A method and device for calibrating a driver's driving state

Abstract

The embodiment of the present invention discloses a method and device for calibrating a driver's driving state, wherein the method comprises: determining whether the speed of a target vehicle is greater than or equal to a target speed threshold; if so, obtaining a target face image corresponding to a target driver of a preset number of frames, and inputting a target neural network model for prediction, obtaining target head horizontal angle data, target head pitch angle data, target sight direction horizontal angle data, and target sight direction pitch angle data; respectively obtaining target head horizontal angle statistical results, target head pitch angle statistical results, target sight direction horizontal angle statistical results, and target sight direction pitch angle statistical results, and determining whether they all meet the target preset statistical requirements; if so, determining that this calibration is successful. The present invention can complete calibration without manual operation, and when the driver is changed or the equipment installation displacement changes, the calibration data is corrected by automatic calibration to improve the accuracy of detection.

Description

Technical Field

[0001] The present invention relates to the technical field of safe driving of automobile drivers, and in particular to a method and device for calibrating a driver's driving state. Background Art

[0002] With the rapid development of society, driving has become the preferred mode of travel for everyone. Due to the large number of vehicles, while it brings convenience to everyone, it also leads to an increase in traffic accidents year by year. Dangerous driving behavior is the main cause of traffic accidents, and dangerous driving behavior mainly includes fatigue detection, distracted driving, making phone calls and smoking. In order to reduce the occurrence of dangerous driving behavior, my country requires relevant vehicles to be equipped with safety auxiliary driving products, including driver monitoring systems. The driver monitoring system captures relevant information about the driver's face through an infrared camera to determine whether the driver is in dangerous driving behavior.

[0003] In actual work, the driver monitoring system often needs to know the normal driving status of the driver, including the driver's head posture facing forward and the driver's line of sight direction, and use it as the basis for judging dangerous driving behavior. At present, the driver monitoring system often obtains this information through manual calibration. For example, after installing the driver monitoring camera, the driver is required to look forward and perform manual calibration to obtain the driver's head posture facing forward and the driver's line of sight direction. This requires that each device needs to be manually calibrated, resulting in a large amount of workload. In addition, there is often only one manually calibrated driver object. When changing different drivers, their head posture facing forward and line of sight direction will change, which is different from the manually calibrated information, resulting in a decrease in detection accuracy. Summary of the invention

[0004] Based on this, it is necessary to propose a driver driving status calibration method and device to address the above-mentioned problems, so as to solve the following problems in the prior art: Currently, the driver monitoring system often calibrates the driver status by manual calibration, and there is often only one manually calibrated driver object. When a different driver is replaced, his head posture and line of sight when facing forward will change, which is different from the manually calibrated information, resulting in a decrease in detection accuracy.

[0005] The first technical solution of the embodiment of the present invention is:

[0006] A method for calibrating a driver's driving state is provided, wherein the method is based on a driver monitoring system and comprises:

[0007] During the process of the target driver driving the target vehicle, it is determined whether the target speed corresponding to the target vehicle is greater than or equal to the target speed threshold value at every preset target time period; if so, the target driver state calibration program is entered, and the target facial image corresponding to the target driver for a preset number of target frames is obtained; the target facial image for the preset number of target frames is input into the target neural network model for prediction, and the target head posture data and target sight line direction data corresponding to the target driver for the preset number of target frames are obtained, wherein the target head posture data includes target head horizontal angle data and target head pitch angle data, and the target sight line direction data includes target sight line direction horizontal angle data and target head pitch angle data. data and target sight direction pitch angle data; respectively perform statistical processing on the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data and the target sight direction pitch angle data to obtain the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results; determine whether the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results all meet the target preset statistical requirements; if so, determine that the calibration of the driver's driving state is successful.

[0008] The second technical solution of the embodiment of the present invention is:

[0009] A device for calibrating a driver's driving state, which performs calibration based on a driver monitoring system, comprises:

[0010] The vehicle speed detection module is used to determine whether the target speed corresponding to the target vehicle is greater than or equal to the target speed threshold at every preset target time period when the target driver is driving the target vehicle; the face image acquisition module is used to enter the target driver state calibration program when the target speed corresponding to the target vehicle is greater than or equal to the target speed threshold, and obtain the target face image corresponding to the target driver for a preset number of frames; the face image prediction module is used to input the target face image for the preset number of frames into the target neural network model for prediction, and obtain the target head posture data and target sight direction data corresponding to the target driver for the preset number of frames, wherein the target head posture data includes target head horizontal angle data and target head pitch angle data, and the target sight direction data includes target sight direction horizontal angle data and target sight direction pitch angle data; the driving data statistics module , used to perform statistical processing on the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data and the target sight direction pitch angle data respectively, to obtain the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results; the statistical result judgment module is used to judge whether the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results all meet the target preset statistical requirements; the statistical result calibration module is used to judge that the calibration of the driver's driving state is successful when the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results all meet the target preset statistical requirements.

[0011] The third technical solution of the embodiment of the present invention is:

[0012] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the following steps:

[0013] During the process of the target driver driving the target vehicle, it is determined whether the target speed corresponding to the target vehicle is greater than or equal to the target speed threshold value at every preset target time period; if so, the target driver state calibration program is entered, and the target facial image corresponding to the target driver for a preset number of target frames is obtained; the target facial image for the preset number of target frames is input into the target neural network model for prediction, and the target head posture data and target sight line direction data corresponding to the target driver for the preset number of target frames are obtained, wherein the target head posture data includes target head horizontal angle data and target head pitch angle data, and the target sight line direction data includes target sight line direction horizontal angle data and target head pitch angle data. data and target sight direction pitch angle data; respectively perform statistical processing on the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data and the target sight direction pitch angle data to obtain the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results; determine whether the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results all meet the target preset statistical requirements; if so, determine that the calibration of the driver's driving state is successful.

[0014] The fourth technical solution of the embodiment of the present invention is:

[0015] A computer-readable storage medium stores a computer program, which, when executed by a processor, causes the processor to perform the following steps:

[0016] During the process of the target driver driving the target vehicle, it is determined whether the target speed corresponding to the target vehicle is greater than or equal to the target speed threshold value at every preset target time period; if so, the target driver state calibration program is entered, and the target facial image corresponding to the target driver for a preset number of target frames is obtained; the target facial image for the preset number of target frames is input into the target neural network model for prediction, and the target head posture data and target sight line direction data corresponding to the target driver for the preset number of target frames are obtained, wherein the target head posture data includes target head horizontal angle data and target head pitch angle data, and the target sight line direction data includes target sight line direction horizontal angle data and target head pitch angle data. data and target sight direction pitch angle data; respectively perform statistical processing on the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data and the target sight direction pitch angle data to obtain the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results; determine whether the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results all meet the target preset statistical requirements; if so, determine that the calibration of the driver's driving state is successful.

[0017] The embodiments of the present invention have the following beneficial effects:

[0018] The present invention first determines whether the speed of a target vehicle is greater than or equal to a target speed threshold value. If so, a target face image corresponding to a target driver of a preset number of frames is obtained, the target face image is predicted, and target head horizontal angle data, target head pitch angle data, target sight direction horizontal angle data, and target sight direction pitch angle data are obtained. Then, statistical results of the target head horizontal angle, target head pitch angle, target sight direction horizontal angle, and target sight direction pitch angle are respectively obtained. Finally, it is determined whether the statistical results of the target head horizontal angle, the target head pitch angle, the target sight direction horizontal angle, and the target sight direction pitch angle all meet the target preset statistical requirements. If so, it is determined that this calibration is successful. The present invention can complete the calibration operation during the normal driving of the vehicle, obtain the calibration data of the driver's status, require less manual operation, and can also solve the problem of inaccurate alarm caused by incorrect manual calibration, and can also solve the problem of inaccurate alarm caused by moving the camera position after manual calibration; secondly, when the driver is changed, the present invention can automatically correct the calibration data to improve the reliability of the calibration data; secondly, the present invention performs statistical analysis based on normal distribution, eliminates the interference of singular values, and improves the accuracy of the calibration data; finally, the present invention can ensure that the calibration data conforms to the situation where the driver drives normally and faces the front by statistically analyzing the standard deviation of multiple calibration data. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] in:

[0021] Figure 1 It is an implementation flow chart of an implementation method of a driver's driving state calibration method in an embodiment;

[0022] Figure 2 It is a framework structure diagram of an implementation method of a device for calibrating a driver's driving state in one embodiment;

[0023] Figure 3 It is a structural block diagram of an implementation scheme of a computer device in one embodiment. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] See also Figure 1 , combined with Figure 1 It can be seen that a method for calibrating a driver's driving state according to an embodiment of the present invention is based on a driver monitoring system and includes the following steps:

[0026] Step S101: When a target driver is driving a target vehicle, it is determined whether a target speed corresponding to the target vehicle is greater than or equal to a target speed threshold at every predetermined target time period.

[0027] The target speed threshold is an empirical value, which can be set to 40km / h but is not limited to this speed. The pre-target setting time is preferably in milliseconds, which is equivalent to real-time monitoring and judgment. When the target speed is greater than or equal to the target speed threshold, the vehicle is prone to safety accidents and its driving status needs to be calibrated.

[0028] Step S102: If the target vehicle speed corresponding to the target vehicle is greater than or equal to the target vehicle speed threshold, the target driver state calibration program is entered, and a target facial image corresponding to the target driver of a preset number of frames is obtained.

[0029] Among them, the target preset number of frames is preferably 200 consecutive frames, and the target facial image corresponding to the target driver includes or covers the driver's head posture image and line of sight direction image.

[0030] Step S103: input the target facial image of the target preset number of frames into the target neural network model for prediction, and obtain the target head posture data and target sight direction data corresponding to the target driver of the target preset number of frames, wherein the target head posture data includes target head horizontal angle data and target head pitch angle data, and the target sight direction data includes target sight direction horizontal angle data and target sight direction pitch angle data.

[0031] Among them, the target neural network model is a deep learning model generated after training. As long as the target face image is input into the target neural network model for prediction, the target head posture data and target line of sight direction data corresponding to the target driver for the preset number of frames of the target can be output.

[0032] Among them, the driver monitoring system can obtain face images in real time, and the driver’s real-time head posture can be detected through the target neural network model [p y ,p p ] and sight direction [g y ,g p ], where p y and p p are the horizontal angle and pitch angle of the head posture, g y and g p The horizontal angle and pitch angle of the sight direction are obtained by recording the driver's head posture and sight direction for consecutive frames (for example, 200 frames). Head posture pitch angle Eye level angle and line of sight elevation angle And so on 4 sets of data.

[0033] Step S104: performing statistical processing on the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data and the target sight direction pitch angle data respectively to obtain the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results.

[0034] Among them, statistical processing of the target head horizontal angle data, the target head pitch angle data, the target line of sight direction horizontal angle data and the target line of sight direction pitch angle data requires multiple steps to respectively count the change degrees of the target head horizontal angle data, the target head pitch angle data, the target line of sight direction horizontal angle data and the target line of sight direction pitch angle data, thereby determining whether the calibration of the driver's driving state is successful.

[0035] Step S105: Determine whether the target head horizontal angle statistics, the target head elevation angle statistics, the target sight direction horizontal angle statistics, and the target sight direction elevation angle statistics all meet the target preset statistical requirements.

[0036] Among them, this step requires that the target head horizontal angle statistical results, the target head pitch angle statistical results, the target line of sight horizontal angle statistical results and the target line of sight pitch angle statistical results all meet the target preset statistical requirements. As long as one of them does not meet the target preset statistical requirements, the calibration of the driver's driving status this time will fail.

[0037] Step S106: If the target head horizontal angle statistics, the target head pitch angle statistics, the target sight direction horizontal angle statistics and the target sight direction pitch angle statistics all meet the target preset statistical requirements, it is determined that the calibration of the driver's driving state is successful.

[0038] In this embodiment, optionally, the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data, and the target sight direction pitch angle data are statistically processed respectively to obtain the target head horizontal angle statistical result, the target head pitch angle statistical result, the target sight direction horizontal angle statistical result, and the target sight direction pitch angle statistical result, including:

[0039] First, respectively calculate the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data and the target sight direction pitch angle data corresponding to the target head horizontal angle first standard deviation, the target head pitch angle first standard deviation, the target sight direction horizontal angle first standard deviation and the target sight direction pitch angle first standard deviation.

[0040] Specifically, this step takes the head posture horizontal angle as an example, and the calculation steps of the remaining data are the same as the head posture horizontal angle. The detailed steps are as follows: First, The array statistics mean0 and standard deviation std0 are shown in formula (1) and formula (2).

[0041]

[0042]

[0043] Wherein, T is a positive integer, and K is a natural number greater than or equal to 0.

[0044] Second, based on the first standard deviation of the target head horizontal angle, the first standard deviation of the target head pitch angle, the first standard deviation of the target sight direction horizontal angle and the first standard deviation of the target sight direction pitch angle, respectively calculate the target head horizontal angle normalized value group, the target head pitch angle normalized value group, the target sight direction horizontal angle normalized value group and the target sight direction pitch angle normalized value group corresponding to the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data and the target sight direction pitch angle data.

[0045] Among them, this step is Normalize to get a normalized value group As shown in formula (3):

[0046]

[0047] Third, the target singular values ​​corresponding to the target head horizontal angle normalized value group, the target head pitch angle normalized value group, the target line of sight horizontal angle normalized value group and the target line of sight pitch angle normalized value group are calculated respectively, and the absolute value of the target singular value is greater than the target preset value.

[0048] In this step, the horizontal angle of head posture is used as an example. During driving, the driver may suddenly turn his head or look to the side, which may lead to singular values. Therefore, statistics need to remove these singular values. It can be approximately regarded as a normal distribution N(0,1) with a mean of 0 and a standard deviation of 1. Assuming that x obeys the normal distribution function N(0,1), then

[0049]

[0050] Formula (4) is a function that obeys the normal distribution N(0,1), and the singular value is determined according to the percentage point function (PPF) of the normal distribution. The percentage point function is the inverse function of the cumulative distribution function (CDF) of the normal distribution. The cumulative distribution function formula is as follows:

[0051]

[0052] This step uses the horizontal coordinate of the normal distribution function to determine the singular value. Formula (5) is the cumulative distribution function of the normal distribution function. The cumulative distribution function is defined as the area under the curve of the normal distribution function N(0,1) from negative infinity to x. The percent point function is the inverse function of the cumulative distribution function. For example, PPF(0.95)=1.96 means that the area under the curve of the normal distribution function N(0,1) from negative infinity to 1.96 (horizontal coordinate) is 0.95 (the area is used as the independent variable, and the dependent variable is the horizontal coordinate). Therefore, this step uses this as the threshold to determine the singular value in the normalized horizontal angle. The independent variable of the CDF function is the horizontal coordinate, and the dependent variable is the area. The PPF function is the inverse function of the CDF function, so the area is the independent variable and the horizontal coordinate is the dependent variable.

[0053] In this step, the CDF function represents the area under the normal distribution N(0,1) curve from negative infinity to x, so PPF(x)=y means that the area under the normal distribution N(0,1) curve from negative infinity to y is x. The singular values ​​of are judged as follows

[0054]

[0055] Where PPF(0.95)=1.96, which satisfies formula (6) It is determined to be a singular value, that is, a value with an absolute value greater than 1.96 is a singular value, and the target preset value of this step is 1.96.

[0056] Fourth, all the target singular values ​​in the target head horizontal angle normalized value group, the target head pitch angle normalized value group, the target line of sight direction horizontal angle normalized value group and the target line of sight direction pitch angle normalized value group are removed respectively to obtain the target head horizontal angle normalized standard data group, the target head pitch angle normalized standard data group, the target line of sight direction horizontal angle normalized standard data group and the target line of sight direction pitch angle normalized standard data group.

[0057] In this step, the horizontal angle of the head posture is taken as an example. After removing all singular values, the data is obtained.

[0058] Fifth, respectively calculate the second standard deviation of the target head horizontal angle, the second standard deviation of the target head pitch angle, the second standard deviation of the target line of sight direction horizontal angle and the second standard deviation of the target line of sight direction pitch angle corresponding to the normalized standard data group of the target head horizontal angle, the normalized standard data group of the target head pitch angle, the normalized standard data group of the target line of sight direction horizontal angle and the normalized standard data group of the target line of sight direction pitch angle.

[0059] Among them, the second standard deviation of the target head horizontal angle, the second standard deviation of the target head pitch angle, the second standard deviation of the target line of sight horizontal angle and the second standard deviation of the target line of sight pitch angle are respectively the statistical results of the target head horizontal angle, the statistical results of the target head pitch angle, the statistical results of the target line of sight horizontal angle and the statistical results of the target line of sight pitch angle.

[0060] In this step, the horizontal angle of head posture is taken as an example, and the statistics are calculated again using formulas (1) and (2) The mean and standard deviation are obtained as mean1 and std1.

[0061] In this embodiment, optionally, the judging whether the target head horizontal angle statistics, the target head pitch angle statistics, the target sight direction horizontal angle statistics, and the target sight direction pitch angle statistics all meet the target preset statistical requirements includes:

[0062] First, determine whether at least one of the second standard deviation of the target head horizontal angle, the second standard deviation of the target head pitch angle, the second standard deviation of the target line of sight horizontal angle, and the second standard deviation of the target line of sight pitch angle is greater than the target preset standard deviation threshold.

[0063] Second, if at least one of the second standard deviation of the target head horizontal angle, the second standard deviation of the target head pitch angle, the second standard deviation of the target line of sight horizontal angle and the second standard deviation of the target line of sight pitch angle is greater than the target preset standard deviation threshold, then it is determined that the calibration of the target driver status has failed, and a new round of calibration of the driver's driving status is performed again.

[0064] Among them, this step takes the horizontal angle of head posture as an example. If the standard deviation std1 after deleting the singular value is greater than the target preset standard deviation threshold, it means that the calibration status of the horizontal angle of head posture is failed, and the statistical analysis ends.

[0065] In this embodiment, optionally, the determining whether at least one of the second standard deviation of the target head horizontal angle, the second standard deviation of the target head pitch angle, the second standard deviation of the target sight direction horizontal angle, and the second standard deviation of the target sight direction pitch angle is greater than a target preset standard deviation threshold comprises:

[0066] First, if the second standard deviation of the target head horizontal angle, the second standard deviation of the target head pitch angle, the second standard deviation of the target line of sight direction horizontal angle and the second standard deviation of the target line of sight direction pitch angle are all less than or equal to the target preset standard deviation threshold, then the normalized standard data group of the target head horizontal angle, the normalized standard data group of the target head pitch angle, the normalized standard data group of the target line of sight direction horizontal angle and the normalized standard data group of the target line of sight direction pitch angle are respectively performed histogram statistics to obtain the target head horizontal angle histogram distribution function data, the target head pitch angle histogram distribution function data, the target line of sight direction horizontal angle histogram distribution function data and the target line of sight direction pitch angle histogram distribution function data.

[0067] Second, respectively calculate the target head horizontal angle KL divergence, target head pitch angle KL divergence, target line of sight direction horizontal angle KL divergence and target line of sight direction pitch angle KL divergence between the four groups of data, namely, the target head horizontal angle histogram distribution function data, the target head pitch angle histogram distribution function data, the target line of sight direction horizontal angle histogram distribution function data and the target line of sight direction pitch angle histogram distribution function data and the preset target normal distribution function data.

[0068] Third, determine whether at least one of the target head horizontal angle KL divergence, the target head pitch angle KL divergence, the target line of sight horizontal angle KL divergence, and the target line of sight pitch angle KL divergence is greater than a preset KL divergence.

[0069] Fourth, if at least one of the target head horizontal angle KL divergence, the target head pitch angle KL divergence, the target line of sight horizontal angle KL divergence and the target line of sight pitch angle KL divergence is greater than the preset KL divergence, it is determined that the calibration of the target driver state has failed, and a new round of calibration of the driver's driving state is performed again.

[0070] In this step, the horizontal angle of head posture is taken as an example. If the standard deviation std1 after deleting the singular value is less than the target preset standard deviation threshold (the calibration is not successful at this time), the data is Perform histogram statistics to obtain the distribution function p(x), then calculate the distribution function p(x) and the normal distribution function f n The KL divergence of (x) is shown in formulas (7) and (8):

[0071]

[0072]

[0073] If D KL (p‖f n ) is less than the preset divergence threshold, indicating that the data Close to normal distribution, the head posture horizontal angle calibration is successful, and the maximum value of p(x) corresponds to x as the calibration angle of the head posture horizontal angle If D KL (p‖f n ) is greater than the preset divergence threshold, it means that the head posture horizontal angle calibration fails.

[0074] The calibration is successful when the target head horizontal angle statistics, the target head pitch angle statistics, the target line of sight horizontal angle statistics and the target line of sight pitch angle statistics all meet the target preset statistical requirements.

[0075] In this embodiment, optionally, the determining whether at least one of the target head horizontal angle KL divergence, the target head pitch angle KL divergence, the target sight direction horizontal angle KL divergence, and the target sight direction pitch angle KL divergence is greater than a preset KL divergence comprises:

[0076] First, if the target head horizontal angle KL divergence, the target head pitch angle KL divergence, the target line of sight horizontal angle KL divergence and the target line of sight pitch angle KL divergence are all less than or equal to the preset KL divergence, it is determined that the calibration of the target driver status is successful.

[0077] Second, respectively calculate the maximum function value of the target head horizontal angle, the maximum function value of the target head pitch angle, the maximum function value of the target line of sight direction horizontal angle and the maximum function value of the target line of sight direction pitch angle corresponding to the target head horizontal angle histogram distribution function data, the target head pitch angle histogram distribution function data, the target line of sight direction horizontal angle histogram distribution function data and the target line of sight direction pitch angle histogram distribution function data.

[0078] Third, respectively obtain the target head horizontal angle horizontal coordinate value, the target head pitch angle horizontal coordinate value, the target line of sight direction horizontal angle horizontal coordinate value and the target line of sight direction pitch angle horizontal coordinate value corresponding to the maximum function value of the target head horizontal angle, the maximum function value of the target head pitch angle, the maximum function value of the target line of sight direction horizontal angle and the maximum function value of the target line of sight direction pitch angle.

[0079] Fourth, the horizontal angle abscissa value of the target head, the pitch angle abscissa value of the target head, the horizontal angle abscissa value of the target sight direction, and the pitch angle abscissa value of the target sight direction are respectively used as the target head horizontal angle calibration angle, the target head pitch angle calibration angle, the target sight direction horizontal angle calibration angle, and the target sight direction pitch angle calibration angle.

[0080] Fifth, the number of successful calibrations of the driver's driving status is increased by one.

[0081] Among them, represents the target head horizontal angle calibration angle, the target head pitch angle calibration angle, the target sight direction horizontal angle calibration angle and the target sight direction pitch angle calibration angle, where is the target head horizontal angle calibration angle, is the target head pitch calibration angle, is the horizontal angle calibration angle of the target sight direction, Calibrate the pitch angle of the target sight direction; if the calibration is successful, the number of successful calibrations is increased by one, and the calibration data is recorded; if the calibration fails, return to recalibrate.

[0082] In this embodiment, optionally, the number of successful calibrations of the driver's driving state is increased by one, and then includes:

[0083] First, determine whether the number of successful calibrations of the target driver status is greater than or equal to a preset successful calibration number threshold.

[0084] Second, if the number of successful calibrations of the target driver status is greater than or equal to a preset threshold value of successful calibrations, then the N consecutive target calibration data that are ranked in front in time (i.e., the latest) are calculated respectively, and the N target calibration data include N head horizontal angle calibration angles, N head pitch angle calibration angles, N line of sight horizontal angle calibration angles, and N line of sight pitch angle calibration angles, where N is a positive integer greater than or equal to the preset threshold value of successful calibrations.

[0085] Third, respectively calculate the target head horizontal angle standard deviation, target head pitch angle standard deviation, target sight direction horizontal angle standard deviation and target sight direction pitch angle standard deviation corresponding to the N head horizontal angle calibration angles, the N head pitch angle calibration angles, the N sight direction horizontal angle calibration angles and the N sight direction pitch angle calibration angles. Wherein, N can be selected as 3 but is not limited to 3.

[0086] Fourth, determine whether the target head horizontal angle standard deviation, the target head pitch angle standard deviation, the target sight direction horizontal angle standard deviation and the target sight direction pitch angle standard deviation are all less than the target preset calibration angle standard deviation threshold.

[0087] Fifth, if the standard deviation of the target head horizontal angle, the standard deviation of the target head pitch angle, the standard deviation of the target line of sight horizontal angle and the standard deviation of the target line of sight pitch angle are all smaller than the target preset calibration angle standard deviation threshold, then it is determined that the calibration of the driver's driving state is successful, and the number of successful calibrations of the driver's driving state is increased by one; otherwise, it is determined that the calibration of the target driver's state has failed, and a new round of calibration of the driver's driving state is performed again.

[0088] Among them, it is possible that the driver has been staring at the rearview mirror for a period of time. At this time, the calibration is successful, but it is not the correct posture to face the front, so return to recalibrate. In this step, if the number of successful calibrations is greater than the preset threshold of successful calibrations, the standard deviation of the last three consecutive calibration data is calculated. The specific formula is as follows:

[0089]

[0090]

[0091]

[0092]

[0093]

[0094]

[0095]

[0096]

[0097] Among them, mean py is the mean of the horizontal calibration angles of the head posture, std py is the standard deviation of the horizontal calibration angle of the head posture, mean pp is the mean of the head posture pitch calibration angle, std pp is the standard deviation of the head posture pitch calibration angle, mean gy is the mean value of the horizontal calibration angle of the line of sight, std gy is the standard deviation of the horizontal calibration angle of sight, mean gp is the mean of the sight pitch calibration angles, std gp The standard deviation of the line of sight pitch calibration angle.

[0098] In this embodiment, optionally, the determining that the calibration of the driver's driving state is successful and adding one to the number of successful calibrations of the driver's driving state may then include:

[0099] First, determine whether the calibration of the driver's driving state has been successful before this calibration.

[0100] Second, if the calibration of the driver's driving state has been successful before this calibration, it is determined whether there is an abnormality between the calibration result data of this time and the calibration result data of the previous time.

[0101] Third, if there is an abnormality between the calibration result data of this time and the calibration result data of the previous time, the calibration result data of this time will replace the calibration result data of the previous time, otherwise the calibration result data of this time will be used as the calibration result data of the current driving state of the driver.

[0102] See also Figure 2 , combined with Figure 2 It can be obtained that a driver driving state calibration device 100 according to an embodiment of the present invention is calibrated based on a driver monitoring system, and includes:

[0103] The vehicle speed detection module 10 is used to determine whether the target vehicle speed corresponding to the target vehicle is greater than or equal to the target vehicle speed threshold at every predetermined target time period during the process of the target driver driving the target vehicle;

[0104] The face image acquisition module 20 is used to enter the target driver state calibration program when the target speed corresponding to the target vehicle is greater than or equal to the target speed threshold, and obtain the target face image corresponding to the target driver for a preset number of frames.

[0105] The face image prediction module 30 is used to input the target face image of the target preset number of frames into the target neural network model for prediction, and obtain the target head posture data and target sight direction data corresponding to the target driver of the target preset number of frames, wherein the target head posture data includes target head horizontal angle data and target head pitch angle data, and the target sight direction data includes target sight direction horizontal angle data and target sight direction pitch angle data.

[0106] The driving data statistics module 40 is used to perform statistical processing on the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data and the target sight direction pitch angle data, respectively, to obtain the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results.

[0107] The statistical result judgment module 50 is used to judge whether the target head horizontal angle statistical result, the target head elevation angle statistical result, the target sight direction horizontal angle statistical result and the target sight direction elevation angle statistical result all meet the target preset statistical requirements.

[0108] The statistical result calibration module 60 is used to determine that the calibration of the driver's driving state is successful when the statistical results of the target head horizontal angle, the target head pitch angle, the target sight direction horizontal angle and the target sight direction pitch angle all meet the target preset statistical requirements.

[0109] Figure 3 FIG. 1 shows an internal structure diagram of a computer device in an embodiment. The computer device may be a terminal or a server. Figure 3 As shown, the computer device includes a processor, a memory and a network interface connected via a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor may implement the above-mentioned method for calibrating the driver's driving state. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor may implement the above-mentioned method for calibrating the driver's driving state. Those skilled in the art will understand that Figure 3The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0110] In another embodiment, a computer device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the following steps:

[0111] During the process of the target driver driving the target vehicle, it is determined whether the target speed corresponding to the target vehicle is greater than or equal to the target speed threshold value at every preset target time period; if so, the target driver state calibration program is entered, and the target facial image corresponding to the target driver for a preset number of target frames is obtained; the target facial image for the preset number of target frames is input into the target neural network model for prediction, and the target head posture data and target sight line direction data corresponding to the target driver for the preset number of target frames are obtained, wherein the target head posture data includes target head horizontal angle data and target head pitch angle data, and the target sight line direction data includes target sight line direction horizontal angle data and target head pitch angle data. data and target sight direction pitch angle data; respectively perform statistical processing on the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data and the target sight direction pitch angle data to obtain the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results; determine whether the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results all meet the target preset statistical requirements; if so, determine that the calibration of the driver's driving state is successful.

[0112] In another embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the processor performs the following steps:

[0113] During the process of the target driver driving the target vehicle, it is determined whether the target speed corresponding to the target vehicle is greater than or equal to the target speed threshold value at every preset target time period; if so, the target driver state calibration program is entered, and the target facial image corresponding to the target driver for a preset number of target frames is obtained; the target facial image for the preset number of target frames is input into the target neural network model for prediction, and the target head posture data and target sight line direction data corresponding to the target driver for the preset number of target frames are obtained, wherein the target head posture data includes target head horizontal angle data and target head pitch angle data, and the target sight line direction data includes target sight line direction horizontal angle data and target head pitch angle data. data and target sight direction pitch angle data; respectively perform statistical processing on the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data and the target sight direction pitch angle data to obtain the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results; determine whether the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results all meet the target preset statistical requirements; if so, determine that the calibration of the driver's driving state is successful.

[0114] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0115] The embodiment of the present invention first determines whether the speed of the target vehicle is greater than or equal to the target speed threshold. If so, the target face image corresponding to the target driver of a preset number of frames is obtained, the target face image is predicted, and the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data and the target sight direction pitch angle data are obtained. Then, the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results are respectively obtained. Finally, it is determined whether the target head horizontal angle statistical results, the target head pitch angle statistical results, the target sight direction horizontal angle statistical results and the target sight direction pitch angle statistical results all meet the target preset statistical requirements. If so, it is determined that this calibration is successful. The present invention can complete the calibration operation during the normal driving of the vehicle, obtain the calibration data of the driver's status, require less manual operation, and can also solve the problem of inaccurate alarm caused by incorrect manual calibration, and can also solve the problem of inaccurate alarm caused by moving the camera position after manual calibration; secondly, when the driver is changed, the present invention can automatically correct the calibration data to improve the reliability of the calibration data; secondly, the present invention performs statistical analysis based on normal distribution, eliminates the interference of singular values, and improves the accuracy of the calibration data; finally, the present invention can ensure that the calibration data conforms to the situation where the driver drives normally and faces the front by statistically analyzing the standard deviation of multiple calibration data.

[0116] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0117] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for calibrating a driver's driving state, characterized in that: Calibration based on the driver monitoring system, which includes: During the process of the target driver driving the target vehicle, determining whether the target vehicle speed corresponding to the target vehicle is greater than or equal to the target vehicle speed threshold at every predetermined target time period; If yes, then enter the target driver state calibration program and obtain the target face image corresponding to the target driver for a target preset number of frames; Inputting the target face images of the target preset number of frames into the target neural network model for prediction, obtaining target head posture data and target sight line direction data corresponding to the target driver of the target preset number of frames, wherein the target head posture data includes target head horizontal angle data and target head pitch angle data, and the target sight line direction data includes target sight line direction horizontal angle data and target sight line direction pitch angle data; Respectively statistically process the target head horizontal angle data, the target head pitch angle data, the target sight line direction horizontal angle data, and the target sight line direction pitch angle data to obtain target head horizontal angle statistical results, target head pitch angle statistical results, target sight line direction horizontal angle statistical results, and target sight line direction pitch angle statistical results; Determine whether the target head horizontal angle statistics, the target head pitch angle statistics, the target sight direction horizontal angle statistics, and the target sight direction pitch angle statistics all meet the target preset statistical requirements; If yes, it is determined that the calibration of the driver's driving state is successful; The statistical processing is performed on the target head horizontal angle data, the target head pitch angle data, the target sight line direction horizontal angle data and the target sight line direction pitch angle data respectively to obtain the target head horizontal angle statistical result, the target head pitch angle statistical result, the target sight line direction horizontal angle statistical result and the target sight line direction pitch angle statistical result, including: respectively calculating the target head horizontal angle first standard deviation, the target head pitch angle first standard deviation, the target sight line direction horizontal angle first standard deviation and the target sight line direction pitch angle first standard deviation corresponding to the target head horizontal angle data, the target head pitch angle data, the target sight line direction horizontal angle data and the target sight line direction pitch angle data; According to the first standard deviation of the target head horizontal angle, the first standard deviation of the target head pitch angle, the first standard deviation of the target sight direction horizontal angle, and the first standard deviation of the target sight direction pitch angle, respectively calculate a target head horizontal angle normalized value group, a target head pitch angle normalized value group, a target sight direction horizontal angle normalized value group, and a target sight direction pitch angle normalized value group corresponding to the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data, and the target sight direction pitch angle data; Calculate target singular values ​​corresponding to the target head horizontal angle normalized value group, the target head elevation angle normalized value group, the target sight direction horizontal angle normalized value group, and the target sight direction elevation angle normalized value group respectively, and the absolute value of the target singular value is greater than the target preset value; Respectively removing all the target singular values ​​in the target head horizontal angle normalized value group, the target head pitch angle normalized value group, the target sight direction horizontal angle normalized value group, and the target sight direction pitch angle normalized value group, to obtain a target head horizontal angle normalized standard data group, a target head pitch angle normalized standard data group, a target sight direction horizontal angle normalized standard data group, and a target sight direction pitch angle normalized standard data group; Respectively calculating the target head horizontal angle second standard deviation, the target head pitch angle second standard deviation, the target sight direction horizontal angle second standard deviation and the target sight direction pitch angle second standard deviation corresponding to the target head horizontal angle normalized standard data group, the target head pitch angle normalized standard data group, the target sight direction horizontal angle normalized standard data group and the target sight direction pitch angle normalized standard data group; The second standard deviation of the target head horizontal angle, the second standard deviation of the target head pitch angle, the second standard deviation of the target sight direction horizontal angle and the second standard deviation of the target sight direction pitch angle are respectively the target head horizontal angle statistics, the target head pitch angle statistics, the target sight direction horizontal angle statistics and the target sight direction pitch angle statistics; The determining whether the target head horizontal angle statistical result, the target head pitch angle statistical result, the target sight direction horizontal angle statistical result and the target sight direction pitch angle statistical result all meet the target preset statistical requirements, including: determining whether at least one of the target head horizontal angle second standard deviation, the target head pitch angle second standard deviation, the target sight direction horizontal angle second standard deviation and the target sight direction pitch angle second standard deviation is greater than the target preset standard deviation threshold; If yes, it is determined that the calibration of the target driver state has failed, and a new round of calibration of the driver's driving state is performed again; The step of judging whether at least one of the second standard deviation of the target head horizontal angle, the second standard deviation of the target head pitch angle, the second standard deviation of the target sight direction horizontal angle and the second standard deviation of the target sight direction pitch angle is greater than a target preset standard deviation threshold value comprises: if the second standard deviation of the target head horizontal angle, the second standard deviation of the target head pitch angle, the second standard deviation of the target sight direction horizontal angle and the second standard deviation of the target sight direction pitch angle are all less than or equal to the target preset standard deviation threshold value, then performing histogram statistics on the target head horizontal angle normalized standard data group, the target head pitch angle normalized standard data group, the target sight direction horizontal angle normalized standard data group and the target sight direction pitch angle normalized standard data group respectively to obtain target head horizontal angle histogram distribution function data, target head pitch angle histogram distribution function data, target sight direction horizontal angle histogram distribution function data and target sight direction pitch angle histogram distribution function data; Respectively calculating the target head horizontal angle KL divergence, target head pitch angle KL divergence, target sight direction horizontal angle KL divergence and target sight direction pitch angle KL divergence between the four groups of data, namely, the target head horizontal angle histogram distribution function data, the target head pitch angle histogram distribution function data, the target sight direction horizontal angle histogram distribution function data and the target sight direction pitch angle histogram distribution function data and the preset target normal distribution function data; Determine whether at least one of the target head horizontal angle KL divergence, the target head pitch angle KL divergence, the target sight direction horizontal angle KL divergence, and the target sight direction pitch angle KL divergence is greater than a preset KL divergence; If so, it is determined that the calibration of the target driver state has failed, and a new round of calibration of the driver's driving state is performed again.

2. The method for calibrating the driver's driving state according to claim 1, characterized in that: The step of judging whether at least one of the target head horizontal angle KL divergence, the target head pitch angle KL divergence, the target sight line direction horizontal angle KL divergence, and the target sight line direction pitch angle KL divergence is greater than a preset KL divergence comprises: if the target head horizontal angle KL divergence, the target head pitch angle KL divergence, the target sight line direction horizontal angle KL divergence, and the target sight line direction pitch angle KL divergence are all less than or equal to the preset KL divergence, then judging that the calibration of the target driver state is successful; Respectively calculating the maximum function value of the target head horizontal angle, the maximum function value of the target head pitch angle, the maximum function value of the target sight direction horizontal angle, and the maximum function value of the target sight direction pitch angle corresponding to the target head horizontal angle histogram distribution function data, the target head pitch angle histogram distribution function data, the target sight direction horizontal angle histogram distribution function data, and the target sight direction pitch angle histogram distribution function data; Respectively obtain the target head horizontal angle abscissa value, the target head pitch angle abscissa value, the target sight direction horizontal angle abscissa value, and the target sight direction pitch angle abscissa value corresponding to the target head horizontal angle maximum function value, the target head pitch angle maximum function value, the target sight direction horizontal angle maximum function value, and the target sight direction pitch angle maximum function value; Respectively taking the target head horizontal angle abscissa value, the target head pitch angle abscissa value, the target sight direction horizontal angle abscissa value and the target sight direction pitch angle abscissa value as the target head horizontal angle calibration angle, the target head pitch angle calibration angle, the target sight direction horizontal angle calibration angle and the target sight direction pitch angle calibration angle; The number of successful calibrations of the driver's driving status is increased by one.

3. The method for calibrating the driver's driving state according to claim 2, characterized in that: The number of successful calibrations of the driver's driving state is increased by one, and then the method includes: determining whether the number of successful calibrations of the target driver's state is greater than or equal to a preset successful calibration threshold; If so, respectively calculate the target calibration data of N consecutive times that are in front in time, wherein the target calibration data of N times include N head horizontal angle calibration angles, N head pitch angle calibration angles, N sight direction horizontal angle calibration angles and N sight direction pitch angle calibration angles, wherein N is a positive integer greater than or equal to the preset calibration success number threshold; Respectively calculating the target head horizontal angle standard deviation, target head pitch angle standard deviation, target sight direction horizontal angle standard deviation and target sight direction pitch angle standard deviation corresponding to the N head horizontal angle calibration angles, the N head pitch angle calibration angles, the N sight direction horizontal angle calibration angles and the N sight direction pitch angle calibration angles; Determine whether the target head horizontal angle standard deviation, the target head pitch angle standard deviation, the target sight direction horizontal angle standard deviation and the target sight direction pitch angle standard deviation are all less than the target preset calibration angle standard deviation threshold; If so, it is determined that the calibration of the driver's driving state is successful, and the number of successful calibrations of the driver's driving state is increased by one; otherwise, it is determined that the calibration of the target driver's state has failed, and a new round of calibration of the driver's driving state is performed again.

4. The method for calibrating the driver's driving state according to claim 3, characterized in that: The determining that the calibration of the driver's driving state is successful and the number of successful calibrations of the driver's driving state is increased by one, then includes: determining whether the calibration of the driver's driving state has been successful before the current calibration; If the calibration of the driver's driving state has been successful before this calibration, then determining whether the calibration result data of this calibration is abnormal compared with the calibration result data of the previous calibration; If there is an abnormality between the calibration result data of this time and the calibration result data of the previous time, the calibration result data of this time will replace the calibration result data of the previous time. If the calibration has not been successful before, the calibration result data of this time will be used as the calibration result data of the current driving state of the driver.

5. A device for calibrating a driver's driving state, characterized in that: Calibration based on the driver monitoring system, which includes: A vehicle speed detection module, used to determine whether a target vehicle speed corresponding to the target vehicle is greater than or equal to a target vehicle speed threshold at predetermined time intervals during the process of the target driver driving the target vehicle; A face image acquisition module is used to enter a target driver state calibration program and acquire a target face image corresponding to the target driver for a preset number of frames when the target vehicle speed corresponding to the target vehicle is greater than or equal to a target vehicle speed threshold; A face image prediction module is used to input the target face image of the target preset number of frames into a target neural network model for prediction, and obtain target head posture data and target sight line direction data corresponding to the target driver of the target preset number of frames, wherein the target head posture data includes target head horizontal angle data and target head pitch angle data, and the target sight line direction data includes target sight line direction horizontal angle data and target sight line direction pitch angle data; A driving data statistics module is used to perform statistical processing on the target head horizontal angle data, the target head pitch angle data, the target sight line direction horizontal angle data and the target sight line direction pitch angle data, respectively, to obtain the target head horizontal angle statistics result, the target head pitch angle statistics result, the target sight line direction horizontal angle statistics result and the target sight line direction pitch angle statistics result; A statistical result judgment module is used to judge whether the target head horizontal angle statistical result, the target head pitch angle statistical result, the target sight direction horizontal angle statistical result and the target sight direction pitch angle statistical result all meet the target preset statistical requirements; A statistical result calibration module, used for determining that the calibration of the driver's driving state is successful when the statistical result of the target head horizontal angle, the statistical result of the target head pitch angle, the statistical result of the target sight direction horizontal angle and the statistical result of the target sight direction pitch angle all meet the preset statistical requirements of the target; The statistical processing is performed on the target head horizontal angle data, the target head pitch angle data, the target sight line direction horizontal angle data and the target sight line direction pitch angle data respectively to obtain the target head horizontal angle statistical result, the target head pitch angle statistical result, the target sight line direction horizontal angle statistical result and the target sight line direction pitch angle statistical result, including: respectively calculating the target head horizontal angle first standard deviation, the target head pitch angle first standard deviation, the target sight line direction horizontal angle first standard deviation and the target sight line direction pitch angle first standard deviation corresponding to the target head horizontal angle data, the target head pitch angle data, the target sight line direction horizontal angle data and the target sight line direction pitch angle data; According to the first standard deviation of the target head horizontal angle, the first standard deviation of the target head pitch angle, the first standard deviation of the target sight direction horizontal angle, and the first standard deviation of the target sight direction pitch angle, respectively calculate a target head horizontal angle normalized value group, a target head pitch angle normalized value group, a target sight direction horizontal angle normalized value group, and a target sight direction pitch angle normalized value group corresponding to the target head horizontal angle data, the target head pitch angle data, the target sight direction horizontal angle data, and the target sight direction pitch angle data; Calculate target singular values ​​corresponding to the target head horizontal angle normalized value group, the target head elevation angle normalized value group, the target sight direction horizontal angle normalized value group, and the target sight direction elevation angle normalized value group respectively, and the absolute value of the target singular value is greater than the target preset value; Respectively removing all the target singular values ​​in the target head horizontal angle normalized value group, the target head pitch angle normalized value group, the target sight direction horizontal angle normalized value group, and the target sight direction pitch angle normalized value group, to obtain a target head horizontal angle normalized standard data group, a target head pitch angle normalized standard data group, a target sight direction horizontal angle normalized standard data group, and a target sight direction pitch angle normalized standard data group; Respectively calculating the target head horizontal angle second standard deviation, the target head pitch angle second standard deviation, the target sight direction horizontal angle second standard deviation and the target sight direction pitch angle second standard deviation corresponding to the target head horizontal angle normalized standard data group, the target head pitch angle normalized standard data group, the target sight direction horizontal angle normalized standard data group and the target sight direction pitch angle normalized standard data group; The second standard deviation of the target head horizontal angle, the second standard deviation of the target head pitch angle, the second standard deviation of the target sight direction horizontal angle and the second standard deviation of the target sight direction pitch angle are respectively the target head horizontal angle statistics, the target head pitch angle statistics, the target sight direction horizontal angle statistics and the target sight direction pitch angle statistics; The determining whether the target head horizontal angle statistical result, the target head pitch angle statistical result, the target sight direction horizontal angle statistical result and the target sight direction pitch angle statistical result all meet the target preset statistical requirements, including: determining whether at least one of the target head horizontal angle second standard deviation, the target head pitch angle second standard deviation, the target sight direction horizontal angle second standard deviation and the target sight direction pitch angle second standard deviation is greater than the target preset standard deviation threshold; If yes, it is determined that the calibration of the target driver state has failed, and a new round of calibration of the driver's driving state is performed again; The step of judging whether at least one of the second standard deviation of the target head horizontal angle, the second standard deviation of the target head pitch angle, the second standard deviation of the target sight direction horizontal angle and the second standard deviation of the target sight direction pitch angle is greater than a target preset standard deviation threshold value comprises: if the second standard deviation of the target head horizontal angle, the second standard deviation of the target head pitch angle, the second standard deviation of the target sight direction horizontal angle and the second standard deviation of the target sight direction pitch angle are all less than or equal to the target preset standard deviation threshold value, then performing histogram statistics on the target head horizontal angle normalized standard data group, the target head pitch angle normalized standard data group, the target sight direction horizontal angle normalized standard data group and the target sight direction pitch angle normalized standard data group respectively to obtain target head horizontal angle histogram distribution function data, target head pitch angle histogram distribution function data, target sight direction horizontal angle histogram distribution function data and target sight direction pitch angle histogram distribution function data; Respectively calculating the target head horizontal angle KL divergence, target head pitch angle KL divergence, target sight direction horizontal angle KL divergence and target sight direction pitch angle KL divergence between the four groups of data, namely, the target head horizontal angle histogram distribution function data, the target head pitch angle histogram distribution function data, the target sight direction horizontal angle histogram distribution function data and the target sight direction pitch angle histogram distribution function data and the preset target normal distribution function data; Determine whether at least one of the target head horizontal angle KL divergence, the target head pitch angle KL divergence, the target sight direction horizontal angle KL divergence, and the target sight direction pitch angle KL divergence is greater than a preset KL divergence; If so, it is determined that the calibration of the target driver state has failed, and a new round of calibration of the driver's driving state is performed again.

6. A computer-readable storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a processor, the processor executes the method for calibrating the driver's driving state as described in any one of claims 1 to 4.

7. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the method for calibrating the driving state of the driver as claimed in any one of claims 1 to 4.

Images