A driver fatigue detection method based on fixed-duration rolling statistical blink frequency

Abstract

The present invention discloses a driver fatigue detection method based on a fixed-duration rolling statistical blink frequency, and the present invention relates to the technical field of driver fatigue detection methods. The driver fatigue detection method based on a fixed-duration rolling statistical blink frequency uses facial image data MT of the target driver within i hours after he or she starts driving to construct a data set Q during the daily driving process of the target driver, and introduces facial brightness data LD in combination with blink frequency data to judge the driver's fatigue driving, thereby obtaining blink alarm data ZYF under different intensities of light, and then determines the blink alarm data ZYF in combination with the real-time facial brightness data LD of the target driver to judge the degree of fatigue driving of the target driver, thereby improving the accuracy of fatigue driving prediction, ensuring that the optimal alarm threshold under a certain intensity of light can be accurately determined, and greatly improving the safety factor.

Description

Technical Field

[0001] The present invention relates to the technical field of driver fatigue detection methods, and in particular to a driver fatigue detection method using fixed-duration rolling statistics of blinking frequency. Background Art

[0002] Driver fatigue detection can greatly ensure driving safety and improve driving safety. It mainly collects facial image data of the driver during driving through the image acquisition module, analyzes the driver's facial features, and then determines whether the driver is driving fatigued by analyzing the facial features. Specifically, it includes analyzing the driver's breathing frequency, the probability of the driver's eye closing time, and the driver's blinking frequency. The collection and analysis of these features can realize the determination of the driver's driving fatigue.

[0003] The existing Chinese invention patent with announcement number CN105054951B is a method for monitoring fatigue driving EEG based on blinking frequency recognition. In the prior art, when using blinking frequency to detect driver fatigue driving, a fixed collection period is usually set to perform periodic blinking frequency data collection, and periodically perform rolling driving fatigue detection. However, in the prior art, the standard for determining fatigue driving detection using blinking frequency, that is, the upper limit of blinking frequency, is mostly predicted by conventional values ​​in the prior art, that is, the real-time acquired blinking frequency is compared with the conventional upper limit of blinking frequency. If it is higher than the upper limit of conventional blinking frequency, it is determined that fatigue driving behavior will occur. Another method is to calculate the average blinking frequency of the driver's daily driving blinking frequency and determine the upper limit of blinking frequency, and then compare and determine. However, both methods have the following limitations:

[0004] The use of conventional lower limit of blinking frequency and the use of the average blinking frequency calculated from the driver's daily driving blinking frequency to determine the upper limit of blinking frequency are both quite limited. Specifically, the blinking frequency of the driver's eyes is different under different ambient light conditions. Specifically, the blinking frequency is higher under strong light, especially when driving in the direction facing the sun or at night when there are many vehicles with their lights turned on. The driver's blinking frequency is higher, so the upper limit of blinking frequency that tends to be fatigue driving is also higher than usual. The driver's blinking frequency is different under different ambient lighting conditions such as cloudy days, overcast weather or rainy days, so the predicted upper limit of fatigue is also different. Therefore, using the conventional upper limit of blinking frequency or judging by the upper limit of blinking frequency based on the average blinking frequency of the driver during daily driving will result in a large error and low prediction accuracy.

[0005] Therefore, it is necessary to provide a driver fatigue detection method with a fixed-duration rolling statistical blinking frequency to solve the above technical problems. Summary of the invention

[0006] 1. Technical issues to be resolved

[0007] In order to solve the above technical problems, the present invention provides a driver fatigue detection method with fixed-duration rolling statistics of blinking frequency.

[0008] (II) Technical solution

[0009] To achieve the above objectives, the present invention is implemented through the following technical solutions: a driver fatigue detection method with fixed-duration rolling statistical blink frequency, comprising the following steps:

[0010] A1. The image acquisition device acquires facial image data MT of the target driver according to the set rolling cycle GZ. The facial image data MT includes facial feature data mb, blinking frequency data ZY, eye closing data BS and facial brightness data LD.

[0011] A2. The data processing device constructs a data set Q using the acquired facial image data MT of the target driver and stores it in a storage device. By analyzing and processing the data set Q, the blink alarm data ZYF of the target driver under different facial brightness data ranges LDF are obtained. The facial image data MT used to construct the data set Q are all continuously collected within a set i hours after the target driver enters the driving seat and starts driving each time.

[0012] A3. After the target driver enters the driving seat, the facial image data MT of the target driver is obtained by using the image acquisition device. Then, the data processing device determines the blinking alarm data ZYF for judging whether the target driver is driving fatigued from the data set Q based on the obtained facial image data MT, and then determines whether the target driver is fatigued by combining the determined blinking alarm data ZYF, the eye closing data BS and the blinking frequency data ZY.

[0013] Preferably, the expression of the data set Q is as follows:

[0014]

[0015] W n represents the nth driver data set, mb n represents the facial feature data of the nth target driver, represents the mth data packet of the nth target driver, represents the facial brightness data range in the mth data packet of the nth target driver, Represents the vth blink frequency data in the mth data packet of the nth target driver.

[0016] Preferably, the facial brightness data range LDF is obtained by artificial setting, and the steps of storing the blink frequency data ZY into the data set Q are as follows:

[0017] A21, matching the acquired facial feature data mb with the facial feature data mb in each driver data set W in the data set Q;

[0018] A22. If the match is not successful, a new driver data set W is created in the data set Q, and the acquired facial feature data mb is stored in the newly created driver data set W. When the data processing device creates the new driver data set W, it uses the artificially set facial brightness data range LDF to construct multiple data packets w in the newly created driver data set W. The subsequent steps are the same as those for successful matching, and the following steps are performed;

[0019] A23, in the driver data set W to which the successfully matched facial feature data mb belongs, or in the newly created driver data set W, determining into which facial brightness data range LDF the acquired facial brightness data LD falls, and then storing the acquired blinking frequency data ZY into a data packet w to which the facial brightness data LD falls;

[0020] A24, after the blink frequency data ZY in the data packet w reaches the set number H, the data processing device calculates the blink frequency data ZY in the data packet w to determine the blink alarm data ZYF, and then uses the blink alarm data ZYF to replace the blink frequency data ZY in the data packet w, and obtains in Represents the blink warning data ZYF in the mth data packet of the nth target driver.

[0021] Preferably, the image acquisition device, data processing device and storage device are all arranged in the vehicle.

[0022] Preferably, the blink alarm data ZYF is calculated using the following formula:

[0023]

[0024] Where T represents the data packet The total number of blink frequency data ZY, where x is the set variable.

[0025] Preferably, the specific steps of determining the blinking alarm data ZYF for judging whether the target driver is driving fatigued from the data set Q in step A3 are as follows:

[0026] B1. The data processing device matches the acquired facial feature data mb with all the facial feature data mb stored in the storage device;

[0027] B2, if the match is not successful, the normal value S is set as the blink alarm data ZYF to judge the driver's fatigue driving, and the acquired facial image data MT is used to execute A22 to A24. If the match is successful, the data processing device calibrates the driver data set W storing a facial feature data mb that has successfully matched, and continues to execute the following steps;

[0028] B3, the data processing device determines which facial brightness data range LDF in the calibrated driver data set W the acquired facial brightness data LD falls into, and calibrates the facial brightness data range LDF in the calibrated driver data set W in which the acquired facial brightness data LD falls;

[0029] B4, the data processing device extracts the blink alarm data ZYF from a data packet w to which the calibrated facial brightness data range LDF belongs, and the extracted blink alarm data ZYF is the blink alarm data ZYF used to determine whether the target driver is driving fatigued. If there is no blink alarm data ZYF in the data packet w to which the calibrated facial brightness data range LDF belongs, the normal value S is set as the blink alarm data ZYF to determine whether the driver is driving fatigued, and the acquired facial image data MT is used to execute steps A23 to A24;

[0030] B5. After the matching is successful in step B2, the facial image data MT subsequently acquired directly uses the driver data set W calibrated in step B2 to execute steps B3 and B4 until the target driver leaves the driving seat.

[0031] Preferably, the step A3 of determining whether the target driver is driving fatigued by combining the determined blink alarm data ZYF, blink frequency data ZY and eye closing data BS specifically comprises the following steps:

[0032] C1. First, the PERCLOS value of the target driver is calculated using the acquired eye-closing data BS, and it is determined whether the PERCLOS value of the target driver is greater than a preset value YP. If the PERCLOS value of the target driver is greater than the preset value YP, it is determined that the target driver is driving fatigued;

[0033] C2. If the PERCLOS value of the target driver is less than the preset value YP, the blinking frequency data ZY is used to determine whether the acquired blinking frequency data ZY is greater than the determined blinking alarm data ZYF. If so, it is determined that the target driver is about to drive fatigued, otherwise the detection work continues.

[0034] Preferably, the image acquisition device specifically includes a camera module, an image processing module and an image analysis module. The camera module at least includes an active near-infrared imaging device and an ordinary camera. The blinking frequency data ZY, the eye closing data BS and the facial feature data mb are all acquired through the active near-infrared imaging device, and the facial brightness data LD is acquired through the ordinary camera.

[0035] (III) Beneficial effects

[0036] The present invention provides a method for detecting driver fatigue by rolling and counting blink frequency over a fixed time period. Compared with the prior art, it has the following beneficial effects:

[0037] During the target driver's daily driving process, the facial image data MT of the target driver within i hours after he started driving is used to construct a data set Q, and the driver's fatigue driving is judged by introducing the facial brightness data LD in combination with the blinking frequency data. The facial brightness data LD can reflect the ambient light intensity, and different facial brightness data LD can reflect light of different intensities, thereby obtaining the blinking alarm data ZYF under light of different intensities. Subsequently, the blinking alarm data ZYF is determined in combination with the target driver's real-time facial brightness data LD to judge the target driver's fatigue driving, which can improve the accuracy of fatigue driving prediction, ensure that the optimal alarm threshold under a certain intensity of light can be accurately determined, and greatly improve the safety factor. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a schematic diagram of the method of the present invention;

[0039] Figure 2 A schematic diagram of the steps of storing the blink frequency data ZY into the data set Q of the present invention;

[0040] Figure 3 It is a schematic diagram of the step of determining blink alarm data ZYF of the present invention;

[0041] Figure 4 A schematic diagram of a process for determining whether a target driver is driving fatigued according to the present invention;

[0042] Figure 5 Schematic diagram of the detection system of the present invention. DETAILED DESCRIPTION

[0043] 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.

[0044] See also Figures 1 to 5 :

[0045] Embodiment 1, the embodiment of the present invention provides a technical solution: a method for detecting driver fatigue by rolling statistics of blinking frequency for a fixed time, comprising the following steps:

[0046] A1. The image acquisition device acquires facial image data MT of the target driver according to the set rolling cycle GZ. The facial image data MT includes facial feature data mb, blinking frequency data ZY, eye closing data BS and facial brightness data LD.

[0047] The blink frequency data ZY is specifically the number of blinks per minute of the target driver obtained by analyzing the facial image of the target driver acquired through facial recognition technology;

[0048] The eye-closing data BS is specifically the percentage of the target driver's eye-closing time per minute obtained by analyzing the target driver's facial image acquired through facial recognition technology;

[0049] The facial brightness data GZ is specifically the average brightness of the face, which refers to the average brightness value of all pixels in the facial image. The brightness value represents the brightness of the pixels in the image, usually expressed as a grayscale value, ranging from 0 to 255. The method to calculate the average brightness of the face is to add the brightness values ​​of all pixels in the image and then divide it by the total number of pixels. The stronger the ambient light, the greater the facial brightness data GZ on the face, which can be obtained through the existing image brightness recognition algorithm.

[0050] By using the collected facial image of the driver, the blinking frequency data ZY and the eye closing data BS are obtained from the image analysis. These are common technical means in the prior art for detecting driver fatigue driving using blinking frequency and eye closing frequency, which can be achieved by existing technologies and algorithms, such as algorithms based on eye aspect ratio (EAR) and algorithms based on image processing, etc.;

[0051] The extraction of facial feature data mb in the facial image can be achieved by using existing algorithms such as Haar CascadeClassifier and Active Shape Model, so it will not be described in detail here;

[0052] A2. The data processing device constructs a data set Q using the acquired facial image data MT of the target driver and stores it in a storage device. By analyzing and processing the data set Q, the blink alarm data ZYF of the target driver under different facial brightness data ranges LDF are obtained. The facial image data MT used to construct the data set Q are all continuously collected within a set i hours after the target driver enters the driving seat and starts driving each time.

[0053] In order to ensure that the facial image data MT used to construct the data set Q are all collected when the target driver is relatively sober, according to daily driving habits, the driver needs to rest when the continuous driving time reaches four hours. If the continuous driving time exceeds 4 hours, it is fatigue driving. Therefore, the value of i hours can be preset according to actual conditions, and can be preset to 1 hour. Then, the facial image of the driver while driving is continuously obtained within one hour after the target driver enters the driving seat and starts driving. In this way, it can be ensured that the acquired blinking frequency data ZY can most truly express the normal blinking frequency of the target driver;

[0054] The expression of the data set Q is as follows:

[0055]

[0056] W n represents the nth driver data set, mb n represents the facial feature data of the nth target driver, represents the mth data packet of the nth target driver, represents the facial brightness data range in the mth data packet of the nth target driver, represents the vth blink frequency data in the mth data packet of the nth target driver;

[0057] The face brightness data range LDF is obtained by artificial setting, and the face brightness data range GZF can be divided into 5 levels. The environmental brightness corresponding to the face brightness data range GZF of the 5 levels is as follows in Table 1:

[0058] Table 1: Face brightness data range GZF corresponding to the ambient brightness table

[0059]

[0060] Therefore, each driver data set W includes 5 data packets w, and each data packet w stores a level of facial brightness data range LDF.

[0061] The specific steps of storing the blink frequency data ZY into the data set Q are as follows:

[0062] A21, matching the acquired facial feature data mb with the facial feature data mb in each driver data set W in the data set Q;

[0063] A22. If the match is not successful, a new driver data set W is created in the data set Q, and the acquired facial feature data mb is stored in the newly created driver data set W. When the data processing device creates the new driver data set W, it uses the artificially set facial brightness data range LDF to construct multiple data packets w in the newly created driver data set W. The subsequent steps are the same as those for successful matching, and the following steps are performed;

[0064] A23, in the driver data set W to which the successfully matched facial feature data mb belongs, or in the newly created driver data set W, determining into which facial brightness data range LDF the acquired facial brightness data LD falls, and then storing the acquired blinking frequency data ZY into a data packet w to which the facial brightness data LD falls;

[0065] A24, after the blink frequency data ZY in the data packet w reaches the set number H, the data processing device calculates the blink frequency data ZY in the data packet w to determine the blink alarm data ZYF, and then uses the blink alarm data ZYF to replace the blink frequency data ZY in the data packet w, and obtains in represents the blink warning data ZYF in the mth data packet of the nth target driver;

[0066] A3. After the target driver enters the driving seat, the facial image data MT of the target driver is obtained by using the image acquisition device. Then, the data processing device determines the blinking alarm data ZYF used to determine whether the target driver is driving fatigued from the data set Q based on the acquired facial image data MT, and then determines whether the target driver is fatigued by combining the determined blinking alarm data ZYF, the eye closing data BS and the blinking frequency data ZY;

[0067] The blink alarm data ZYF is calculated using the following formula:

[0068]

[0069] Where T represents the data packet The total number of blink frequency data ZY, where x is the set variable.

[0070] The specific steps of determining the blink warning data ZYF for judging whether the target driver is driving fatigued from the data set Q in step A3 are as follows:

[0071] B1. The data processing device uses a facial recognition algorithm to match the acquired facial feature data mb with all facial feature data mb stored in the storage device;

[0072] The facial feature data mb is obtained from the acquired facial image of the target driver. The key feature information is extracted from the facial image. This is usually achieved through deep learning methods, such as convolutional neural network (CNN). CNN can convert the input facial image into a feature vector of fixed dimension. This feature vector can represent the key features of the face.

[0073] The pairing of two sets of facial feature data mb is also an existing technology and can be achieved using existing technology and algorithms;

[0074] B2. If the match is not successful, the normal value S is set as the blink alarm data ZYF to judge the driver's fatigue driving, and the acquired facial image data MT is used to execute A22 to A24, and the setting of "the facial image data MT used to construct the data set Q are all continuously collected within a set i hours after the target driver enters the driving seat and starts driving each time" in step A2 is followed. If the match is successful, the data processing device calibrates the driver data set W that stores a facial feature data mb that has successfully matched, and continues to execute the following steps;

[0075] B3, the data processing device determines which facial brightness data range LDF in the calibrated driver data set W the acquired facial brightness data LD falls into, and calibrates the facial brightness data range LDF in the calibrated driver data set W in which the acquired facial brightness data LD falls;

[0076] B4, the data processing device extracts the blink alarm data ZYF from a data packet w to which the calibrated facial brightness data range LDF belongs, and the extracted blink alarm data ZYF is the blink alarm data ZYF used to determine whether the target driver is driving fatigued. If there is no blink alarm data ZYF in the data packet w to which the calibrated facial brightness data range LDF belongs, a normal value S is set as the blink alarm data ZYF to determine whether the driver is driving fatigued, and steps A23 to A24 are performed using the acquired facial image data MT, and the setting in step A2 that "the facial image data MT used to construct the data set Q are all continuously collected within a set i hours after each target driver enters the driving seat and starts driving" is followed;

[0077] B5. After the matching is successful in step B2, the facial image data MT subsequently acquired directly uses the driver data set W calibrated in step B2 to execute steps B3 and B4 until the target driver leaves the driving seat;

[0078] In this way, the fatigue driving prediction and judgment of the target driver is carried out while constructing the data set Q. The setting of the conventional value S is a commonly used value in the prior art for predicting whether the target driver is fatigued by using the blinking frequency data ZY. It can be obtained by big data acquisition and analysis. For example, according to big data, it can be obtained that the human eye blinks 15-20 times per minute under normal circumstances. Then 20 can be used as the setting of the conventional value S, that is, the upper limit of the blinking frequency for predicting fatigue driving, to participate in the judgment of the fatigue driving of the target driver, that is, it can be ensured that the fatigue driving judgment of the target driver can be carried out when the data packet w is not established and perfected;

[0079] Embodiment 2: This embodiment provides a further technical solution based on embodiment 1. In step A3, combining the determined blink alarm data ZYF, blink frequency data ZY and eye closing data BS to determine whether the target driver is driving fatigued specifically includes the following steps:

[0080] C1. First, the PERCLOS value of the target driver is calculated using the acquired eye-closing data BS, and it is determined whether the PERCLOS value of the target driver is greater than a preset value YP. If the PERCLOS value of the target driver is greater than the preset value YP, it is determined that the target driver is driving fatigued;

[0081] C2. If the PERCLOS value of the target driver is less than the preset value YP, the blink frequency data ZY is used to determine whether the acquired blink frequency data ZY is greater than the determined blink alarm data ZYF. If so, it is determined that the target driver is about to drive fatigued, otherwise the detection work continues;

[0082] PERCLOS is a parameter used to measure the percentage of time that the eyes are closed. It is often used to assess the driver's fatigue or attention level. It specifically refers to the percentage of time that the eyes are closed for more than a certain proportion (such as 70% or 80%) per unit time (usually 1 minute).

[0083] Measurement method:

[0084] PERCLOS usually has three measurement methods, two of which are described below:

[0085] P70: refers to the eye being considered closed when the area of ​​the pupil covered by the eyelid exceeds 70%, and the percentage of time the eye is closed within a certain period of time is counted;

[0086] P80: refers to the eye being considered closed when the area of ​​the pupil covered by the eyelid exceeds 80%, and the percentage of time the eye is closed within a certain period of time is counted;

[0087] Usually, the P70 measurement method is selected to calculate the PERCLOS value. When the P70 measurement method is selected, the preset value YP is set to 30%;

[0088] Example description:

[0089] In actual use, the image acquisition device, data processing device and storage device constitute a detection system, which is arranged in the vehicle and connected to the vehicle computer for data;

[0090] When in use, i is set to 1, the rolling cycle GZ is set to one minute, and H is set to 60. Then, when the target driver enters the driving seat and starts driving, the image acquisition device is used to acquire the target driver's facial image once a minute, and a facial image of one minute is obtained each time. Facial image data MT is obtained from the facial image, and the obtained facial image data MT is used to construct the data set Q. When the number of blinking frequency data ZY in a data packet w in a driver data set W that matches the target driver's facial feature data mb reaches the set number of 60, the blinking alarm data ZYF can be calculated. In this way, the blinking alarm data ZYF of the corresponding target driver in the corresponding facial brightness data range LDF can be obtained, thereby obtaining the blinking alarm data ZYF of the target driver under different lighting conditions. For the calculation of the blinking alarm data ZYF, an example calculation is performed:

[0091] The unit of blink frequency data is times / minute. The values ​​of 60 blink frequency data ZY of the nth target driver in the mth data packet are set as shown in Table 2 below:

[0092] Table 2: ZY value table of 60 blink frequency data

[0093]

[0094]

[0095] Substitute the data in Table 2 into the blink alarm data ZYF calculation formula, and the value of x in the blink alarm data ZYF calculation formula is set to 3, and the calculation is as follows:

[0096]

[0097] The calculation result of the blink alarm data ZYF adopts the principle of rounding off, and the blink frequency data ZY also adopts the principle of rounding off;

[0098] Therefore, it can be obtained that the target driver's blinking alarm data ZYF is 21 times under a certain face brightness data range GZF;

[0099] If the PERCLOS value calculated by obtaining the target driver's facial image data MT at a certain moment is 28%, according to step C1, 28%<30%, and therefore less than the preset value YP, then the blink alarm data ZYF is used for judgment, and the facial image data MT is used to determine the blink alarm data ZYF in the data set Q. If the blink alarm data ZYF finally determined is 17 seconds / minute, and the blink frequency data ZY in the target driver's facial image data MT obtained is 20 seconds / minute, then the blink frequency data ZY> the blink alarm data ZYF, and then it can be determined that the target driver will have fatigue driving behavior in the future, and the vehicle computer can be used for alarm work;

[0100] There are two solutions for defining the blink alarm data ZYF in the prior art. One is to take the upper limit value, that is, if the actual blink frequency data ZY is greater than the upper limit value blink alarm data ZYF, it is determined that the driver is about to have fatigue driving behavior. The specific basis is that after long-term driving and fatigue, the eyes continue to work, and the eyelids will increase the number of blinks to moisten the eyeball part. This solution can achieve the prediction of fatigue driving;

[0101] The other is to take the lower limit, that is, the actual blink frequency data ZY is less than the lower limit blink alarm data ZYF, then it is determined that the driver is in a fatigue driving state, and the eye closing rate, that is, the PERCLOS value, is greater than 30%, and the blink frequency data ZY drops significantly, the driver is in a highly relaxed state, and driving is extremely dangerous at this time;

[0102] In summary, the present application uses the upper limit value of the blink alarm data ZYF as a prediction of the driver's upcoming fatigue driving, thereby achieving the effect of early warning;

[0103] During the process of constructing the data set Q, fatigue driving detection can also be carried out simultaneously. Once the data set Q is fully constructed, the target driver's blinking alarm data ZYF under different lighting conditions can be accurately understood, which greatly improves the accuracy of fatigue driving prediction and greatly improves the safety factor.

[0104] Embodiment 3. This embodiment provides a detection system, including an image acquisition device, a data processing device and a storage device. The image acquisition device, the data processing device and the storage device are all arranged in the vehicle. The image acquisition device specifically includes a camera module, an image processing module and an image analysis module. The camera module at least includes an active near-infrared imaging device and an ordinary camera. The blinking frequency data ZY, the closed eyes data BS and the facial feature data mb are all acquired by the active near-infrared imaging device, and the facial brightness data LD is acquired by the ordinary camera.

[0105] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0106] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0107] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A driver fatigue detection method using fixed-duration rolling statistics of blink frequency, characterized in that: The steps include: A1. The image acquisition device acquires facial image data MT of the target driver according to the set rolling cycle GZ. The facial image data MT includes facial feature data mb, blinking frequency data ZY, eye closing data BS and facial brightness data LD. A2. The data processing device constructs a data set Q using the acquired facial image data MT of the target driver and stores it in a storage device. By analyzing and processing the data set Q, the blink alarm data ZYF of the target driver under different facial brightness data ranges LDF are obtained. The facial image data MT used to construct the data set Q are all continuously collected within a set i hours after the target driver enters the driving seat and starts driving each time. A3. After the target driver enters the driving seat, the facial image data MT of the target driver is obtained by using the image acquisition device. Then, the data processing device determines the blinking alarm data ZYF used to determine whether the target driver is driving fatigued from the data set Q based on the acquired facial image data MT, and then determines whether the target driver is fatigued by combining the determined blinking alarm data ZYF, the eye closing data BS and the blinking frequency data ZY; The expression of the data set Q is as follows: ； W n represents the nth driver data set, mb n represents the facial feature data of the nth target driver, represents the mth data packet of the nth target driver, represents the facial brightness data range in the mth data packet of the nth target driver, represents the vth blink frequency data in the mth data packet of the nth target driver; The facial brightness data range LDF is obtained by artificial setting, and the steps of storing the blink frequency data ZY into the data set Q are as follows: A21, matching the acquired facial feature data mb with the facial feature data mb in each driver data set W in the data set Q; A22. If the match is not successful, a new driver data set W is created in the data set Q, and the acquired facial feature data mb is stored in the newly created driver data set W. When the data processing device creates the new driver data set W, it uses the artificially set facial brightness data range LDF to construct multiple data packets w in the newly created driver data set W. The subsequent steps are the same as those for successful matching, and the following steps are performed; A23, in the driver data set W to which the successfully matched facial feature data mb belongs, or in the newly created driver data set W, determining into which facial brightness data range LDF the acquired facial brightness data LD falls, and then storing the acquired blinking frequency data ZY into a data packet w to which the facial brightness data LD falls; A24. After the blink frequency data ZY in the data packet w reaches the set number H, the data processing device calculates the blink frequency data ZY in the data packet w to determine the blink alarm data ZYF, and then uses the blink alarm data ZYF to replace the blink frequency data ZY in the data packet w.

2. The driver fatigue detection method of fixed-duration rolling statistical blink frequency according to claim 1 is characterized by: The image acquisition device, data processing device and storage device are all arranged in the vehicle.

3. The driver fatigue detection method of fixed-duration rolling statistical blink frequency according to claim 1 is characterized by: The blink alarm data ZYF is calculated using the following formula: ； Where T represents the data packet The total number of blink frequency data ZY, where x is the set variable.

4. The driver fatigue detection method of fixed-duration rolling statistical blink frequency according to claim 3 is characterized by: The specific steps of determining the blink warning data ZYF for judging whether the target driver is driving fatigued from the data set Q in step A3 are as follows: B1. The data processing device matches the acquired facial feature data mb with all the facial feature data mb stored in the storage device; B2, if the match is not successful, the normal value S is set as the blink alarm data ZYF to judge the driver's fatigue driving, and the acquired facial image data MT is used to execute A22 to A24. If the match is successful, the data processing device calibrates the driver data set W storing a facial feature data mb that has successfully matched, and continues to execute the following steps; B3, the data processing device determines which facial brightness data range LDF in the calibrated driver data set W the acquired facial brightness data LD falls into, and calibrates the facial brightness data range LDF in the calibrated driver data set W in which the acquired facial brightness data LD falls; B4, the data processing device extracts the blink alarm data ZYF from a data packet w to which the calibrated facial brightness data range LDF belongs, and the extracted blink alarm data ZYF is the blink alarm data ZYF used to determine whether the target driver is driving fatigued. If there is no blink alarm data ZYF in the data packet w to which the calibrated facial brightness data range LDF belongs, the normal value S is set as the blink alarm data ZYF to determine whether the driver is driving fatigued, and the acquired facial image data MT is used to execute steps A23 to A24; B5. After the matching is successful in step B2, the facial image data MT subsequently acquired directly uses the driver data set W calibrated in step B2 to execute steps B3 and B4 until the target driver leaves the driving seat.

5. The driver fatigue detection method of fixed-duration rolling statistical blink frequency according to claim 4 is characterized by: The step A3 of determining whether the target driver is driving fatigued by combining the determined blinking alarm data ZYF, blinking frequency data ZY and closed eye data BS specifically includes the following steps: C1. First, the PERCLOS value of the target driver is calculated using the acquired eye-closing data BS, and it is determined whether the PERCLOS value of the target driver is greater than a preset value YP. If the PERCLOS value of the target driver is greater than the preset value YP, it is determined that the target driver is driving fatigued; C2. If the PERCLOS value of the target driver is less than the preset value YP, the blinking frequency data ZY is used to determine whether the acquired blinking frequency data ZY is greater than the determined blinking alarm data ZYF. If so, it is determined that the target driver is about to drive fatigued, otherwise the detection work continues.

6. The driver fatigue detection method of fixed-duration rolling statistical blink frequency according to claim 2 is characterized by: The image acquisition device specifically includes a camera module, an image processing module and an image analysis module. The camera module at least includes an active near-infrared imaging device and an ordinary camera. The blinking frequency data ZY, the eye closing data BS and the facial feature data mb are all acquired through the active near-infrared imaging device, and the facial brightness data LD is acquired through the ordinary camera.

Images