Kss drowsiness rating assessment method and device, computer equipment and storage medium

Abstract

The application relates to the technical field of image processing, and particularly discloses a KSS drowsiness grade evaluation method and device, computer equipment and a storage medium. The application can detect a face image of an evaluated person according to preset drowsiness grade influence factors, perform sliding window statistics on the detection result, obtain a target feature probability value, generate a target feature probability value set, and finally compare the target feature probability value with a preset KSS drowsiness grade evaluation rule to obtain a KSS drowsiness grade evaluation result. The drowsiness grade influence factors can be set as required, and the drowsiness grade of the evaluated person can be comprehensively evaluated in combination with multiple parameters. The KSS drowsiness grade evaluation rule is introduced, accidental errors can be eliminated through statistical analysis and conditional judgment of multiple data, and the accuracy and stability of the evaluation are improved. In addition, the application only needs to collect a to-be-detected image, detect the to-be-detected image and perform sliding window statistics on the detection result, so that the complexity of KSS drowsiness grade evaluation is reduced.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a KSS drowsiness level assessment method, apparatus, computer equipment, and storage medium. Background Technology

[0002] Currently, some vehicle manufacturers in the automotive industry have begun to introduce Driver Drowsiness and Attention Warning (DDAW) systems. These systems typically include sensor- and camera-based technologies to monitor the driver's facial expressions, eye movements, and vehicle status to assess the driver's alertness and fatigue levels. DDAW uses the Karolinska Sleepiness Scale (KSS), a commonly used self-report scale for assessing an individual's subjective level of drowsiness. However, existing Driver Drowsiness and Attention Warning systems face some challenges in accurately assessing KSS drowsiness levels: 1. Subjectivity and misjudgment issues: Recognition based on facial expressions and eye movements is often subjective, as different drivers may behave differently, which affects the accuracy of the assessment results.

[0003] 2. Operational complexity: Some systems may be too complex, requiring a large number of sensors and computing resources to accurately assess drowsiness levels, increasing the cost and complexity of the system.

[0004] 3. Insufficient real-time performance: Some systems do not react quickly enough when warning the driver, and cannot detect the driver's fatigue in time, which affects the effectiveness of the system.

[0005] Therefore, how to reduce the complexity of KSS drowsiness level assessment and improve its accuracy and real-time performance has become an urgent problem to be solved. Summary of the Invention

[0006] This application provides a method, apparatus, computer device, and storage medium for assessing KSS drowsiness levels, in order to reduce the complexity of KSS drowsiness level assessment and improve the accuracy and real-time performance of KSS drowsiness level assessment.

[0007] Firstly, this application provides a KSS drowsiness level assessment method, the method comprising: Acquire continuous facial images of the person to be evaluated as the images to be detected; Each frame of the image to be detected is inspected to obtain face detection information, eye detection information, expression detection information, and face angle detection information; Based on a preset threshold, sliding window statistics are performed on the target features present in the face detection information, the eye detection information, the expression detection information, and the face angle detection information to obtain the probability value of each target feature and generate a set of probability values. Based on the preset KSS drowsiness level assessment rules, the probability value of at least one of the target features is obtained from the probability value set, and the probability value of each target feature is compared with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed.

[0008] Secondly, this application also provides a KSS drowsiness level assessment device, the device comprising: The image acquisition module is used to acquire continuous facial images of the person to be evaluated as images to be detected. The detection information acquisition module is used to detect each frame of the image to be detected and obtain face detection information, eye detection information, expression detection information and face angle detection information; The probability value set acquisition module is used to perform sliding window statistics on the target features present in the face detection information, the eye detection information, the expression detection information, and the face angle detection information based on a preset threshold, to obtain the probability value of each target feature and generate a probability value set. The KSS drowsiness level assessment module is used to obtain the probability value of at least one of the target features from the probability value set based on the preset KSS drowsiness level assessment rules, and compare the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed.

[0009] Thirdly, this application also provides a computer device, the computer device including a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and implement the KSS drowsiness level assessment method as described above when executing the computer program.

[0010] Fourthly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the KSS drowsiness level assessment method as described above.

[0011] This application discloses a KSS drowsiness level assessment method, apparatus, computer equipment, and storage medium. The method involves acquiring continuous facial images of the person to be assessed as images to be detected; detecting each frame of the image to be detected to obtain face detection information, eye detection information, expression detection information, and face angle detection information; based on preset thresholds, performing sliding window statistics on target features present in the face detection information, eye detection information, expression detection information, and face angle detection information to obtain probability values ​​for each target feature and generate a probability value set; based on preset KSS drowsiness level assessment rules, obtaining the probability value of at least one target feature from the probability value set, and comparing the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed. This application can detect facial images of the person being evaluated based on preset factors influencing drowsiness levels, obtain detection results, and perform sliding window statistics on the detection results to obtain accurate probability values ​​of target features, generating a set of target feature probability values. Finally, the target feature probability values ​​are compared with preset KSS drowsiness level evaluation rules to obtain the final KSS drowsiness level evaluation result. Drowsiness level influencing factors can be set as needed, combining multiple parameters to comprehensively evaluate the drowsiness level of the person being evaluated. Introducing KSS drowsiness level evaluation rules allows for statistical analysis and conditional judgment of various data, eliminating random errors and improving the accuracy and stability of the evaluation. Furthermore, this application only requires collecting the image to be detected, performing detection on the image, and performing sliding window statistics on the detection results, reducing the complexity of KSS drowsiness level evaluation. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a schematic flowchart of a first embodiment of the KSS drowsiness level assessment method provided in this application; Figure 2 This is a schematic flowchart of a drowsiness level assessment method provided in an embodiment of this application. Figure 3 This is a schematic flowchart of a second embodiment of the KSS drowsiness level assessment method provided in this application; Figure 4 A flowchart illustrating the dynamic learning of the human eye closing threshold for a KSS drowsiness level assessment method provided in an embodiment of this application; Figure 5 A schematic block diagram of a KSS drowsiness level assessment device provided for embodiments of this application; Figure 6 A schematic block diagram of the structure of a computer device provided for an embodiment of this application. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0016] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0017] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0018] This application provides a KSS drowsiness level assessment method, apparatus, computer device, and storage medium. The KSS drowsiness level assessment method can be applied to a server, allowing for the setting of drowsiness level influencing factors as needed, and combining multiple parameters to comprehensively assess the drowsiness level of the person being assessed. By introducing KSS drowsiness level assessment rules, statistical analysis and conditional judgment of various data can eliminate random errors, improving the accuracy and stability of the assessment. Furthermore, this application only requires the acquisition of the image to be detected, the detection of the image, and the sliding window statistics of the detection results, reducing the complexity of the KSS drowsiness level assessment. The server can be a standalone server or a server cluster.

[0019] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0020] Please see Figure 1 , Figure 1 This is a schematic flowchart illustrating a KSS drowsiness level assessment method provided in an embodiment of this application. This KSS drowsiness level assessment method can be applied to servers, allowing for the setting of drowsiness level influencing factors as needed, and combining multiple parameters to comprehensively assess the drowsiness level of the person being assessed. By introducing KSS drowsiness level assessment rules, statistical analysis and conditional judgment of various data can eliminate random errors, improving the accuracy and stability of the assessment. Furthermore, this application only requires the acquisition of the image to be detected, the detection of the image, and the sliding window statistics of the detection results, reducing the complexity of the KSS drowsiness level assessment.

[0021] like Figure 1 As shown, the KSS drowsiness level assessment method specifically includes steps S101 to S104.

[0022] S101. Obtain continuous facial images of the person to be evaluated as images to be detected; In one embodiment, an infrared camera is invoked to capture an image including the face of the person to be evaluated, which is then used as the image to be detected.

[0023] In one embodiment, the person being assessed may be a driver, a machine operator, or another person who needs to undergo a KSS drowsiness level assessment.

[0024] S102. Detect each frame of the image to be detected to obtain face detection information, eye detection information, expression detection information and face angle detection information; In one embodiment, factors influencing drowsiness levels may include facial angle features, yawning features, left eye closing features, and right eye closing features, and may also include other features set by the user according to actual needs.

[0025] In one embodiment, the image to be processed is fed into various detection algorithms to obtain basic features such as the face and eyes. Specifically, a face detection algorithm is used to detect face information, including face score, face location, and 106 feature points. An eye detection algorithm is used to detect eye information, including left and right eye scores and eye aspect ratio. A yawn detection algorithm is used to detect facial expression information, including mouth score and yawn score. A face angle detection algorithm is used to detect face angle information, including scores in the four directions (up, down, left, and right) of the face.

[0026] S103. Based on a preset threshold, perform sliding window statistics on the target features present in the face detection information, the eye detection information, the expression detection information, and the face angle detection information respectively to obtain the probability value of each target feature and generate a probability value set. Further, the step of performing sliding window statistics on target features present in the face detection information, eye detection information, expression detection information, and face angle detection information based on a preset threshold to obtain the probability value of each target feature includes: adding the face detection information, eye detection information, expression detection information, and face angle detection information of each image to be detected into preset short-term sliding windows and long-term sliding windows according to the frame number of the image to be detected; judging and statistically analyzing whether each target feature exists in the face detection information, eye detection information, expression detection information, and face angle detection information based on the short-term sliding window, the long-term sliding window, and the preset threshold to obtain the statistical results of the short-term sliding window and the statistical results of the long-term sliding window; and performing probability calculations based on the statistical results of the short-term sliding window and the long-term sliding window to obtain the short-term probability value of each target feature in the short-term sliding window and the long-term probability value of each target feature in the long-term sliding window.

[0027] In one embodiment, multi-frame statistics are performed on a frame-by-frame basis, divided into short-term and long-term groups, denoted as shortTermData and longTermData, respectively. ShortTermData uses a 20-frame sliding window, and longTermData uses a 600-frame sliding window. The probability values ​​of each feature in the short term can be calculated using the feature statistics of shortTermData, and the probability values ​​of each feature in the long term can be calculated using the feature statistics of longTermData.

[0028] In one embodiment, the sliding window statistics can include the following 13 features: face presence, face to the left, face to the right, head down, head up, left eye present, left eye closed, left eye open, right eye present, right eye closed, right eye open, yawning, and rubbing eyes.

[0029] In one embodiment, facial angle features may include a face tilted to the left, a face tilted to the right, a head tilted down, or a head tilted up.

[0030] In one embodiment, for the statistics of facial presence features: if the face score is greater than a threshold, the face is considered to exist and added to the statistics queue. Otherwise, the face is considered not to exist and added to the statistics queue. For the statistics of left-leaning face features: the face angle in the detected face detection results is added to the statistics queue, and the probability of the face angle reaching a given angle threshold is calculated. The threshold can be adjusted according to the algorithm used. Other category statistics: such as "face leaning to the right", "head down", "head up", "left eye closed", "right eye closed", etc., refer to the similar implementation method of "face leaning to the left". Eye rubbing logic inference: when a face is detected, if the detected left or right eye score is lower than the threshold, it is considered that no eye has been detected, i.e., eye rubbing.

[0031] In one embodiment, the probability value set of the target feature stores the short-term probability value and the long-term probability value corresponding to each target feature. The short-term probability value is the probability obtained by short-term sliding window statistics, and the long-term probability value is the probability obtained by long-term sliding window statistics. Examples include the long-term and short-term probability values ​​of features such as face leaning to the right, face leaning to the left, head down, head up, presence of left eye, left eye closed, left eye open, right eye present, right eye closed, right eye open, yawning, and rubbing eyes.

[0032] In one embodiment, a sliding window statistical method is used to achieve real-time and accurate fatigue detection. Whenever a new frame of data is detected, the data is updated within the sliding window, while the oldest frame of data is removed.

[0033] In one embodiment, two sliding windows are used: a short-term sliding window and a long-term sliding window. The short-term sliding window contains the most recent 20 frames of data. The long-term sliding window contains the most recent 600 frames of data.

[0034] In one embodiment, for each sliding window, the frequency of feature occurrence for each category is counted. For example: If a face is detected in 10 frames within a short sliding window (20 frames), the probability of a face being present is 50%. The statistical methods for other features (such as closed eyes, yawning, etc.) are similar.

[0035] In one embodiment, each time a new frame is detected, the data of the new frame is added to the sliding window, and the oldest frame data in the sliding window is removed. Then, the occurrence probability of each category within the sliding window is recalculated.

[0036] In one embodiment, a sliding window statistical mechanism allows for rapid response to short-term changes in the driver's condition. A long-term sliding window can be used to assess the driver's overall fatigue trend. This dynamic, real-time assessment of the fatigue state of the person being evaluated, combined with more accurate judgments based on both short-term and long-term statistical data, enables a more comprehensive assessment.

[0037] S104. Based on the preset KSS drowsiness level assessment rules, obtain the probability value of at least one of the target features from the probability value set, and compare the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed.

[0038] In one embodiment, the KSS level is simplified to five levels: Level 1: LEVEL_03, Level 2: LEVEL_02, Level 3: LEVEL_01, Level 4: LEVEL_00, and Level 5: LEVEL_NONE. These represent: LEVEL_NONE: alert; LEVEL_00: neither alert nor drowsy; LEVEL_01: somewhat drowsy; LEVEL_02: sleepy, but able to stay awake with effort; and LEVEL_03: very drowsy.

[0039] In one embodiment, short-term assessment: assesses the immediate drowsiness level of the person being assessed based on short-term statistical data. Long-term assessment: assesses the overall fatigue trend of the person being assessed based on long-term statistical data.

[0040] Furthermore, before obtaining the probability value of at least one of the target features from the probability value set based on the preset KSS drowsiness level assessment rule, and comparing the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, the method further includes: obtaining the long-term probability value of facial presence features based on the probability value set; when the long-term probability value of facial presence features is less than the preset facial presence probability threshold, re-acquiring the image to be detected and detecting the probability of the target features, so that when the long-term probability value of facial presence features is greater than or equal to the facial presence threshold, based on the preset KSS drowsiness level assessment rule, obtaining the probability value of at least one of the target features from the probability value set, and comparing the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed.

[0041] In one embodiment, a face presence check is first performed: if the probability of a face presence in long-term data is lower than a threshold, i.e., the long-term probability value of the face presence feature is less than a preset face presence probability threshold, the detection is considered to have failed, and no further evaluation is performed. The image to be detected needs to be acquired again and detected.

[0042] In one embodiment, such as Figure 2 As shown, Figure 2 This is a flowchart illustrating the drowsiness level assessment. The drowsiness level assessment rules include: face presence check; high fatigue state: if the probability of yawning or eye closure exceeds a threshold in long-term data, it is judged as a high fatigue state; special fatigue state: if the probability of eye closure is abnormally high in long-term data, it may indicate prolonged periods without blinking, and is judged as a special fatigue state; multiple fatigue states: combining short-term and long-term data to determine if there are multiple fatigue indicators, such as the probability of eye closure and yawning simultaneously exceeding the threshold; face position threshold exceeding the limit state; and high eye closure state.

[0043] Further, the step of obtaining the probability value of at least one of the target features from the probability value set based on the preset KSS drowsiness level assessment rule, and comparing the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, includes: based on the KSS drowsiness level assessment rule, obtaining the long-term probability value of yawning, the long-term probability value of left eye presence and left eye closed, the long-term probability value of right eye presence and right eye closed, the long-term probability value of left eye presence and left eye open, and the long-term probability value of right eye presence and right eye open from the probability value set. The probability values ​​include: long-term probability values ​​of features such as left-leaning face, right-leaning face, head-down, head-up, eye-rubbing, left-eye presence, left-eye closed, right-eye presence, and right-eye closed; and long-term probability values ​​of features such as yawning, left-eye presence and left-eye closed, right-eye presence and right-eye closed, left-eye presence and left-eye open, and right-eye presence and right-eye open. The probability values ​​of the left-leaning yawning feature, the right-leaning yawning feature, the left-leaning eye closing feature, and the right-leaning eye closing feature are compared with preset first yawning probability thresholds, preset first left-leaning eye opening probability thresholds, preset first right-leaning eye opening probability thresholds, and preset first right-leaning eye opening probability thresholds, respectively, to obtain a first comparison result. The long-term probability values ​​of the left-leaning face feature, the right-leaning face feature, the head-down feature, the head-up feature, and the eye-rubbing feature are compared with preset first left-leaning probability thresholds, preset first right-leaning probability thresholds, preset first head-down probability thresholds, and preset head-up probability thresholds, respectively. The first comparison result is obtained by comparing the probability of rubbing the eyes with a preset first probability threshold. The second comparison result is obtained by comparing the long-term probability value of the left eye presence feature, the short-term probability value of the left eye closure feature, the long-term probability value of the right eye presence feature, and the short-term probability value of the right eye closure feature with preset left eye presence probability thresholds, preset second left eye closure probability thresholds, preset right eye presence probability thresholds, and preset second right eye closure probability thresholds. The third comparison result is obtained by comparing the first comparison result, the second comparison result, and the third comparison result. The drowsiness level of the person to be evaluated is determined.

[0044] In one embodiment, if the first comparison result shows that the long-term probability value of the yawning feature is greater than the first yawning probability threshold, or the long-term probability value of the left eye being present and the left eye being closed is greater than the first left eye closed probability threshold, or the long-term probability value of the right eye being present and the right eye being closed is greater than the first right eye closed probability threshold, then the person to be evaluated is determined to be in a state of high fatigue and the drowsiness level is determined to be Level 1. If the long-term probability value of the left eye being present and the left eye being open is greater than the first left eye open probability threshold, or the long-term probability value of the right eye being present and the right eye being open is greater than the first right eye open probability threshold, then the person to be evaluated is determined to be in a state of special fatigue and the drowsiness level is determined to be Level 1.

[0045] In a specific embodiment, a very high fatigue index is defined as follows: if the probability of yawning in longTermData exceeds a preset yawning probability threshold (e.g., 0.6), or if the left eye is present and the probability of closing the eye exceeds a preset left eye closing threshold (e.g., 0.80), or if the right eye is present and the probability of closing the eye exceeds a preset right eye closing probability threshold (e.g., 0.80), then the fatigue level is determined to be very high, and the drowsiness level is increased to a maximum of LEVEL_03.

[0046] In a specific embodiment, a special fatigue indicator is defined as follows: if the left eye is present in longTermData and its open-eye probability exceeds a preset threshold (e.g., 0.995), or the right eye is present and its open-eye probability exceeds a preset threshold (e.g., 0.995), then a special fatigue state is identified, and the drowsiness level is increased to a maximum of LEVEL_03. This rule addresses the special case of prolonged periods without blinking.

[0047] In one embodiment, if the second comparison result is that the long-term probability value of the face-left-leaning feature is greater than the first left-leaning probability threshold, or the long-term probability value of the face-right-leaning feature is greater than the first right-leaning probability threshold, or the long-term probability value of the head-down feature is greater than the first head-down probability threshold, or the long-term probability value of the head-up feature is greater than the first head-up probability threshold, or the long-term probability value of the eye-rubbing feature is greater than the first eye-rubbing probability threshold, then the drowsiness level of the person to be evaluated is determined to be the first level.

[0048] In a specific embodiment, if the probability of a face being tilted to the left, right, looking down, looking up, or rubbing its eyes in longTermData exceeds a preset probability threshold (e.g., 0.8), it is determined that the face position threshold has exceeded the limit, and the drowsiness level is increased to a maximum of LEVEL_03.

[0049] In one embodiment, if the third comparison result is that the long-term probability value of the left eye presence feature is greater than the preset left eye presence probability threshold and the short-term probability value of the left eye closed feature is greater than the preset second left eye closed probability threshold, or the long-term probability value of the right eye presence feature is greater than the preset right eye presence probability threshold and the short-term probability value of the right eye closed feature is greater than the preset second right eye closed probability threshold, the drowsiness level of the person to be evaluated is determined to be the second level.

[0050] In one embodiment, a high probability of eye closure occurs when the left eye is present in longTermDat and the probability of the left eye being closed in shortTermData exceeds a preset threshold for the probability of the left eye being closed (e.g., 0.99), or when the right eye is present and the probability of the right eye being closed in shortTermData exceeds a preset threshold for the probability of the right eye being closed (e.g., 0.99), then the drowsiness level is increased to a maximum of LEVEL_02.

[0051] Further, the step of obtaining the probability value of at least one of the target features from the probability value set based on the preset KSS drowsiness level assessment rule, and comparing the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, also includes: based on the KSS drowsiness level assessment rule, comparing the long-term probability value of the left eye being present and the left eye being closed, the long-term probability value of the right eye being present and the right eye being closed, and the long-term probability value of the yawning feature with preset third left eye closed probability thresholds, preset third right eye closed probability thresholds, and preset third yawning probability thresholds, respectively; if the long-term probability value of the left eye being present and the left eye being closed is greater than the third left eye closed probability threshold, or the long-term probability value of the right eye being present and the right eye being closed is greater than the third left eye closed probability threshold, or the long-term probability value of the right eye being present and the right eye being closed is greater than the third left eye closed probability threshold, then the long-term probability value of the left eye being present and the right eye being closed is greater than the third left eye closed probability threshold. If the long-term probability value of the yawning feature is greater than the third right eye closing probability threshold, or the long-term probability value of the yawning feature is greater than the third yawning probability threshold, then the short-term probability values ​​of the left eye closing feature, the right eye closing feature, the yawning feature, and the eye rubbing feature are obtained from the probability value set. The short-term probability values ​​of the left eye closing feature, the right eye closing feature, the yawning feature, and the eye rubbing feature are compared with preset fourth left eye closing probability threshold, preset fourth right eye closing probability threshold, preset fourth yawning probability threshold, and preset second eye rubbing probability threshold, respectively, to obtain a fourth comparison result. Based on the fourth comparison result, the drowsiness level of the person to be evaluated is determined.

[0052] In one embodiment, if the fourth comparison result is that the short-term probability value of the left eye closing feature is greater than the fourth left eye closing probability threshold, or the short-term probability value of the right eye closing feature is greater than the fourth right eye closing probability threshold, or the short-term probability value of the yawning feature is greater than the fourth yawning probability threshold, or the short-term probability value of the eye rubbing feature is greater than the second eye rubbing probability threshold, then the drowsiness level of the person to be evaluated is determined to be level one; otherwise, the drowsiness level of the person to be evaluated is determined to be level two.

[0053] In a specific embodiment, multiple fatigue indicators (high threshold): If the left eye is present in longTermData and the probability of eye closure exceeds a preset left eye closure threshold (e.g., 0.50), or the right eye is present and the probability of eye closure exceeds a preset right eye closure threshold (e.g., 0.50), or the probability of yawning exceeds a preset yawning probability threshold (e.g., 0.4), then if the probability of the left eye being closed in shortTermData exceeds a preset left eye closure probability threshold (e.g., 0.8), or the probability of the right eye being closed exceeds a preset right eye closure probability threshold (e.g., 0.8), or the probability of rubbing eyes exceeds a preset eye rubbing probability threshold (e.g., 0.8), or the probability of yawning exceeds a preset yawning probability threshold (e.g., 0.8), then the drowsiness level is increased to a maximum of LEVEL_03. Otherwise, the drowsiness level is increased to a maximum of LEVEL_02.

[0054] Further, the step of obtaining the probability value of at least one of the target features from the probability value set based on the preset KSS drowsiness level assessment rule, and comparing the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, also includes: based on the KSS drowsiness level assessment rule, comparing the long-term probability value of the left eye being present and the left eye being closed, the long-term probability value of the right eye being present and the right eye being closed, and the long-term probability value of the yawning feature with preset fifth left eye closed probability thresholds, preset fifth right eye closed probability thresholds, and preset fifth yawning probability thresholds, respectively: if the long-term probability value of the left eye being present and the left eye being closed is greater than the fifth left eye closed probability threshold, or the long-term probability value of the right eye being present and the right eye being closed is greater than the fifth left eye closed probability threshold, or the long-term probability value of the right eye being present and the right eye being closed is greater than the fifth left eye closed probability threshold, then the long-term probability value of the left eye being present and the right eye being closed is greater than the fifth left eye closed probability threshold. If the long-term probability value of the yawning feature is greater than the fifth right eye closing probability threshold, or the long-term probability value of the yawning feature is greater than the fifth yawning probability threshold, then the short-term probability values ​​of the left eye closing feature, the right eye closing feature, the yawning feature, and the eye rubbing feature are obtained from the probability value set. The short-term probability values ​​of the left eye closing feature, the right eye closing feature, the yawning feature, and the eye rubbing feature are compared with the preset sixth left eye closing probability threshold, the preset sixth right eye closing probability threshold, the preset sixth yawning probability threshold, and the preset third eye rubbing probability threshold, respectively, to obtain a fifth comparison result. Based on the fifth comparison result, the drowsiness level of the person to be evaluated is determined.

[0055] In one embodiment, if the fifth comparison result is that the short-term probability value of the left eye closing feature is greater than the sixth left eye closing probability threshold, or the short-term probability value of the right eye closing feature is greater than the sixth right eye closing probability threshold, or the short-term probability value of the yawning feature is greater than the sixth yawning probability threshold, or the short-term probability value of the eye rubbing feature is greater than the third eye rubbing probability threshold, then the drowsiness level of the person to be evaluated is determined to be level two; otherwise, the drowsiness level of the person to be evaluated is determined to be level three.

[0056] In a specific embodiment, multiple fatigue indicators (medium threshold): If the left eye exists in longTermData and the probability of closing the eye exceeds a preset left eye closing probability threshold (e.g., 0.30), or the right eye exists and the probability of closing the eye exceeds a preset right eye closing probability threshold (e.g., 0.30), or the probability of yawning exceeds a preset yawning probability threshold (e.g., 0.3), then if the probability of the left eye closing in shortTermData exceeds a preset left eye closing probability threshold (e.g., 0.8), or the probability of the right eye closing exceeds a preset right eye closing probability threshold (e.g., 0.8), or the probability of rubbing the eyes exceeds a preset eye rubbing probability threshold (e.g., 0.8), or the probability of yawning exceeds a preset yawning probability threshold (e.g., 0.8), then the drowsiness level is increased to a maximum of LEVEL_02. Otherwise, the drowsiness level is increased to a maximum of LEVEL_01.

[0057] Further, the step of obtaining the probability value of at least one of the target features from the probability value set based on the preset KSS drowsiness level assessment rule, and comparing the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, also includes: based on the KSS drowsiness level assessment rule, comparing the long-term probability value of the left eye being present and the left eye being closed, the long-term probability value of the right eye being present and the right eye being closed, and the long-term probability value of the yawning feature with preset seventh left eye closed probability thresholds, preset seventh right eye closed probability thresholds, and preset seventh yawning probability thresholds, respectively: if the long-term probability value of the left eye being present and the left eye being closed is greater than the seventh left eye closed probability threshold, or the long-term probability value of the right eye being present and the right eye being closed is greater than the seventh left eye closed probability threshold, or the long-term probability value of the right eye being present and the right eye being closed is greater than the seventh left eye closed probability threshold, then the long-term probability value of the left eye being present and the right eye being closed is greater than the seventh left eye closed probability threshold. If the long-term probability value of the yawning feature is greater than the seventh right eye closing probability threshold, or the long-term probability value of the yawning feature is greater than the seventh yawning probability threshold, then the short-term probability values ​​of the left eye closing feature, the right eye closing feature, the yawning feature, and the eye rubbing feature are obtained from the probability value set. The short-term probability values ​​of the left eye closing feature, the right eye closing feature, the yawning feature, and the eye rubbing feature are compared with the preset eighth left eye closing probability threshold, the preset eighth right eye closing probability threshold, the preset eighth yawning probability threshold, and the preset fourth eye rubbing probability threshold, respectively, to obtain a sixth comparison result. Based on the sixth comparison result, the drowsiness level of the person to be evaluated is determined.

[0058] In one embodiment, if the sixth comparison result is that the short-term probability value of the left eye closing feature is greater than the eighth left eye closing probability threshold, or the short-term probability value of the right eye closing feature is greater than the eighth right eye closing probability threshold, or the short-term probability value of the yawning feature is greater than the eighth yawning probability threshold, or the short-term probability value of the eye rubbing feature is greater than the fourth eye rubbing probability threshold, then the drowsiness level of the person to be evaluated is determined to be level three; otherwise, the drowsiness level of the person to be evaluated is determined to be level four.

[0059] In a specific embodiment, multiple fatigue indicators (low threshold): If the left eye is present in longTermData and the probability of eye closure exceeds a preset left eye closure threshold (e.g., 0.20), or the right eye is present and the probability of eye closure exceeds a preset right eye closure threshold (e.g., 0.20), or the probability of yawning exceeds 0.2, then if the left eye closure probability in shortTermData exceeds a preset left eye closure threshold (e.g., 0.8), or the right eye closure probability exceeds a preset right eye closure threshold (e.g., 0.8), or the probability of rubbing eyes exceeds a preset eye rubbing probability threshold (e.g., 0.8), or the probability of yawning exceeds a preset yawning probability threshold (e.g., 0.8), then the drowsiness level is increased to a maximum of LEVEL_01. Otherwise, the drowsiness level is increased to a maximum of LEVEL_00.

[0060] Further, the step of obtaining the probability value of at least one of the target features from the probability value set based on the preset KSS drowsiness level assessment rule, and comparing the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, also includes: obtaining the short-term probability value of the left-leaning face feature, the short-term probability value of the right-leaning face feature, the short-term probability value of the head-down feature, and the short-term probability value of the head-up feature based on the KSS drowsiness level assessment rule; and obtaining the long-term probability values ​​of the eye-rubbing feature, the yawning feature, the left eye present and closed, the right eye present and closed, the left-leaning face feature, the right-leaning face feature, the head-down feature, the head-up feature, and the left eye closed feature. The short-term probability values ​​of eye features, the short-term probability value of the right eye closing feature, the short-term probability value of the face left-leaning feature, the short-term probability value of the face right-leaning feature, the short-term probability value of the head-down feature, and the short-term probability value of the head-up feature are compared with preset sixth eye-rubbing probability thresholds, preset ninth yawning probability thresholds, preset ninth left eye closing thresholds, preset ninth right eye closing thresholds, preset second left-leaning probability thresholds, preset second right-leaning probability thresholds, preset second head-down probability thresholds, preset second head-up probability thresholds, preset tenth left eye closing probability thresholds, preset tenth right eye closing probability thresholds, preset third left-leaning probability thresholds, preset third right-leaning probability thresholds, preset third head-down probability thresholds, and preset third head-up probability thresholds to obtain a seventh comparison result; based on the seventh comparison result, the drowsiness level of the person to be evaluated is determined.

[0061] In one embodiment, if the seventh comparison result is that the long-term probability value of the yawning feature is greater than the ninth yawning probability threshold, and the long-term probability value of the eye rubbing feature is greater than the sixth eye rubbing probability threshold, and the long-term probability value of the left eye being present and the left eye being closed is greater than the ninth left eye closed probability threshold, or the long-term probability value of the right eye being present and the right eye being closed is greater than the ninth right eye closed probability threshold, or the long-term probability value of the face being left-leaning feature is greater than the second left-leaning probability threshold, or the long-term probability value of the face being right-leaning feature is greater than the second right-leaning probability threshold, or the long-term probability value of the head-down feature is greater than the second head-down feature. If the long-term probability value of the head-up feature is greater than the second head-up probability threshold, or the short-term probability value of the left eye closed feature is greater than the tenth left eye closed probability threshold, or the short-term probability value of the right eye closed feature is greater than the tenth right eye closed probability threshold, or the short-term probability value of the face-left-leaning feature is greater than the third left-leaning probability threshold, or the short-term probability value of the face-right-leaning feature is greater than the third right-leaning probability threshold, or the short-term probability value of the head-down feature is greater than the third head-down probability threshold, or the short-term probability value of the head-up feature is greater than the third head-up probability threshold, then the drowsiness level of the person to be evaluated is determined to be level three.

[0062] In a specific embodiment, the threshold for closing eyes or yawning exceeds the following: If the probability of rubbing eyes in longTermData exceeds a preset threshold for rubbing eyes (e.g., 0.05), and the probability of yawning exceeds a preset threshold for yawning (e.g., 0.05), and the left eye is present and the probability of closing eyes exceeds a preset threshold for closing eyes in the left eye (e.g., 0.1), or the right eye is present and the probability of closing eyes exceeds a preset threshold for closing eyes in the right eye (e.g., 0.1), or the probability of the face being tilted to the left, right, down, or up exceeds a preset probability threshold of 0.1, or the probability of the left eye being closed in shortTermData exceeds a preset threshold for closing eyes in the left eye (e.g., 0.6), or the probability of the right eye being closed exceeds a preset threshold for closing eyes in the right eye (e.g., 0.6), or the probability of the face being tilted to the left, right, down, or up exceeds a preset probability threshold of 0.8, then the drowsiness level is increased to a maximum of LEVEL_01. This rule mainly considers the situation where multiple fatigue characteristics occur simultaneously, but the probability values ​​are low.

[0063] The above embodiments provide a KSS drowsiness level assessment method, apparatus, computer equipment, and storage medium. The method involves acquiring continuous facial images of the person to be assessed as images to be detected; detecting each frame of the image to be detected to obtain face detection information, eye detection information, expression detection information, and face angle detection information; based on preset thresholds, performing sliding window statistics on target features present in the face detection information, eye detection information, expression detection information, and face angle detection information to obtain the probability value of each target feature and generate a probability value set; based on preset KSS drowsiness level assessment rules, obtaining the probability value of at least one target feature from the probability value set, and comparing the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed. This application can detect facial images of the person being evaluated based on preset factors influencing drowsiness levels, obtain detection results, perform sliding window statistics on the detection results to obtain accurate probability values ​​of target features, and generate a set of target feature probability values. Finally, the target feature probability values ​​are compared with preset KSS drowsiness level assessment rules to obtain the final KSS drowsiness level assessment result. Drowsiness level influencing factors can be set as needed, combining multiple parameters to comprehensively assess the drowsiness level of the person being evaluated. Introducing KSS drowsiness level assessment rules allows for statistical analysis and conditional judgment of various data, and comprehensive judgment based on multiple fatigue indicator parameters, thereby more accurately assessing the driver's fatigue level and issuing timely warnings; eliminating random errors and improving the accuracy and stability of the assessment. Furthermore, this application only requires collecting the image to be detected, performing detection on the image, and performing sliding window statistics on the detection results, reducing the complexity of KSS drowsiness level assessment.

[0064] Please see Figure 3 , Figure 3 This is a schematic flowchart illustrating a KSS drowsiness level assessment method provided in an embodiment of this application. This KSS drowsiness level assessment method can be applied in a server to dynamically adjust the eye-closing threshold based on the actual situation of the person being assessed, adapting to different individuals and eye sizes, thereby improving the sensitivity and accuracy of fatigue detection.

[0065] like Figure 3 As shown, the KSS drowsiness level assessment method specifically includes steps S201 to S204.

[0066] S201. Based on the human eye detection information, obtain the left eye aspect ratio data and the right eye aspect ratio data of the person to be evaluated; S202. Perform frequency distribution analysis on the aspect ratio data of the left eye and the aspect ratio data of the right eye respectively to obtain the first target aspect ratio interval corresponding to the left eye and the second target aspect ratio interval corresponding to the right eye. S203. Based on the preset scaling factor and the first median value of the first target aspect ratio interval and the second median value of the second target aspect ratio interval, the first target eye-closing threshold and the second target eye-closing threshold are obtained respectively. S204. When the first target eye-closing threshold is within the preset eye-closing threshold range, the left eye eye-closing threshold is updated to the first target eye-closing threshold, and when the second target eye-closing threshold is within the eye-closing threshold range, the right eye eye-closing threshold is updated to the second target eye-closing threshold.

[0067] In one embodiment, obtaining a first target eye-closing threshold and a second target eye-closing threshold based on a preset scaling factor, the first target aspect ratio interval, and the second target aspect ratio interval includes: obtaining a first median value of the first target aspect ratio interval and a second median value of the second target aspect ratio interval; obtaining the first target eye-closing threshold based on the first median value and the preset scaling factor; and obtaining the second target eye-closing threshold based on the second median value and the preset scaling factor.

[0068] In one embodiment, the eye-closing threshold is dynamically adjusted based on real-time detected eye aspect ratio data to improve the accuracy of fatigue detection. The appropriate eye-closing threshold for the current person being assessed is determined by statistically analyzing the frequency distribution of eye aspect ratio.

[0069] In specific embodiments, such as Figure 4 As shown, Figure 4This is a flowchart illustrating the dynamic learning of the human eye closing threshold. 1. Initialization: During initialization, the statistical interval and the number of statistical sliding window frames are set. The statistical interval refers to the unit interval of the human eye size score. A default closing threshold range is set; for example, the open-eye threshold is greater than 0.2, and vice versa.

[0070] 2. Data Update: Real-time collection of aspect ratio data for the left and right eyes, stored separately according to eye type (left or right eye). A sliding window mechanism is used to store the aspect ratio data for the most recent period, ensuring the real-time nature and validity of the statistical data.

[0071] 3. Dynamic Threshold Calculation: Within the sliding window, frequency distribution analysis is performed on the collected eye aspect ratio data to identify the most frequently occurring aspect ratio intervals, which are then designated as target aspect ratio intervals. The left eye corresponds to the first target aspect ratio interval, and the right eye corresponds to the second target aspect ratio interval. Based on the most frequent aspect ratio intervals, a dynamic eye-closing threshold is calculated. Specifically, the median of this interval is multiplied by a preset ratio (e.g., 0.6) to obtain the new eye-closing threshold. Ensure that the calculated eye-closing threshold falls within the preset threshold range (e.g., between 0.1 and 0.2). If it meets the condition, the dynamic eye-closing threshold is updated. If it is outside the preset threshold range, it is considered an anomaly and should be discarded.

[0072] 4. Threshold Update: Dynamic thresholds are calculated and updated separately for the left and right eyes to ensure that the threshold for each eye can be adjusted independently to adapt to different eye characteristics. The calculated dynamic thresholds are output for subsequent fatigue detection.

[0073] In the above embodiments, the eye-closing threshold can be dynamically adjusted according to the actual situation of the person being evaluated, adapting to different people being evaluated and different eye sizes, thereby improving the sensitivity and accuracy of fatigue detection.

[0074] Please see Figure 5 , Figure 5 This application provides a schematic block diagram of a KSS drowsiness level assessment device, which is used to perform the aforementioned KSS drowsiness level assessment method. The KSS drowsiness level assessment device can be configured on a server.

[0075] like Figure 5 As shown, the KSS drowsiness level assessment device 400 includes: The image acquisition module 401 is used to acquire continuous facial images of the person to be evaluated as images to be detected. The detection information acquisition module 402 is used to detect each frame of the image to be detected and obtain face detection information, eye detection information, expression detection information and face angle detection information. The probability value set acquisition module 403 is used to perform sliding window statistics on the target features present in the face detection information, the eye detection information, the expression detection information and the face angle detection information based on a preset threshold, to obtain the probability value of each target feature and generate a probability value set. The KSS drowsiness level assessment module 404 is used to obtain the probability value of at least one of the target features from the probability value set based on the preset KSS drowsiness level assessment rules, and compare the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed.

[0076] Furthermore, the preset thresholds include a left eye closure threshold and a right eye closure threshold. The KSS drowsiness level assessment device also includes a dynamic threshold generation module, which includes: The data acquisition unit is used to acquire the left eye aspect ratio data and the right eye aspect ratio data of the person to be evaluated based on the human eye detection information. The interval determination unit is used to perform frequency distribution analysis on the left eye aspect ratio data and the right eye aspect ratio data respectively, to obtain the first target aspect ratio interval corresponding to the left eye and the second target aspect ratio interval corresponding to the right eye. The threshold acquisition unit is used to obtain the first target closed-eye threshold and the second target closed-eye threshold based on a preset scaling factor, the first median of the first target aspect ratio interval, and the second median of the second target aspect ratio interval. The threshold update unit is configured to update the left eye closing threshold to the first target closing threshold when the first target closing threshold is within a preset closing threshold range, and to update the right eye closing threshold to the second target closing threshold when the second target closing threshold is within the closing threshold range.

[0077] Furthermore, the target features include features of the presence of a face, yawning, left-leaning face, right-leaning face, head down, head up, presence of the left eye, left eye closed, left eye open, presence of the right eye, right eye closed, right eye open, and rubbing the eyes.

[0078] Furthermore, the probability value set acquisition module 403 includes: The information adding unit is used to add the face detection information, the eye detection information, the expression detection information and the face angle detection information of each of the images to be detected into preset short-term sliding windows and long-term sliding windows according to the frame number of the images to be detected. The sliding window statistics unit is used to judge and count whether each of the target features exists in the face detection information, the eye detection information, the expression detection information, and the face angle detection information based on the short-term sliding window, the long-term sliding window, and the preset threshold, respectively, to obtain the statistical results of the short-term sliding window and the statistical results of the long-term sliding window; The probability value acquisition unit is used to perform probability calculations based on the statistical results of the short-term sliding window and the statistical results of the long-term sliding window, respectively, to obtain the short-term probability value of each target feature in the short-term sliding window and the long-term probability value of each target feature in the long-term sliding window.

[0079] Furthermore, the KSS drowsiness level assessment device also includes: The probability value acquisition unit is used to acquire long-term probability values ​​of facial features based on the probability value set. The probability value comparison unit is used to reacquire the image to be detected when the long-term probability value of the facial feature is less than a preset facial presence probability threshold, and to detect the probability of the target feature. When the long-term probability value of the facial feature is greater than or equal to the facial presence threshold, based on the preset KSS drowsiness level assessment rule, the unit obtains the probability value of at least one of the target features from the probability value set, and compares the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed.

[0080] Furthermore, the KSS drowsiness level assessment module 404 includes: The probability value acquisition unit is used to acquire, based on the KSS drowsiness level assessment rule, the long-term probability value of the yawning feature, the long-term probability value of the left eye present and the left eye closed, the long-term probability value of the right eye present and the right eye closed, the long-term probability value of the left eye present and the left eye open, the long-term probability value of the right eye present and the right eye open, the long-term probability value of the face leaning to the left, the long-term probability value of the face leaning to the right, the long-term probability value of the head-down feature, the long-term probability value of the head-up feature, the long-term probability value of the eye-raising feature, the long-term probability value of the eye-rubbing feature, the long-term probability value of the left eye present, the short-term probability value of the left eye closed, the long-term probability value of the right eye present, and the short-term probability value of the right eye closed from the probability value set. The first comparison result obtaining unit is used to compare the long-term probability value of the yawning feature, the long-term probability value of the left eye being present and the left eye being closed, the long-term probability value of the right eye being present and the right eye being closed, the long-term probability value of the left eye being present and the left eye being open, and the long-term probability value of the right eye being present and the right eye being open with preset first yawning probability thresholds, preset first left eye closed probability thresholds, preset first right eye closed probability thresholds, preset first left eye open probability value, and preset first right eye open probability value, respectively, to obtain a first comparison result; The second comparison result obtaining unit is used to compare the long-term probability value of the face left-leaning feature, the long-term probability value of the face right-leaning feature, the long-term probability value of the head-down feature, the long-term probability value of the head-up feature, and the long-term probability value of the eye-rubbing feature with preset first left-leaning probability threshold, preset first right-leaning probability threshold, preset first head-down probability threshold, preset head-up probability threshold, and preset first eye-rubbing probability threshold, respectively, to obtain a second comparison result; The third comparison result obtaining unit is used to compare the long-term probability value of the left eye presence feature, the short-term probability value of the left eye closed feature, the long-term probability value of the right eye presence feature, and the short-term probability value of the right eye closed feature with preset left eye presence probability threshold, preset second left eye closed probability threshold, preset right eye presence probability threshold, and preset second right eye closed probability threshold, respectively, to obtain the third comparison result. The drowsiness level determination unit is used to determine the drowsiness level of the person to be evaluated based on the first comparison result, the second comparison result, and the third comparison result.

[0081] Furthermore, the KSS drowsiness level assessment module 404 also includes: The probability value comparison unit is used to compare the long-term probability values ​​of the left eye being present and the left eye being closed, the long-term probability values ​​of the right eye being present and the right eye being closed, and the long-term probability values ​​of the yawning feature with preset third left eye closed probability thresholds, preset third right eye closed probability thresholds, and preset third yawning probability thresholds, respectively, based on the KSS drowsiness level assessment rules. The probability value acquisition unit is used to acquire, from the probability value set, the short-term probability value of the left eye closing feature, the short-term probability value of the right eye closing feature, the short-term probability value of the yawning feature, and the short-term probability value of the eye rubbing feature if the long-term probability value of the left eye closing feature is greater than the third left eye closing probability threshold, or the long-term probability value of the right eye closing feature is greater than the third right eye closing probability threshold, or the long-term probability value of the yawning feature is greater than the third yawning probability threshold. The fourth comparison result obtaining unit is used to compare the short-term probability value of the left eye closing feature, the short-term probability value of the right eye closing feature, the short-term probability value of the yawning feature, and the short-term probability value of the eye rubbing feature with the preset fourth left eye closing probability threshold, the preset fourth right eye closing probability threshold, the preset fourth yawning probability threshold, and the preset second eye rubbing probability threshold, respectively, to obtain the fourth comparison result; The drowsiness level determination unit is used to determine the drowsiness level of the person to be evaluated based on the fourth comparison result.

[0082] Furthermore, the KSS drowsiness level assessment module 404 also includes: The probability value comparison unit is used to compare the long-term probability values ​​of the presence and closure of the left eye, the presence and closure of the right eye, and the yawning feature with preset fifth left eye closure probability thresholds, preset fifth right eye closure probability thresholds, and preset fifth yawning probability thresholds, respectively, based on the KSS drowsiness level assessment rules. The probability value acquisition unit is used to acquire, from the probability value set, the short-term probability value of the left eye closing feature, the short-term probability value of the right eye closing feature, the short-term probability value of the yawning feature, and the short-term probability value of the eye rubbing feature if the long-term probability value of the left eye closing feature is greater than the fifth left eye closing probability threshold, or the long-term probability value of the right eye closing feature is greater than the fifth right eye closing probability threshold, or the long-term probability value of the yawning feature is greater than the fifth yawning probability threshold. The fifth comparison result obtaining unit is used to compare the short-term probability value of the left eye closing feature, the short-term probability value of the right eye closing feature, the short-term probability value of the yawning feature, and the short-term probability value of the eye rubbing feature with the preset sixth left eye closing probability threshold, the preset sixth right eye closing probability threshold, the preset sixth yawning probability threshold, and the preset third eye rubbing probability threshold, respectively, to obtain the fifth comparison result; The drowsiness level determination unit is used to determine the drowsiness level of the person to be evaluated based on the fifth comparison result.

[0083] Furthermore, the KSS drowsiness level assessment module 404 also includes: The probability comparison unit is used to compare the long-term probability values ​​of the presence and closure of the left eye, the presence and closure of the right eye, and the yawning feature with preset seventh left eye closure probability thresholds, preset seventh right eye closure probability thresholds, and preset seventh yawning probability thresholds, respectively, based on the KSS drowsiness level assessment rules. The probability value acquisition unit is used to acquire, from the probability value set, the short-term probability value of the left eye closing feature, the short-term probability value of the right eye closing feature, the short-term probability value of the yawning feature, and the short-term probability value of the eye rubbing feature if the long-term probability value of the left eye closing feature is greater than the seventh left eye closing probability threshold, or the long-term probability value of the right eye closing feature is greater than the seventh right eye closing probability threshold, or the long-term probability value of the yawning feature is greater than the seventh yawning probability threshold. The sixth comparison result obtaining unit is used to compare the short-term probability value of the left eye closing feature, the short-term probability value of the right eye closing feature, the short-term probability value of the yawning feature, and the short-term probability value of the eye rubbing feature with the preset eighth left eye closing probability threshold, the preset eighth right eye closing probability threshold, the preset eighth yawning probability threshold, and the preset fourth eye rubbing probability threshold, respectively, to obtain the sixth comparison result; The drowsiness level determination unit is used to determine the drowsiness level of the person to be evaluated based on the sixth comparison result.

[0084] Furthermore, the KSS drowsiness level assessment module 404 also includes: The probability value acquisition unit is used to acquire short-term probability values ​​of features such as left-leaning face, right-leaning face, head-downward, and head-up based on the KSS drowsiness level assessment rules. The seventh comparison result obtaining unit is used to obtain the long-term probability values ​​of the following features: rubbing eyes, yawning, left eye presence and left eye closed, right eye presence and right eye closed, face left-leaning feature, face right-leaning feature, head down, head up, left eye closed, right eye closed, face left-leaning feature, face right-leaning feature, and head down. The value and the short-term probability value of the head-up feature are compared with the preset sixth eye-rubbing probability threshold, the preset ninth yawning probability threshold, the preset ninth left eye closing probability threshold, the preset ninth right eye closing probability threshold, the preset second left-leaning probability threshold, the preset second right-leaning probability threshold, the preset second head-down probability threshold, the preset second head-up probability threshold, the preset tenth left eye closing probability threshold, the preset tenth right eye closing probability threshold, the preset third left-leaning probability threshold, the preset third right-leaning probability threshold, the preset third head-down probability threshold, and the preset third head-up probability threshold to obtain the seventh comparison result; The drowsiness level determination unit is used to determine the drowsiness level of the person to be evaluated based on the seventh comparison result.

[0085] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the above-described apparatus and modules can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0086] The aforementioned device can be implemented as a computer program, which can be used in, for example... Figure 6 It runs on the computer device shown.

[0087] Please see Figure 6 , Figure 6 This is a schematic block diagram illustrating the structure of a computer device according to an embodiment of this application. The computer device may be a server.

[0088] See Figure 6 The computer device includes a processor, memory, and network interface connected via a system bus, wherein the memory may include non-volatile storage media and internal memory.

[0089] The non-volatile storage medium can store the operating system and computer program. The computer program includes program instructions that, when executed, cause the processor to perform any KSS drowsiness level assessment method.

[0090] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0091] Internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to perform any KSS drowsiness level assessment method.

[0092] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0093] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0094] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: Acquire continuous facial images of the person to be evaluated as the images to be detected; Each frame of the image to be detected is inspected to obtain face detection information, eye detection information, expression detection information, and face angle detection information; Based on a preset threshold, sliding window statistics are performed on the target features present in the face detection information, the eye detection information, the expression detection information, and the face angle detection information to obtain the probability value of each target feature and generate a set of probability values. Based on the preset KSS drowsiness level assessment rules, the probability value of at least one of the target features is obtained from the probability value set, and the probability value of each target feature is compared with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed.

[0095] In one embodiment, the preset threshold includes a left eye closing threshold and a right eye closing threshold. Before implementing the sliding window statistics of target features existing in the face detection information, the eye detection information, the expression detection information, and the face angle detection information based on the preset threshold to obtain the probability value of each target feature and generate a probability value set, the processor is also used to implement: Based on the human eye detection information, the aspect ratio data of the left eye and the aspect ratio data of the right eye of the person to be evaluated are obtained; Frequency distribution analysis was performed on the aspect ratio data of the left eye and the aspect ratio data of the right eye to obtain the first target aspect ratio interval for the left eye and the second target aspect ratio interval for the right eye. Based on the preset scaling factor and the first median of the first target aspect ratio interval and the second median of the second target aspect ratio interval, the first target eye-closing threshold and the second target eye-closing threshold are obtained respectively. When the first target eye-closing threshold is within the preset eye-closing threshold range, the left eye eye-closing threshold is updated to the first target eye-closing threshold, and when the second target eye-closing threshold is within the eye-closing threshold range, the right eye eye-closing threshold is updated to the second target eye-closing threshold.

[0096] In one embodiment, the target features include facial presence features, yawning features, facial left-leaning features, facial right-leaning features, head-down features, head-up features, left-eye presence features, left-eye closed features, left-eye open features, right-eye presence features, right-eye closed features, right-eye open features, and eye-rubbing features.

[0097] In one embodiment, when the processor performs sliding window statistics on target features present in the face detection information, the eye detection information, the expression detection information, and the face angle detection information based on a preset threshold to obtain the probability value of each target feature, it is configured to: According to the frame number of the image to be detected, the face detection information, eye detection information, expression detection information and face angle detection information of each image to be detected are added to a preset short-term sliding window and a long-term sliding window respectively; Based on the short-term sliding window, the long-term sliding window, and the preset threshold, the presence of each target feature in the face detection information, the eye detection information, the expression detection information, and the face angle detection information is judged and statistically analyzed to obtain the statistical results of the short-term sliding window and the statistical results of the long-term sliding window. Probability calculations are performed based on the statistical results of the short-term sliding window and the statistical results of the long-term sliding window to obtain the short-term probability value of each target feature in the short-term sliding window and the long-term probability value of each target feature in the long-term sliding window.

[0098] In one embodiment, before the processor implements a preset KSS drowsiness level assessment rule, obtains the probability value of at least one of the target features from the probability value set, and compares the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, it is further configured to: Based on the set of probability values, obtain long-term probability values ​​of facial features. When the long-term probability value of the facial feature is less than the preset facial presence probability threshold, the image to be detected is reacquired, and the probability of the target feature is detected. When the long-term probability value of the facial feature is greater than or equal to the facial presence threshold, based on the preset KSS drowsiness level assessment rule, at least one probability value of the target feature is obtained from the probability value set, and the probability value of each target feature is compared with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed.

[0099] In one embodiment, when the processor implements a preset KSS drowsiness level assessment rule, obtains the probability value of at least one target feature from the probability value set, and compares the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, it is configured to: Based on the KSS drowsiness level assessment rule, the following long-term probability values ​​are obtained from the probability value set: yawning, left eye presence and left eye closed, right eye presence and right eye closed, left eye presence and left eye open, right eye presence and right eye open, face left-leaning, face right-leaning, head down, head up, rubbing eyes, left eye presence, left eye closed, right eye closed, right eye presence, and right eye closed. The long-term probability values ​​of the yawning feature, the long-term probability values ​​of the left eye being present and the left eye being closed, the long-term probability values ​​of the right eye being present and the right eye being closed, the long-term probability values ​​of the left eye being present and the left eye being open, and the long-term probability values ​​of the right eye being present and the right eye being open are compared with preset first yawning probability thresholds, preset first left eye closed probability thresholds, preset first right eye closed probability thresholds, preset first left eye open probability values, and preset first right eye open probability values, respectively, to obtain a first comparison result; The long-term probability values ​​of the left-leaning face feature, the right-leaning face feature, the head-down feature, the head-up feature, and the eye-rubbing feature are compared with preset first left-leaning probability thresholds, preset first right-leaning probability thresholds, preset first head-down probability thresholds, preset head-up probability thresholds, and preset first eye-rubbing probability thresholds to obtain a second comparison result. The long-term probability value of the left eye presence feature, the short-term probability value of the left eye closed feature, the long-term probability value of the right eye presence feature, and the short-term probability value of the right eye closed feature are compared with preset left eye presence probability threshold, preset second left eye closed probability threshold, preset right eye presence probability threshold, and preset second right eye closed probability threshold, respectively, to obtain a third comparison result. Based on the first comparison result, the second comparison result, and the third comparison result, the drowsiness level of the person to be evaluated is determined.

[0100] In one embodiment, when the processor implements a preset KSS drowsiness level assessment rule, obtains the probability value of at least one of the target features from the probability value set, and compares the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, it is also configured to: Based on the KSS drowsiness level assessment rule, the long-term probability values ​​of the presence of the left eye and the left eye closed, the long-term probability values ​​of the presence of the right eye and the right eye closed, and the long-term probability values ​​of the yawning feature are compared with the preset third left eye closed probability threshold, the preset third right eye closed probability threshold, and the preset third yawning probability threshold, respectively. If the long-term probability value of the left eye being present and the left eye being closed is greater than the third probability threshold for the left eye being closed, or the long-term probability value of the right eye being present and the right eye being closed is greater than the third probability threshold for the right eye being closed, or the long-term probability value of the yawning feature is greater than the third probability threshold for the yawning feature, then the short-term probability values ​​of the left eye being closed, the right eye being closed, the yawning feature, and the eye rubbing feature are obtained from the probability value set. The short-term probability values ​​of the left eye closing feature, the right eye closing feature, the yawning feature, and the eye rubbing feature are compared with preset fourth left eye closing probability thresholds, preset fourth right eye closing probability thresholds, preset fourth yawning probability thresholds, and preset second eye rubbing probability thresholds to obtain a fourth comparison result. Based on the fourth comparison result, the level of drowsiness of the person to be evaluated is determined.

[0101] In one embodiment, when the processor implements a preset KSS drowsiness level assessment rule, obtains the probability value of at least one of the target features from the probability value set, and compares the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, it is also configured to: Based on the KSS drowsiness level assessment rule, the long-term probability values ​​of the presence of the left eye and the left eye closed, the long-term probability values ​​of the presence of the right eye and the right eye closed, and the long-term probability value of the yawning feature are compared with preset fifth left eye closed probability thresholds, preset fifth right eye closed probability thresholds, and preset fifth yawning probability thresholds, respectively: If the long-term probability value of the left eye being present and the left eye being closed is greater than the fifth probability threshold for the left eye being closed, or the long-term probability value of the right eye being present and the right eye being closed is greater than the fifth probability threshold for the right eye being closed, or the long-term probability value of the yawning feature is greater than the fifth probability threshold for the yawning feature, then the short-term probability values ​​of the left eye being closed, the right eye being closed, the yawning feature, and the eye rubbing feature are obtained from the probability value set. The short-term probability values ​​of the left eye closing feature, the right eye closing feature, the yawning feature, and the eye rubbing feature are compared with the preset sixth left eye closing probability threshold, the preset sixth right eye closing probability threshold, the preset sixth yawning probability threshold, and the preset third eye rubbing probability threshold, respectively, to obtain the fifth comparison result. Based on the fifth comparison result, the level of drowsiness of the person to be evaluated is determined.

[0102] In one embodiment, when the processor implements a preset KSS drowsiness level assessment rule, obtains the probability value of at least one of the target features from the probability value set, and compares the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, it is also configured to: Based on the KSS drowsiness level assessment rule, the long-term probability values ​​of the presence and closure of the left eye, the presence and closure of the right eye, and the yawning feature are compared with preset seventh left eye closure probability thresholds, preset seventh right eye closure probability thresholds, and preset seventh yawning probability thresholds, respectively: If the long-term probability value of the left eye being present and the left eye being closed is greater than the seventh probability threshold for the left eye being closed, or the long-term probability value of the right eye being present and the right eye being closed is greater than the seventh probability threshold for the right eye being closed, or the long-term probability value of the yawning feature is greater than the seventh probability threshold for the yawning feature, then the short-term probability values ​​of the left eye being closed, the right eye being closed, the yawning feature, and the eye rubbing feature are obtained from the probability value set. The short-term probability values ​​of the left eye closing feature, the right eye closing feature, the yawning feature, and the eye rubbing feature are compared with the preset eighth left eye closing probability threshold, the preset eighth right eye closing probability threshold, the preset eighth yawning probability threshold, and the preset fourth eye rubbing probability threshold, respectively, to obtain the sixth comparison result. Based on the sixth comparison result, the level of drowsiness of the person to be evaluated is determined.

[0103] In one embodiment, when the processor implements a preset KSS drowsiness level assessment rule, obtains the probability value of at least one of the target features from the probability value set, and compares the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, it is also configured to: Based on the KSS drowsiness level assessment rule, the short-term probability values ​​of the left-leaning face feature, the right-leaning face feature, the head-down feature, and the head-up feature are obtained. The long-term probability values ​​of the following features are included: the long-term probability value of rubbing eyes, the long-term probability value of yawning, the long-term probability value of the left eye being present and closed, the long-term probability value of the right eye being present and closed, the long-term probability value of the face being skewed to the left, the long-term probability value of the face being skewed to the right, the long-term probability value of looking down, the long-term probability value of looking up, the short-term probability value of the left eye being closed, the short-term probability value of the right eye being closed, the short-term probability value of the face being skewed to the left, the short-term probability value of the face being skewed to the right, the short-term probability value of looking down, and the short-term probability value of looking up. The short-term probability values ​​of the head features are compared with the preset sixth probability threshold for rubbing eyes, the preset ninth probability threshold for yawning, the preset ninth probability threshold for closing the left eye, the preset ninth probability threshold for closing the right eye, the preset second probability threshold for leaning to the left, the preset second probability threshold for leaning to the right, the preset second probability threshold for looking down, the preset second probability threshold for looking up, the preset tenth probability threshold for closing the left eye, the preset tenth probability threshold for closing the right eye, the preset third probability threshold for leaning to the left, the preset third probability threshold for leaning to the right, the preset third probability threshold for looking down, and the preset third probability threshold for looking up, respectively, to obtain the seventh comparison result; Based on the seventh comparison result, the level of drowsiness of the person to be evaluated is determined.

[0104] The embodiments of this application also provide a computer-readable storage medium storing a computer program, the computer program including program instructions, and the processor executing the program instructions to implement any of the KSS drowsiness level assessment methods provided in the embodiments of this application.

[0105] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0106] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A KSS drowsiness level assessment method, characterized in that, include: Acquire continuous facial images of the person to be evaluated as the images to be detected; Each frame of the image to be detected is detected to obtain face detection information, eye detection information, expression detection information, and face angle detection information for each frame of the image to be detected. Based on a preset threshold, sliding window statistics are performed on the target features present in the face detection information, the eye detection information, the expression detection information, and the face angle detection information to obtain the probability value of each target feature and generate a set of probability values. Based on the preset KSS drowsiness level assessment rules, the probability value of at least one of the target features is obtained from the probability value set, and the probability value of each target feature is compared with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed; the target features include facial presence features, yawning features, facial left-leaning features, facial right-leaning features, head-down features, head-up features, left-eye presence features, left-eye closed features, left-eye open features, right-eye presence features, right-eye closed features, right-eye open features, and eye-rubbing features; Based on a preset threshold, a sliding window statistical analysis is performed on the target features present in the face detection information, the eye detection information, the expression detection information, and the face angle detection information to obtain the probability value of each target feature, including: According to the frame number of the image to be detected, the face detection information, eye detection information, expression detection information and face angle detection information of each image to be detected are added to a preset short-term sliding window and a long-term sliding window respectively; Based on the short-term sliding window, the long-term sliding window, and the preset threshold, the presence of each target feature in the face detection information, the eye detection information, the expression detection information, and the face angle detection information is judged and statistically analyzed to obtain the statistical results of the short-term sliding window and the statistical results of the long-term sliding window. Probability calculations are performed based on the statistical results of the short-term sliding window and the statistical results of the long-term sliding window to obtain the short-term probability value of each target feature in the short-term sliding window and the long-term probability value of each target feature in the long-term sliding window.

2. The KSS drowsiness level assessment method according to claim 1, characterized in that, The preset thresholds include a left eye closing threshold and a right eye closing threshold. Before performing sliding window statistics on target features present in the face detection information, the eye detection information, the expression detection information, and the face angle detection information based on the preset thresholds to obtain the probability value of each target feature and generate a probability value set, the method further includes: Based on the human eye detection information, the aspect ratio data of the left eye and the aspect ratio data of the right eye of the person to be evaluated are obtained; Frequency distribution analysis was performed on the aspect ratio data of the left eye and the aspect ratio data of the right eye to obtain the first target aspect ratio interval for the left eye and the second target aspect ratio interval for the right eye. Based on the preset scaling factor and the first median of the first target aspect ratio interval and the second median of the second target aspect ratio interval, the first target eye-closing threshold and the second target eye-closing threshold are obtained respectively. When the first target eye-closing threshold is within the preset eye-closing threshold range, the left eye eye-closing threshold is updated to the first target eye-closing threshold, and when the second target eye-closing threshold is within the eye-closing threshold range, the right eye eye-closing threshold is updated to the second target eye-closing threshold.

3. The KSS drowsiness level assessment method according to claim 1, characterized in that, Before obtaining the drowsiness level of the person to be assessed by comparing the probability value of at least one of the target features with the probability value threshold corresponding to each target feature, based on the preset KSS drowsiness level assessment rule, the method further includes: Based on the set of probability values, obtain long-term probability values ​​of facial features. When the long-term probability value of the facial feature is less than the preset facial presence probability threshold, the image to be detected is reacquired, and the probability of the target feature is detected. When the long-term probability value of the facial feature is greater than or equal to the facial presence threshold, based on the preset KSS drowsiness level assessment rule, at least one probability value of the target feature is obtained from the probability value set, and the probability value of each target feature is compared with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed.

4. The KSS drowsiness level assessment method according to claim 1, characterized in that, The method based on the preset KSS drowsiness level assessment rule obtains the probability value of at least one of the target features from the probability value set, and compares the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, including: Based on the KSS drowsiness level assessment rule, the following long-term probability values ​​are obtained from the probability value set: yawning, left eye presence and left eye closed, right eye presence and right eye closed, left eye presence and left eye open, right eye presence and right eye open, face left-leaning, face right-leaning, head down, head up, rubbing eyes, left eye presence, left eye closed, right eye closed, right eye presence, and right eye closed. The long-term probability values ​​of the yawning feature, the long-term probability values ​​of the left eye being present and the left eye being closed, the long-term probability values ​​of the right eye being present and the right eye being closed, the long-term probability values ​​of the left eye being present and the left eye being open, and the long-term probability values ​​of the right eye being present and the right eye being open are compared with preset first yawning probability thresholds, preset first left eye closed probability thresholds, preset first right eye closed probability thresholds, preset first left eye open probability values, and preset first right eye open probability values, respectively, to obtain a first comparison result; The long-term probability values ​​of the left-leaning face feature, the right-leaning face feature, the head-down feature, the head-up feature, and the eye-rubbing feature are compared with preset first left-leaning probability thresholds, preset first right-leaning probability thresholds, preset first head-down probability thresholds, preset head-up probability thresholds, and preset first eye-rubbing probability thresholds to obtain a second comparison result. The long-term probability value of the left eye presence feature, the short-term probability value of the left eye closed feature, the long-term probability value of the right eye presence feature, and the short-term probability value of the right eye closed feature are compared with preset left eye presence probability threshold, preset second left eye closed probability threshold, preset right eye presence probability threshold, and preset second right eye closed probability threshold, respectively, to obtain a third comparison result. Based on the first comparison result, the second comparison result, and the third comparison result, the drowsiness level of the person to be evaluated is determined.

5. The KSS drowsiness level assessment method according to claim 4, characterized in that, The method based on the preset KSS drowsiness level assessment rule, which obtains the probability value of at least one of the target features from the probability value set, and compares the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, further includes: Based on the KSS drowsiness level assessment rule, the long-term probability values ​​of the presence of the left eye and the left eye closed, the long-term probability values ​​of the presence of the right eye and the right eye closed, and the long-term probability values ​​of the yawning feature are compared with the preset third left eye closed probability threshold, the preset third right eye closed probability threshold, and the preset third yawning probability threshold, respectively. If the long-term probability value of the left eye being present and the left eye being closed is greater than the third probability threshold for the left eye being closed, or the long-term probability value of the right eye being present and the right eye being closed is greater than the third probability threshold for the right eye being closed, or the long-term probability value of the yawning feature is greater than the third probability threshold for the yawning feature, then the short-term probability values ​​of the left eye being closed, the right eye being closed, the yawning feature, and the eye rubbing feature are obtained from the probability value set. The short-term probability values ​​of the left eye closing feature, the right eye closing feature, the yawning feature, and the eye rubbing feature are compared with preset fourth left eye closing probability thresholds, preset fourth right eye closing probability thresholds, preset fourth yawning probability thresholds, and preset second eye rubbing probability thresholds to obtain a fourth comparison result. Based on the fourth comparison result, the level of drowsiness of the person to be evaluated is determined.

6. The KSS drowsiness level assessment method according to claim 4, characterized in that, The method based on the preset KSS drowsiness level assessment rule, which obtains the probability value of at least one of the target features from the probability value set, and compares the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, further includes: Based on the KSS drowsiness level assessment rule, the long-term probability values ​​of the presence of the left eye and the left eye closed, the long-term probability values ​​of the presence of the right eye and the right eye closed, and the long-term probability value of the yawning feature are compared with preset fifth left eye closed probability thresholds, preset fifth right eye closed probability thresholds, and preset fifth yawning probability thresholds, respectively: If the long-term probability value of the left eye being present and the left eye being closed is greater than the fifth probability threshold for the left eye being closed, or the long-term probability value of the right eye being present and the right eye being closed is greater than the fifth probability threshold for the right eye being closed, or the long-term probability value of the yawning feature is greater than the fifth probability threshold for the yawning feature, then the short-term probability values ​​of the left eye being closed, the right eye being closed, the yawning feature, and the eye rubbing feature are obtained from the probability value set. The short-term probability values ​​of the left eye closing feature, the right eye closing feature, the yawning feature, and the eye rubbing feature are compared with the preset sixth left eye closing probability threshold, the preset sixth right eye closing probability threshold, the preset sixth yawning probability threshold, and the preset third eye rubbing probability threshold, respectively, to obtain the fifth comparison result. Based on the fifth comparison result, the level of drowsiness of the person to be evaluated is determined.

7. The KSS drowsiness level assessment method according to claim 4, characterized in that, The method based on the preset KSS drowsiness level assessment rule, which obtains the probability value of at least one of the target features from the probability value set, and compares the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, further includes: Based on the KSS drowsiness level assessment rule, the long-term probability values ​​of the presence and closure of the left eye, the presence and closure of the right eye, and the yawning feature are compared with preset seventh left eye closure probability thresholds, preset seventh right eye closure probability thresholds, and preset seventh yawning probability thresholds, respectively: If the long-term probability value of the left eye being present and the left eye being closed is greater than the seventh probability threshold for the left eye being closed, or the long-term probability value of the right eye being present and the right eye being closed is greater than the seventh probability threshold for the right eye being closed, or the long-term probability value of the yawning feature is greater than the seventh probability threshold for the yawning feature, then the short-term probability values ​​of the left eye being closed, the right eye being closed, the yawning feature, and the eye rubbing feature are obtained from the probability value set. The short-term probability values ​​of the left eye closing feature, the right eye closing feature, the yawning feature, and the eye rubbing feature are compared with the preset eighth left eye closing probability threshold, the preset eighth right eye closing probability threshold, the preset eighth yawning probability threshold, and the preset fourth eye rubbing probability threshold, respectively, to obtain the sixth comparison result. Based on the sixth comparison result, the level of drowsiness of the person to be evaluated is determined.

8. The KSS drowsiness level assessment method according to claim 4, characterized in that, The method based on the preset KSS drowsiness level assessment rule, which obtains the probability value of at least one of the target features from the probability value set, and compares the probability value of each target feature with the probability value threshold corresponding to each target feature to obtain the drowsiness level of the person to be assessed, further includes: Based on the KSS drowsiness level assessment rule, the short-term probability values ​​of the left-leaning face feature, the right-leaning face feature, the head-down feature, and the head-up feature are obtained. The long-term probability values ​​of the following features are included: the long-term probability value of rubbing eyes, the long-term probability value of yawning, the long-term probability value of the left eye being present and closed, the long-term probability value of the right eye being present and closed, the long-term probability value of the face being skewed to the left, the long-term probability value of the face being skewed to the right, the long-term probability value of looking down, the long-term probability value of looking up, the short-term probability value of the left eye being closed, the short-term probability value of the right eye being closed, the short-term probability value of the face being skewed to the left, the short-term probability value of the face being skewed to the right, the short-term probability value of looking down, and the short-term probability value of looking up. The short-term probability values ​​of the head features are compared with the preset sixth probability threshold for rubbing eyes, the preset ninth probability threshold for yawning, the preset ninth probability threshold for closing the left eye, the preset ninth probability threshold for closing the right eye, the preset second probability threshold for leaning to the left, the preset second probability threshold for leaning to the right, the preset second probability threshold for looking down, the preset second probability threshold for looking up, the preset tenth probability threshold for closing the left eye, the preset tenth probability threshold for closing the right eye, the preset third probability threshold for leaning to the left, the preset third probability threshold for leaning to the right, the preset third probability threshold for looking down, and the preset third probability threshold for looking up, respectively, to obtain the seventh comparison result; Based on the seventh comparison result, the level of drowsiness of the person to be evaluated is determined.

Images