A method for detecting an abnormal driving event

Abstract

A method for detecting abnormal driving events is provided, applicable to detecting an abnormal event occurring to a driver, and implemented using an in-vehicle driving monitoring device including an angle acquisition unit and a processing unit. The processing unit first converts an angle signal into a time-frequency graph related to the driver's head rotation angle, and then obtains a set of characteristic signals and a base frequency signal based on the time-frequency graph. The processing unit obtains at least one candidate time point corresponding to a frequency greater than a dynamic frequency threshold based on the set of characteristic signals, and obtains an extreme time point corresponding to the maximum angle value based on a base angle signal generated from the base frequency signal. The processing unit obtains the correct abnormal event based on the extreme time point and the at least one candidate time point.

Claims

1. A method for detecting abnormal driving events, applicable to detecting the occurrence of abnormal events of a driver, and implemented using an in-vehicle driving monitoring device, the in-vehicle driving monitoring device comprising an angle acquisition unit for acquiring an angle signal related to the driver's head rotation angle within a time interval, and a processing unit electrically connected to the angle acquisition unit, the method for detecting abnormal driving events comprising the following steps: (A) the processing unit converts the angle signal into a time-frequency graph related to the driver's head rotation angle using a time-frequency analysis; (B) the processing unit obtains a set of feature signals related to the driver's head rotation angle based on a relatively high-frequency signal region in the time-frequency graph; (C) the processing unit obtains at least one candidate time point corresponding to a frequency greater than a dynamic frequency threshold based on the set of feature signals; (D) the processing unit obtains a base frequency signal based on the time-frequency graph; (E) the processing unit converts the base frequency signal into a base angle signal using a signal conversion method. (F) The processing unit obtains an extreme time point corresponding to the maximum angle value based on the base angle signal; and (G) The processing unit obtains a period of occurrence related to the abnormal event based on the extreme time point and the at least one candidate time point, wherein, Step (G) includes the following sub-steps: (G-1) For each candidate time point, the processing unit subtracts the candidate time point from the extreme time point to obtain a time change value; (G-2) The processing unit obtains a target time point corresponding to the minimum time change value from the at least one time change value; and (G-3) The processing unit obtains the occurrence period based on the target time point and the extreme time point.

2. The method for detecting abnormal driving events as described in claim 1, wherein, Step (D) includes the following sub-steps: (D-1) The processing unit obtains at least one candidate signal component corresponding to a time-frequency window greater than a time-frequency window threshold value from the time-frequency diagram; and (D-2) The processing unit obtains the base frequency signal based on the at least one candidate signal component.

3. The method for detecting abnormal driving events as described in claim 1, wherein, The angle acquisition unit includes an image capturing module for acquiring multiple surveillance images related to the driver's head within a time interval, and a processing module connected to the image capturing module for converting the surveillance images into the angle signal. Before step (A), the unit further includes the following steps: (H) For each surveillance image, the processing module obtains a rotation angle value related to the driver's head rotation based on the head portion of the surveillance image; (I) The processing module obtains an initial angle signal within the time interval based on all rotation angle values; and (J) The processing module obtains the angle signal based on the initial angle signal using a correction method.

4. The method for detecting abnormal driving events as described in claim 3, wherein, Step (H) further includes the following sub-steps: (H-1) The processing module uses a feature algorithm to obtain multiple head feature positions of the head portion of the surveillance image; and (H-2) The processing module obtains the rotation angle value related to the rotation of the driver's head based on the head feature positions.

5. The method for detecting abnormal driving events as described in claim 1, wherein, In step (A), the processing unit uses a wavelet transform method to convert the angle signal into the time-frequency diagram. In step (E), the processing unit uses an inverse wavelet transform method to convert the base frequency signal into the base angle signal.

6. The method for detecting abnormal driving events as described in claim 1, wherein, In step (C), the at least one candidate time point is multiple candidate time points, and step (G) further includes the following sub-steps: (G-1) The processing unit obtains two target time points from the candidate time points that cover the extreme time points and have the shortest corresponding time intervals; and (G-2) The processing unit obtains the occurrence period based on the target time points.

Images