Vehicle violation detection method and vehicle violation detection system

1 . A vehicle violation detection method, comprising:

obtaining a video clip comprising a plurality of consecutive frames, wherein the video clip is generated through photographing an intersection by an image capture device;

detecting a traffic sign object corresponding to a traffic sign and a license plate object corresponding to a license plate from each of the frames;

obtaining vehicle behavior information of each of the frames according to a sign position of the traffic sign object and a plate position of the license plate object in each of the frames; and

determining whether a vehicle violation event has occurred by conducting regression analysis on the vehicle behavior information of each of the frames,

wherein determining whether the vehicle violation event has occurred by conducting the regression analysis on the vehicle behavior information of each of the frames comprises:

conducting the regression analysis according to the vehicle behavior information of the frames to generate a coefficient of determination; and

determining whether the vehicle violation event has occurred according to the coefficient of determination of the regression analysis,

wherein obtaining the vehicle behavior information of each of the frames according to the sign position of the traffic sign object and the plate position of the license plate object in each of the frames comprises:

conducting depth estimation on each of the frame to obtain a depth map;

obtaining a first depth of the traffic sign object in each of the frames by using the depth map of each of the frames according to the sign position of the traffic sign object in each of the frames;

obtaining a second depth of the license plate object by using the depth map of each of the frames according to the plate position of the license plate object in each of the frames; and

obtaining a depth difference between the first depth of the traffic sign object and the second depth of the license plate object in each of the frames.

2 . The vehicle violation detection method according to claim 1 , wherein the regression analysis comprises establishing a linear regression model according to a plurality of depth differences of the frames, the depth differences of the frames comprise a first depth difference corresponding to an earlier reference time and a second depth difference corresponding to a later reference time, and determining whether the vehicle violation event has occurred according to the coefficient of determination of the regression analysis comprises:

in response to the coefficient of determination being greater than or equal to a first threshold value, the first depth difference being less than 0, and the second depth difference being greater than 0, determining that a red light running event occurs, wherein the red light running event is a vehicle proceeding straight through a red traffic sign.

3 . The vehicle violation detection method according to claim 1 , wherein obtaining the vehicle behavior information of each of the frames according to the sign position of the traffic sign object and the plate position of the license plate object in each of the frames comprises:

adjusting the plate position of the license plate object in each of the frames by using the sign position of the traffic sign object in each of the frames as an adjustment basis;

comparing plate positions of any two adjacent frames in the frames, and obtaining a movement trajectory vector associated with each of the frames; and

obtaining a movement angle of each of the frames according to the movement trajectory vector of each of the frames.

4 . The vehicle violation detection method according to claim 3 , wherein the regression analysis comprises establishing a linear regression model according to movement angles of the frames, and determining whether the vehicle violation event has occurred according to the coefficient of determination of the linear regression model comprises:

in response to the coefficient of determination being greater than or equal to a second threshold value, determining that a red light turning event occurs.

5 . The vehicle violation detection method according to claim 4 , wherein in response to the coefficient of determination being greater than or equal to a second threshold value, determining that the red light turning event occurs further comprises:

determining whether the red light turning event is a red light right turning event or a red light left turning event according to changing trends of the movement angles in the frames.

6 . The vehicle violation detection method according to claim 1 , wherein the coefficient of determination is R squared.

7 . The vehicle violation detection method according to claim 1 , wherein in each of the frames, a depth value of the traffic sign object corresponding to the traffic sign is less than a depth value of other traffic sign objects corresponding to other traffic signs.

8 . The vehicle violation detection method according to claim 1 , wherein the traffic sign comprises a traffic light, and the traffic light in the video clip is in a red light state.

9 . A vehicle violation detection system, comprising:

a storage circuit;

a processor, coupled to the storage circuit and configured to:

obtain a video clip comprising a plurality of consecutive frames, wherein the video clip is generated through photographing an intersection by an image capture device;

detect a traffic sign object corresponding to a traffic sign and a license plate object corresponding to a license plate from each of the frames;

obtain vehicle behavior information of each of the frames according to a sign position of the traffic sign object and a plate position of the license plate object in each of the frames; and

determine whether a vehicle violation event has occurred by conducting regression analysis on the vehicle behavior information of each of the frames,

wherein the processor is configured to:

conduct the regression analysis according to the vehicle behavior information of the frames to generate a coefficient of determination; and

determine whether the vehicle violation event has occurred according to the coefficient of determination of the regression analysis,

wherein the processor is configured to:

conduct depth estimation on each of the frame to obtain a depth map;

obtain a first depth of the traffic sign object in each of the frames by using the depth map of each of the frames according to the sign position of the traffic sign object in each of the frames;

obtain a second depth of the license plate object by using the depth map of each of the frames according to the plate position of the license plate object in each of the frames; and

obtain a depth difference between the first depth of the traffic sign object and the second depth of the license plate object in each of the frames.

10 . The vehicle violation detection system according to claim 9 , wherein the regression analysis comprises establishing a linear regression model according to a plurality of depth differences of the frames, the depth differences of the frames comprise a first depth difference corresponding to an earlier reference time and a second depth difference corresponding to a later reference time, the processor is configured to:

in response to the coefficient of determination being greater than or equal to a first threshold value, the first depth difference being less than 0, and the second depth difference being greater than 0, determine that a red light running event occurs, wherein the red light running event is a vehicle proceeding straight through a red traffic sign.

11 . The vehicle violation detection system according to claim 9 , wherein the processor is configured to:

adjust the plate position of the license plate object in each of the frames by using the sign position of the traffic sign object in each of the frames as an adjustment basis;

compare plate positions of any two adjacent frames in the frames, and obtain a movement trajectory vector associated with each of the frames; and

obtain a movement angle of each of the frames according to the movement trajectory vector of each of the frames.

12 . The vehicle violation detection system according to claim 11 , wherein the regression analysis comprises establishing a linear regression model according to movement angles of the frames, the processor is configured to:

in response to the coefficient of determination being greater than or equal to a second threshold value, determine that a red light turning event occurs.

13 . The vehicle violation detection system according to claim 12 , wherein the processor is configured to:

determine whether the red light turning event is a red light right turning event or a red light left turning event according to changing trends of the movement angles in the frames.

14 . The vehicle violation detection system according to claim 9 , wherein the coefficient of determination is R squared.

15 . The vehicle violation detection system according to claim 9 , wherein in each of the frames, a depth value of the traffic sign object corresponding to the traffic sign is less than a depth value of other traffic sign objects corresponding to other traffic signs.

16 . The vehicle violation detection system according to claim 9 , wherein the traffic sign comprises a traffic light, and the traffic light in the video clip is in a red light state.