate violator.
7. The method of claim 1 further comprising:
in response to the candidate violator being classified as double-parked, associating the detected vehicle as being a double-parked vehicle and providing the user with a notification as output.
8. The method of claim 1 , further comprising:
considering multiple events for confirming whether the candidate violator is double-parked.
9. A system for detecting a double-parked vehicle, the system comprising:
a computer including at least one processor adapted to execute a set of instructions stored in a memory, the computer:
identifying a parking lane and an enforcement traffic lane at least partially surrounding the parking lane in a sequence of frames received from an image capture device monitoring the parking lane;
detecting a stationary candidate double-parked vehicle in the enforcement lane, the detecting including:
detecting a presence of vehicles in a difference image formed between a selected frame and a background estimate,
generating a binary image using the difference image, and
determining if each detected vehicle is moving or stationary using the binary image and a current frame, wherein the current frame is different from the selected frame;
in response to the detected vehicle being stationary, performing a normalized cross-correlation between a first frame where the vehicle was detected and subsequent frames to determine a duration the detected vehicle remains stationary, the determining including:
in response to the cross-correlation meeting or exceeding a predetermined threshold for a given frame, determining that the detected vehicle has not moved away from a location where it was originally detected and is a candidate violator, and
calculating the duration based on a number of frames that the detected vehicle remained stationary;
in response to the duration meeting a predetermined time threshold, determining whether a second vehicle is located in front of the candidate violator; and,
in response to a distance between the second vehicle and the stationary candidate double-parked vehicle being below a predetermined distance threshold, classify the detected vehicle as not being double-parked.
10. The system of claim 9 wherein the processor is operative to:
in response to the distance between the second vehicle and the candidate violator being above the predetermine distance threshold, determine if the candidate violator is one of an emergency vehicle and a commercial vehicle; and
in response to the candidate violator being one of the emergency vehicle and the commercial vehicle, classify the detected vehicle as not being double-parked.
11. The system of claim 9 , wherein the processor is further operative to:
in response to the distance between the second vehicle and the candidate being equal to or above the predetermined distance threshold, determine if the object is a stop light;
determine pixel colors in the stop light region in the sequence of frames; and,
in response to the pixel colors being green, classify the candidate violator as being double-parked.
12. The system of claim 9 , wherein the computer is further programmed to detect the stationary candidate double-parked a vehicle using background estimation and subtraction, temporal difference, optical flow, and an initialization process.
13. The system of claim 9 , wherein the computer is further programmed to:
in response to the candidate violator being classified as double-parked, associate the detected vehicle as being a double-parked vehicle and provide the user with a notification as output.
14. The system of claim 9 , wherein the computer is further programmed:
consider multiple events for confirming whether the detected vehicle is double-parked.
15. The system of claim 9 further comprising an output device for providing the classification to a user.
16. A computer-implemented method for detecting a double-parked vehicle, the method comprising:
at a server computer, executing instructions for:
identifying a parking region in video data received from an image capture device monitoring the parking lane;
defining an enforcement traffic lane at least partially surrounding the parking lane;
detecting a stationary candidate double-parked vehicle in the enforcement lane, the detecting including:
detecting a presence of vehicles in a difference image formed between a selected frame and a background estimate,
generating a binary image using the difference image, and
determining if each detected vehicle is moving or stationary using the binary image and a current frame, wherein the current frame is different from the selected frame;
in response to the detected vehicle being stationary, performing a normalized cross-correlation between a first frame where the vehicle was detected and subsequent frames to determine a duration that the detected vehicle remains stationary, the determining including:
in response to the cross-correlation meeting or exceeding a predetermined threshold for a given frame, determining that the detected vehicle has not moved away from a location where it was originally detected and is a candidate violator, and
calculating the duration based on a number of frames that the detected vehicle remained stationary;
in response to the duration meeting a predetermined time threshold, searching for a second vehicle in the sequence of frames;
determining a trajectory of the second vehicle relative to the candidate violator across a number of frames;
determining whether the trajectory moves around the candidate violator; and,
in response the trajectory moving around the candidate violator, classifying the detected-vehicle as being double-parked.
17. The method of claim 16 further comprising:
in response the trajectory not moving around the candidate violator, determining if the candidate violator is one of an emergency vehicle and a commercial vehicle; and
in response to the candidate violator being one of the emergency vehicle and the commercial vehicle, classifying the detected vehicle as not being double-parked.
18. The method of claim 16 further comprising:
detecting a second stationary vehicle in the parking lane located within proximity to the detected vehicle in the enforcement lane.
19. The method of claim 16 , wherein the detecting each vehicle is performed using one of background estimation and subtraction, tracking, temporal difference, optical flow, and an initialization process.
20. The method of claim 16 further comprising:
in response to the candidate violator being classified as double-parked, associating the detected vehicle as being a double-parked vehicle and providing the user with a notification as output.
21. The method of claim 1 further comprising:
considering multiple events for confirming whether the candidate violator is double-parked.
22. The system of claim 21 , wherein the multiple events are selected from a group consisting of:
hazard lights detected as operating on the candidate violator;
a detected object located in front of the candidate violator; and
a combination of the above.
23. The method of claim 9 , wherein the multiple events are selected from a group consisting of:
a detected object located in front of the candidate violator;
a tracked object located in front of the candidate violator; and
a combination of the above.
24. The method of claim 14 , wherein the multiple events are selected from a group consisting of:
hazard lights detected as operating on the candidate violator;
a tracked object moving around the candidate violator; and
a combination of the above.