Optimized driver fatigue monitoring method and apparatus

Abstract

Disclosed in the present invention is an optimized driver fatigue monitoring method. On the basis of performing feature extraction on driving images by means of face detection, face position prediction is incorporated, and whether to perform feature extraction on the driving images by means of face detection or perform feature extraction on the driving images by means of face prediction is determined on the basis of a prediction confidence level of a predicted face position. Only when the prediction confidence level of the predicted face position does not reach a confidence level threshold, face prediction is used to perform feature extraction on the driving images, thereby reducing the number of times of performing feature extraction on the driving images by means of face prediction, and saving computing power and time. Further disclosed is an apparatus for implementing the driver fatigue monitoring method.

Description

An optimized method and device for monitoring driver fatigue. Technical Field

[0001] This invention relates to safe driving technology, and more particularly to an optimized method and device for monitoring driver fatigue. Background Technology

[0002] Fatigue driving is one of the major causes of traffic accidents. Because it cannot be explicitly prohibited by traffic regulations like drunk driving or driving while talking on the phone, it is easily overlooked. However, when a driver is fatigued, their attention will be distracted, their reaction ability will decrease, they will not be able to observe the road conditions and the surrounding environment in a timely manner, and they may even fall asleep for a short period of time, completely losing their ability to drive, which can easily lead to traffic accidents.

[0003] The World Report on Prevention of Road Traffic Injuries lists excessive or improper driving speed, driving under the influence of alcohol, and fatigued driving as the leading causes of accidents, deaths, and serious injuries to all vehicle occupants. According to statistics from my country's transportation authorities, fatigued driving accounts for approximately 20% of all traffic accidents, 40% of major traffic accidents, and 83% of traffic fatalities.

[0004] The China New Car Assessment Program (C-NCAP) Management Rules (2024 Edition) will officially come into effect on July 1, 2024. This includes the addition of a Driver Monitoring System (DMS) assessment item to the active safety section. Article 6 of EU Regulation 2019 / 2144 requires Category M and N vehicles to be equipped with advanced vehicle systems, including a Driver Drowsiness and Attention Warning (DDAW) system. The DDAW system should be able to monitor driver drowsiness and issue warnings to the driver via an HMI (Human-Machine Interface). The DDAW system's privacy and data protection procedures should comply with the EU's GDPR (General Data Protection Regulation). The DDAW regulation was officially promulgated and came into effect in August 2021, and will be mandatory for all new vehicle models from July 2022 to July 2024.

[0005] Vision-based driver monitoring (DMS) systems employ direct monitoring and non-contact methods, characterizing driver fatigue levels by acquiring image signals of the driver's face, eyes, and head movements. Compared to indirect monitoring based on vehicle behavior features, direct monitoring offers higher accuracy and real-time performance, enabling faster detection of fatigue. For example, Chinese patent application CN202410210718.7 discloses a driver fatigue detection method based on lightweight neural network image enhancement, comprising: real-time acquisition of dynamic images of the driver in the cockpit; enhancement processing of the dynamic images; capturing a stream of the driver's face images in the enhanced dynamic images using a lightweight fast face detector; inputting the face image stream into a preset facial key feature point extraction network model to extract facial key feature points from each frame of the face image stream, wherein the facial key feature point extraction network model is obtained by constructing and training a heavily parameterized lightweight convolutional network; determining whether the driver exhibits facial fatigue features based on the facial key feature points, and determining that the driver is driving while fatigued when the facial fatigue features are detected.

[0006] However, face detection requires processing the entire dynamic image, which consumes a lot of computing resources and takes a long time, seriously affecting the timeliness and safety of driver fatigue monitoring. Summary of the Invention

[0007] To address the shortcomings of the existing technology, this invention provides an optimized method and apparatus for monitoring driver fatigue, which can reduce the number of face detections to save computing power and time.

[0008] The technical problem to be solved by the present invention is achieved through the following technical solution:

[0009] An optimized method for monitoring driver fatigue includes the following steps:

[0010] Step 100: Obtain the Nth driving image, N≥1;

[0011] Step 200: Determine whether there is a face prediction location in the Nth driving image. If the face prediction location exists, proceed to steps 300-700. If the face prediction location does not exist, proceed to steps 500-700. The face prediction location in the Nth driving image is obtained based on the prediction of the (N-1)th driving image.

[0012] Step 300: Extract features from the Nth driving image based on the predicted face location to obtain first feature information;

[0013] Step 400: Calculate the prediction confidence based on the first feature information, and determine whether the prediction confidence reaches the confidence threshold. If it does not reach the confidence threshold, proceed to steps 500-700. If it reaches the confidence threshold, use the first feature information as facial feature information and proceed to step 700.

[0014] Step 500: Perform face detection on the Nth driving image to obtain the face detection location;

[0015] Step 600: Extract features from the Nth driving image based on the face detection location to obtain second feature information, and use the second feature information as face feature information;

[0016] Step 700: Determine whether the driver is driving while fatigued based on the facial feature information.

[0017] 1. An optimized method for monitoring driver fatigue, characterized by comprising the following steps:

[0018] Step 100: Obtain the Nth driving image, N≥1;

[0019] Step 200: Determine whether there is a face prediction location in the Nth driving image. If the face prediction location exists, proceed to steps 300-700. If the face prediction location does not exist, proceed to steps 500-700. The face prediction location in the Nth driving image is obtained based on the prediction of the (N-1)th driving image.

[0020] Step 300: Extract features from the Nth driving image based on the predicted face location to obtain first feature information;

[0021] Step 400: Calculate the prediction confidence based on the first feature information, and determine whether the prediction confidence reaches the confidence threshold. If it does not reach the confidence threshold, proceed to steps 500-700. If it reaches the confidence threshold, use the first feature information as facial feature information and proceed to step 700.

[0022] Step 500: Perform face detection on the Nth driving image to obtain the face detection location;

[0023] Step 600: Extract features from the Nth driving image based on the face detection location to obtain second feature information, and use the second feature information as face feature information;

[0024] Step 700: Determine whether the driver is driving while fatigued based on the facial feature information.

[0025] Furthermore, in step 300, the step of extracting features from the Nth driving image based on the predicted face location is as follows:

[0026] Step 310: Crop the face in the Nth driving image according to the predicted face position to obtain the first face image;

[0027] Step 320: Extract features from the first face image to obtain the first feature information.

[0028] Furthermore, in step 310, when cropping the face in the Nth driving image, a corresponding face recognition box is set in the Nth driving image to crop the image based on the coordinate information predicted by the feature points in the Nth image and the length and width dimensions of the face in the (N-1)th driving image.

[0029] Furthermore, the face prediction location includes the coordinate information of at least one feature point in the Nth driving image. This coordinate information is calculated by a location prediction model based on the coordinate information of the same feature point in the N-1th driving image and the time difference between the two driving images.

[0030] Furthermore, in step 600, the step of extracting features from the Nth driving image based on the face detection location is as follows:

[0031] Step 610: Crop the face in the Nth driving image according to the face detection location to obtain the second face image;

[0032] Step 620: Extract features from the second face image to obtain the second feature information.

[0033] Furthermore, a feature extraction network is used to extract features from the first face image and the second face image. The backbone of the feature extraction network adopts an improved MobileNet network and combines it with a DenseNet network. At the same time, some modules are replaced with lighter Ghost modules, and the accuracy is improved by adding convolutional layers.

[0034] Furthermore, in step 700, the driver's fatigue level is calculated based on the proportion of time the eyes are closed per unit time. The facial feature information includes eyelid edge feature points and pupil edge feature points. The proportion of the area of ​​the pupil covered by the eyelid is calculated based on the coordinate information of the eyelid edge feature points and pupil edge feature points. When the proportion of the area reaches a first proportion threshold, it is determined that the eyes are in a closed state. The proportion of time the eyes are in a closed state per unit time is counted. When the proportion of time reaches a second proportion threshold, it is determined that the driver is driving while fatigued.

[0035] Furthermore, following step 700, the following steps are also included:

[0036] Step 800: Based on the facial feature information, perform position prediction on the (N+1)th driving image to obtain the predicted facial position of the (N+1)th driving image.

[0037] An optimized driver fatigue monitoring device is provided for implementing the aforementioned driver fatigue monitoring method; the driver fatigue monitoring device includes:

[0038] The image acquisition module is used to acquire driving images;

[0039] The location prediction module is used to predict the face in the next driving image based on the previous driving image, so as to obtain the predicted face location.

[0040] A location detection module is used to perform face detection on the driving image to obtain the face detection location;

[0041] The feature extraction module is used to extract features from the driving image based on the predicted face location or the face detection location to obtain first feature information or second feature information;

[0042] The confidence module is used to determine the first feature information or the second feature information as facial feature information based on the predicted confidence of the first feature information.

[0043] The fatigue detection module is used to determine whether the driver is driving while fatigued based on the facial feature information.

[0044] An optimized driver fatigue monitoring device includes a processor and a memory connected to the processor. The memory stores a computer program that the processor can execute. When the processor executes the computer program, it performs the aforementioned driver fatigue monitoring method.

[0045] The present invention has the following beneficial effects: The driving fatigue monitoring method of the present invention, based on the use of face detection to extract features from driving images, introduces face position prediction, and determines whether to use face detection or face prediction to extract features from the driving image based on the prediction confidence of the face position. Only when the prediction confidence of the face position does not reach the confidence threshold is face prediction used to extract features from the driving image, thereby reducing the number of times face prediction is used to extract features from the driving image, thus saving computing power and time. Attached Figure Description

[0046] Figure 1 is a flowchart of the driving fatigue monitoring method provided by the present invention.

[0047] Figure 2 is a schematic diagram of the architecture of the feature extraction network provided by the present invention.

[0048] Figure 3 is a schematic diagram of the architecture principle of separable convolutional layers in the feature extraction network provided by the present invention.

[0049] Figure 4 is a schematic diagram of the driving fatigue monitoring device provided by the present invention.

[0050] Figure 5 is a schematic diagram of another driver fatigue monitoring device provided by the present invention. Detailed Implementation

[0051] The present invention will now be described in detail with reference to the accompanying drawings and embodiments, examples of which are shown in the drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0052] In the description of this invention, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0053] Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first," "second," or "third" may explicitly or implicitly include one or more of that feature. In the description of this invention, "multiple" means two or more, unless otherwise explicitly specified.

[0054] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," "fixing," and "setting," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances. Example 1

[0055] As shown in Figure 1, an optimized method for monitoring driver fatigue includes the following steps:

[0056] Step 100: Obtain the Nth driving image, where N≥1.

[0057] In step 100, a monitoring camera can be installed in front of the driver's cab of the vehicle (such as at the A-pillar or the upper part of the windshield) to capture real-time images of the driver while the vehicle is in motion.

[0058] Preferably, the surveillance camera is an infrared camera.

[0059] Step 200: Determine whether there is a face prediction location in the Nth driving image. If the face prediction location exists, proceed to steps 300-700. If the face prediction location does not exist, proceed to steps 500-700. The face prediction location in the Nth driving image is obtained based on the prediction of the (N-1)th driving image.

[0060] In step 200, for the first driving image, since there is no zeroth driving image, it is impossible to predict the face position of the first driving image based on the zeroth driving image. That is, there is no face prediction position for the first driving image. Therefore, steps 300-700 are performed to calculate the fatigue level of the first driving image by face detection. For the second and subsequent driving images, since there is a first and subsequent driving image, it is possible to predict the face position of the second and subsequent driving images based on the first and subsequent driving images. Therefore, steps 500-700 are performed to calculate the fatigue level of the second and subsequent driving images by face prediction or face detection.

[0061] Step 300: Extract features from the Nth driving image based on the predicted face location to obtain first feature information.

[0062] In step 300, the face prediction location includes the coordinate information of at least one feature point in the Nth driving image. This coordinate information is calculated by a location prediction model based on the coordinate information of the same feature point in the N-1th driving image and the time difference between the two driving images.

[0063] In this embodiment, the feature points used for location prediction include, but are not limited to, the intersection points between the face's central axis and the line connecting the two eyes; the location prediction model can be, but is not limited to, a Kalman filter model, a Markov chain model, or a particle filter model.

[0064] In step 300, the step of extracting features from the Nth driving image based on the predicted face location is as follows:

[0065] Step 310: Crop the face in the Nth driving image according to the predicted face position to obtain the first face image.

[0066] In step 310, since the predicted face position is obtained based on the prediction of the (N-1)th driving image, when extracting features from the (N-1)th driving image, the extracted face feature information includes the coordinate information of the feature points and the length and width of the face. When cropping the face in the Nth driving image, a corresponding face recognition box can be set in the Nth driving image to crop the image based on the predicted coordinate information of the feature points in the Nth image and the length and width of the face in the (N-1)th driving image.

[0067] Step 320: Extract features from the first face image to obtain the first feature information.

[0068] In step 320, a feature extraction network can be used to extract features from the first face image and the second face image. The backbone of the feature extraction network adopts an improved MobileNet network combined with DenseNet, while some modules are replaced with lighter Ghost modules, and the accuracy is improved by adding convolutional layers.

[0069] Specifically, as shown in Figures 2 and 3, the MobileNet network includes an input layer, multiple separable convolutional layers, a classification layer, and an output layer. The input layer, multiple separable convolutional layers, classification layer, and output layer are connected sequentially. The output of each separable convolutional layer is connected to the classification layer, except that it is connected to the input of the next separable convolutional layer. Each separable convolutional layer includes multiple ghost modules, and the input of each ghost module is connected to the output of all ghost modules in the upper layer.

[0070] Step 400: Calculate the prediction confidence based on the first feature information, and determine whether the prediction confidence reaches the confidence threshold. If the confidence threshold is not reached, proceed to steps 500-700. If the confidence threshold is reached, use the first feature information as facial feature information and proceed to step 700.

[0071] In step 400, the prediction confidence of the predicted face position is judged to determine whether there is a large deviation in the predicted face position obtained from the (N-1)th driving image. If there is a large deviation, that is, the confidence threshold is not reached, then steps 500-700 are performed, and the fatigue level is still calculated for the Nth driving image by face detection. If there is a small deviation, that is, the confidence threshold is reached, then step 700 is performed, and the fatigue level is calculated for the Nth driving image by face prediction.

[0072] The prediction confidence can be calculated using the Euclidean distance between a feature point in the predicted face location of the Nth driving image and the face recognition bounding box. Of course, the face recognition bounding box can also be calculated using other parameters according to actual needs, and should not be limited to this.

[0073] In this embodiment, the feature points used to calculate the prediction confidence level include, but are not limited to, the intersection point between the face midline and the line connecting the two eyes.

[0074] Step 500: Perform face detection on the Nth driving image to obtain the face detection location.

[0075] In step 500, a face detection network can be used to detect faces in the Nth driving image. The face detection network traverses all pixels in the Nth driving image to obtain the coordinate information of all pixels that constitute the face in the Nth driving image, the face recognition box, and the length and width of the face. The whole process requires a lot of computing resources and takes a long time to calculate.

[0076] In this embodiment, the face detection network may be, but is not limited to, a YOLO network, a CNN network, an MTCNN network, or an RNN network.

[0077] Step 600: Extract features from the Nth driving image based on the face detection location to obtain second feature information, and use the second feature information as face feature information.

[0078] In step 600, the face detection location includes the coordinate information of all pixels constituting the face, the face recognition box, and the length and width of the face, which is obtained by the face detection network through traversing all pixels in the Nth driving image.

[0079] In this embodiment, the facial feature information may, but is not limited to, using the intersection point between the facial midline and the line connecting the two eyes as feature points; the location prediction model may, but is not limited to, a Kalman filter model, a Markov chain model, or a particle filter model, etc.

[0080] In step 600, the feature extraction step for the Nth driving image based on the face detection location is as follows:

[0081] Step 610: Crop the face in the Nth driving image according to the face detection location to obtain the second face image.

[0082] In step 610, the face detection location includes the coordinate information of all feature points and the length and width of the face. When cropping the face in the Nth driving image, the corresponding face recognition box can be set in the Nth driving image to crop the image based on the coordinate information of all feature points detected in the Nth image and the length and width of the face.

[0083] Step 620: Extract features from the second face image to obtain the second feature information.

[0084] In step 620, a feature extraction network can be used to extract features from the second face image. The backbone of the feature extraction network adopts an improved MobileNet network and combines it with DenseNet. At the same time, some modules are replaced with lighter Ghost modules, and the accuracy is improved by adding convolutional layers.

[0085] Specifically, as shown in Figures 2 and 3, the MobileNet network includes an input layer, multiple separable convolutional layers, a classification layer, and an output layer. The input layer, multiple separable convolutional layers, classification layer, and output layer are connected sequentially. The output of each separable convolutional layer is connected to the classification layer, except that it is connected to the input of the next separable convolutional layer. Each separable convolutional layer includes multiple ghost modules, and the input of each ghost module is connected to the output of all ghost modules in the upper layer.

[0086] Step 700: Determine whether the driver is driving while fatigued based on the facial feature information.

[0087] In step 700, the driver's fatigue level is calculated based on the proportion of time the eyes are closed per unit time. The facial feature information includes eyelid edge feature points and pupil edge feature points. The proportion of the pupil area covered by the eyelid is calculated based on the coordinate information of the eyelid edge feature points and pupil edge feature points. When the proportion of the area reaches a first proportion threshold, it is determined that the eyes are in a closed state. The proportion of time the eyes are in a closed state per unit time is counted. When the proportion of time reaches a second proportion threshold, it is determined that the driver is driving while fatigued.

[0088] Step 800: Based on the facial feature information, perform position prediction on the (N+1)th driving image to obtain the predicted facial position of the (N+1)th driving image.

[0089] In step 800, the facial feature information includes the coordinate information of at least one feature point in the Nth driving image. A position prediction model is used to predict the coordinate information of the same feature point in the N+1th driving image based on the coordinate information of a certain feature point in the Nth driving image and the time difference between the Nth and N+1th driving images.

[0090] In this embodiment, the feature points used for location prediction include, but are not limited to, the intersection points between the face's central axis and the line connecting the two eyes; the location prediction model can be, but is not limited to, a Kalman filter model, a Markov chain model, or a particle filter model.

[0091] The driver fatigue monitoring method of the present invention introduces face position prediction in addition to face detection for feature extraction of driving images. It determines whether to use face detection or face prediction for feature extraction of the driving image based on the prediction confidence of the face position. Only when the prediction confidence of the face position does not reach the confidence threshold is face prediction used to extract features of the driving image, thereby reducing the number of times face prediction is used for feature extraction of the driving image, thus saving computing power and time. Example 2

[0092] As shown in Figure 4, an optimized driver fatigue monitoring device is used to implement the driver fatigue monitoring method described in Embodiment 1; the driver fatigue monitoring device includes:

[0093] The image acquisition module is used to acquire driving images;

[0094] The location prediction module is used to predict the face in the next driving image based on the previous driving image, so as to obtain the predicted face location.

[0095] A location detection module is used to perform face detection on the driving image to obtain the face detection location;

[0096] The feature extraction module is used to extract features from the driving image based on the predicted face location or the face detection location to obtain first feature information or second feature information;

[0097] The confidence module is used to determine the first feature information or the second feature information as facial feature information based on the predicted confidence of the first feature information.

[0098] The fatigue detection module is used to determine whether the driver is driving while fatigued based on the facial feature information. Example 3

[0099] As shown in Figure 5, an optimized driver fatigue monitoring device includes a processor and a memory connected to the processor. The memory stores a computer program for the processor to execute. When the processor executes the computer program, it performs the driver fatigue monitoring method described in Embodiment 1.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention and not to limit them. Although the embodiments of the present invention have been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the embodiments of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An optimized method for monitoring driver fatigue, characterized in that, Includes the following steps: Step 100: Obtain the Nth driving image, N≥1; Step 200: Determine whether there is a face prediction location in the Nth driving image. If the face prediction location exists, proceed to steps 300-700. If the face prediction location does not exist, proceed to steps 500-700. The face prediction location in the Nth driving image is obtained based on the prediction of the (N-1)th driving image. Step 300: Extract features from the Nth driving image based on the predicted face location to obtain first feature information; Step 400: Calculate the prediction confidence based on the first feature information, and determine whether the prediction confidence reaches the confidence threshold. If it does not reach the confidence threshold, proceed to steps 500-700. If it reaches the confidence threshold, use the first feature information as facial feature information and proceed to step 700. Step 500: Perform face detection on the Nth driving image to obtain the face detection location; Step 600: Extract features from the Nth driving image based on the face detection location to obtain second feature information, and use the second feature information as face feature information; Step 700: Determine whether the driver is driving while fatigued based on the facial feature information.

2. The driving fatigue monitoring method according to claim 1, characterized in that, In step 300, the feature extraction steps for the Nth driving image based on the predicted face location are as follows: Step 310: Crop the face in the Nth driving image according to the predicted face position to obtain the first face image; Step 320: Extract features from the first face image to obtain the first feature information.

3. The driving fatigue monitoring method according to claim 2, characterized in that, In step 310, when cropping the face in the Nth driving image, a corresponding face recognition box is set in the Nth driving image to crop the image based on the coordinate information predicted by the feature points in the Nth image and the length and width of the face in the (N-1)th driving image.

4. The driver fatigue monitoring method according to any one of claims 1-3, characterized in that, The face prediction location includes the coordinate information of at least one feature point in the Nth driving image. This coordinate information is calculated by a location prediction model based on the coordinate information of the same feature point in the N-1th driving image and the time difference between the two driving images.

5. The driving fatigue monitoring method according to claim 1, characterized in that, In step 600, the feature extraction step for the Nth driving image based on the face detection location is as follows: Step 610: Crop the face in the Nth driving image according to the face detection location to obtain the second face image; Step 620: Extract features from the second face image to obtain the second feature information.

6. The driving fatigue monitoring method according to claim 1, characterized in that, A feature extraction network is used to extract features from the first and second face images. The backbone of the feature extraction network adopts an improved MobileNet network and combines it with a DenseNet network. At the same time, some modules are replaced with lighter Ghost modules, and the accuracy is improved by adding convolutional layers.

7. The driving fatigue monitoring method according to claim 1, characterized in that, In step 700, the driver's fatigue level is calculated based on the proportion of time the eyes are closed per unit time. The facial feature information includes eyelid edge feature points and pupil edge feature points. The proportion of the pupil area covered by the eyelid is calculated based on the coordinate information of the eyelid edge feature points and pupil edge feature points. When the proportion of the area reaches a first proportion threshold, it is determined that the eyes are in a closed state. The proportion of time the eyes are in a closed state per unit time is counted. When the proportion of time reaches a second proportion threshold, it is determined that the driver is driving while fatigued.

8. The driving fatigue monitoring method according to claim 1, characterized in that, Following step 700, the following steps are also included: Step 800: Based on the facial feature information, perform position prediction on the (N+1)th driving image to obtain the predicted facial position of the (N+1)th driving image.

9. An optimized driver fatigue monitoring device, characterized in that, Used to implement the driving fatigue monitoring method as described in claim 1; The driver fatigue monitoring device includes: The image acquisition module is used to acquire driving images; The location prediction module is used to predict the face in the next driving image based on the previous driving image, so as to obtain the predicted face location. A location detection module is used to perform face detection on the driving image to obtain the face detection location; The feature extraction module is used to extract features from the driving image based on the predicted face location or the face detection location to obtain first feature information or second feature information; The confidence module is used to determine the first feature information or the second feature information as facial feature information based on the predicted confidence of the first feature information. The fatigue detection module is used to determine whether the driver is driving while fatigued based on the facial feature information.

10. An optimized driver fatigue monitoring device, characterized in that, The device includes a processor and a memory connected to the processor. The memory stores a computer program that the processor can execute. When the processor executes the computer program, it performs the driving fatigue monitoring method as described in claim 1.

Images