Cloud video processing method based on truck driving recorder

Abstract

The invention relates to the technical field of video processing, and discloses a cloud video processing method based on a truck driving recorder, which comprises the steps of video data uploading, video feature extraction, video feature analysis and warning signal display. Through complex calculation and analysis models, the fatigue degree of the driver can be accurately evaluated, and the accuracy and reliability of driving analysis are improved; multiple factors such as the continuous driving duration of the driver, the eye closing rate of different time periods, the eye closing judgment factor and the abnormal driving coefficient are comprehensively considered, the state of the driver is comprehensively evaluated, and the limitation of single-index evaluation is avoided; the warning signal is visually displayed to related personnel, so that the related personnel can quickly take measures, such as reminding a driver of having a rest or performing other interventions, and the safety of the driver and road traffic is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of video processing, and in particular to a cloud video processing method based on a truck driving recorder. Background Art

[0002] In the field of modern transportation, the safe operation of trucks is of vital importance. However, as truck transportation usually involves long distances and long driving hours, the problem of driver fatigue is becoming increasingly prominent. Fatigue driving not only reduces the driver's reaction ability and concentration, but also greatly increases the risk of traffic accidents.

[0003] Traditional truck driving monitoring methods often rely on manual observation or simple equipment recording, which makes it difficult to obtain real-time and accurate information on the driver's driving status. Although dashcams can record video data during driving, most of this data is only stored in local devices and cannot be effectively processed and analyzed centrally.

[0004] With the development of cloud computing technology, it is possible to upload the video data of truck driving recorders to the cloud for processing. However, there is currently a lack of a complete, efficient and accurate cloud video processing method to extract and analyze the driver's behavioral characteristics, so as to timely detect abnormal driving behaviors such as fatigue driving, and send warning signals to relevant personnel to ensure the safety and stability of truck driving.

[0005] Therefore, there is an urgent need for a cloud-based video processing method based on truck driving recorders that can accurately extract the driver's behavioral characteristics, realize real-time analysis and early warning of driving status, and reduce safety risks in truck transportation. Summary of the invention

[0006] The purpose of the present invention is to provide a cloud video processing method based on a truck driving recorder, which solves the technical problems raised in the background technology.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] The cloud video processing method based on the truck driving recorder includes the following steps:

[0009] Step 1: Upload video data

[0010] Upload the video data recorded by the truck's driving recorder to the cloud server in real time;

[0011] Step 2: Video feature extraction

[0012] The cloud server extracts the driver’s behavioral characteristics from the video data using facial recognition technology;

[0013] Step 3: Video feature analysis

[0014] According to the behavioral characteristics, the driver is analyzed and the corresponding warning signal is obtained according to the analysis results;

[0015] Step 4: Warning signal display

[0016] The corresponding warning signals are displayed to relevant personnel in an intuitive form.

[0017] As a further solution of the present invention: wherein the behavioral characteristics include the continuous driving time and eye closure rate of the same driver.

[0018] As a further solution of the present invention: the behavior feature extraction method is as follows:

[0019] Step 1: Video frame disassembly and timestamp extraction

[0020] Decompose the video data into several video frames and extract the timestamp of each frame;

[0021] Step 2: Facial image recognition

[0022] In each video frame, the facial image of the driver is identified using facial recognition technology;

[0023] Step 3: Mark the first facial image time frame and timestamp

[0024] In the video data, a time frame corresponding to the first appearance of the driver's facial image and its corresponding timestamp are obtained, and they are marked as the first video frame and the first timestamp respectively;

[0025] Step 4: Subsequent facial image detection and processing

[0026] Then, facial images in other corresponding video frames are obtained in sequence according to the time trend;

[0027] When the driver's facial image does not appear in a video frame, a timestamp corresponding to the video frame is obtained;

[0028] Then, n video frames subsequent to the video frame are acquired, and facial recognition technology is used to identify whether the n video frames contain the driver's facial image, where n is a preset value;

[0029] If none of the n video frames contains the driver's facial image, then the previous video frame corresponding to the video frame and its corresponding timestamp are obtained, and they are marked as the last video frame and the last timestamp respectively;

[0030] If at least one of the n video frames contains the driver's facial image, then the facial images in the subsequent corresponding video frames are obtained in sequence according to the time trend until there are n+1 video frames that do not contain the driver's facial image, and then a video frame before the n+1 video frame and its corresponding timestamp are obtained, and they are marked as the last video frame and the last timestamp respectively;

[0031] Step 5: Capture the video segment during driving time

[0032] Then, according to the first time frame and the last time frame, the corresponding video segment is intercepted from the video data and marked as the driving time period;

[0033] Step 6. Calculate the continuous driving time

[0034] At the same time, the timestamps of the start time node and the end time node of the driving time period in the video data are obtained, and the time difference between the two timestamps is calculated, and then recorded as the continuous driving time of the driver;

[0035] Step 7: Extract video frames during driving time

[0036] Extract all video frames during the driving time period;

[0037] Step 8: Eye area recognition and closed eye marking

[0038] In each video frame, the driver's eye area is identified using facial recognition technology;

[0039] Then, the pre-trained eye recognition model is used to identify whether the eye area in each video frame is closed, and a closed eye mark is added to the video frame where the eye area is closed;

[0040] Step 9: Driving time division and eye-closing frame number acquisition

[0041] Divide the driving time period into a number of predetermined standard time periods;

[0042] Wherein, each standard time period contains n video frames, and n is a fixed value;

[0043] Then, the number of video frames corresponding to the eye-closing mark is obtained from each standard time period. j , j = 1, 2, ... m, m represents the number of all standard time periods in the driving time period;

[0044] Step 10. Calculate the eye closure rate during the standard time period

[0045] By L j =u j / n, calculate the driver's eye closing rate L in each standard time period j ;

[0046] Step 11: Calculate the eye closure determination factor

[0047] Through L0 = L1 × β1 + L2 × β2 + ... L m ×β m , calculate the driver's eye closing determination factor L0 during the driving period;

[0048] Among them, β j is the preset weight factor;

[0049] Step 12: Calculate the abnormal driving coefficient

[0050] Obtain the driver's eye-closing determination factor L0 during the driving period, and mark the continuous driving time as K;

[0051] Then the abnormal driving coefficient Q is calculated by Q=L0×α1+K×α2;

[0052] Among them, α1 and α2 are both preset weight factors.

[0053] As a further embodiment of the present invention: wherein, β j The value of L j The value of is selected from the preset weight factor set [β a , β b , β c , β d ] is determined;

[0054] where β a <β b <β c <β d ;

[0055] When 0.75<L j ≤1, then β j =β a ;

[0056] When 0.5<L j ≤0.75, then β j =β b ;

[0057] When 0.25<L j ≤0.5, then β j =β c ;

[0058] When 0≤L j ≤0.25, then β j =β d .

[0059] As a further solution of the present invention: wherein, the eye recognition model recognizes whether the eye region in the video frame is closed in the following manner:

[0060] SH1. The eye recognition model uses a convolutional neural network for model training. During model training, a large amount of eye image data with open and closed eyes labeled is used to train the model so that it can learn to distinguish between open and closed eyes and create open eye templates and closed eye templates.

[0061] SH2. Calculate the similarity between the eye region in the video frame and a plurality of eye-opening templates and eye-closed templates respectively through an image similarity algorithm, and obtain the similarity between the eye region in the video frame and each eye-opening template and eye-closed template;

[0062] SH3. Then obtain the template corresponding to the similarity with the largest value. If the template is an open-eye template, it means that the driver's eye area in the video frame is in an open-eye state. If the template is a closed-eye template, it means that the driver's eye area in the video frame is in a closed-eye state, and a closed-eye mark is added to the video frame.

[0063] As a further solution of the present invention: the driving analysis method is as follows:

[0064] SA1. Compare the continuous driving time of the same driver with the preset driving time threshold:

[0065] If the continuous driving time is greater than or equal to the driving time threshold, it means that the driver is driving fatigued, and a warning signal 1 is generated; otherwise, no corresponding warning signal is generated;

[0066] SA2. In each standard time period corresponding to the continuous driving duration, respectively obtain the eye closure rate of the driver in each standard time period;

[0067] Then the driver's eye closure rate in each standard time period is compared with the preset eye closure rate threshold:

[0068] If the eye closing rate of any standard time period in all standard time periods is greater than or equal to the eye closing rate threshold, it means that the driver is driving fatigued and a second warning signal is generated. Otherwise, no corresponding warning signal is generated.

[0069] SA3. Obtain the driver's eye-closing determination factor during the driving time period, and then compare the driver's eye-closing determination factor during the continuous driving time with a preset determination factor threshold:

[0070] If the eye-closing determination factor is greater than or equal to the determination factor threshold, it indicates that the driver is driving fatigued, and a warning signal 3 is generated; otherwise, no corresponding warning signal is generated;

[0071] SA4. Obtain the abnormal driving coefficient of the driver during the driving time period, and then compare the abnormal driving coefficient of the driver with a preset abnormal driving threshold:

[0072] If the abnormal driving coefficient is greater than or equal to the abnormal driving threshold, it indicates that the driver is driving fatigued and a warning signal 4 is generated; otherwise, no corresponding warning signal is generated.

[0073] Beneficial effects of the present invention:

[0074] Real-time monitoring and early warning: By uploading the video data of the truck's driving recorder to the cloud in real time, the present invention can timely obtain the driver's driving behavior characteristics, realize real-time monitoring and early warning of abnormal driving behaviors such as fatigue driving, and effectively reduce the possibility of accidents.

[0075] Accurate analysis: Facial recognition technology is used to extract the driver's behavioral characteristics such as continuous driving time and eye closure rate. Through complex calculation and analysis models, the driver's fatigue level can be accurately assessed, improving the accuracy and reliability of driving analysis.

[0076] Comprehensive assessment: It comprehensively considers multiple factors such as the driver's continuous driving time, eye closure rate in different time periods, eye closure judgment factor and abnormal driving coefficient to conduct a comprehensive assessment of the driver's status, avoiding the limitations of single indicator evaluation.

[0077] Intelligent processing: Utilizing the powerful computing power and advanced algorithms of cloud servers, automatic processing and analysis of video data is achieved, reducing manual intervention and improving work efficiency and processing accuracy.

[0078] Effective warning: The warning signals are displayed to relevant personnel in an intuitive form, so that they can take prompt measures, such as reminding the driver to rest or taking other interventions, to ensure the safety of the driver and road traffic.

[0079] Accident prevention: It helps to detect driver fatigue in advance and intervene in time, thereby preventing traffic accidents and reducing casualties and property losses.

[0080] Improve management level: It provides effective management tools for transportation companies and regulatory authorities, helps to strengthen the management of truck drivers and improve the safety management level of the entire industry.

[0081] Data traceability: Video data and analysis results stored in the cloud facilitate subsequent query and tracing, providing strong evidence support for accident investigation and responsibility determination. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] The present invention will be further described below in conjunction with the accompanying drawings.

[0083] Figure 1 It is a system block diagram of the cloud video processing method based on the truck driving recorder of the present invention.

[0084] Figure 2 It is a flow chart of the cloud video processing method based on the truck driving recorder of the present invention. DETAILED DESCRIPTION

[0085] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0086] Embodiment 1

[0087] See also Figure 1 and Figure 2 As shown, the present invention is a cloud video processing method based on a truck driving recorder, comprising the following steps:

[0088] Step 1: Upload video data

[0089] Upload the video data recorded by the truck's driving recorder to the cloud server in real time;

[0090] Step 2: Video feature extraction

[0091] The cloud server extracts the driver’s behavioral characteristics from the video data using facial recognition technology;

[0092] Among them, the behavioral characteristics include the continuous driving time of the same driver;

[0093] The continuous driving time is extracted as follows:

[0094] StepA1, split the video data into several video frames, and extract the timestamp of each frame;

[0095] Step A2, in each video frame, using facial recognition technology to identify the driver's facial image;

[0096] Step A3, in the video data, obtain the time frame corresponding to the first appearance of the driver's facial image and its corresponding timestamp, and mark them as the first video frame and the first timestamp respectively;

[0097] Step A4, then sequentially obtain facial images in other corresponding video frames according to the time trend;

[0098] When the driver's facial image does not appear in a video frame, a timestamp corresponding to the video frame is obtained;

[0099] Then, n video frames subsequent to the video frame are acquired, and facial recognition technology is used to identify whether the n video frames contain the driver's facial image, where n is a preset value;

[0100] If none of the n video frames contains the driver's facial image, then the previous video frame corresponding to the video frame and its corresponding timestamp are obtained, and they are marked as the last video frame and the last timestamp respectively;

[0101] If at least one of the n video frames contains the driver's facial image, then the facial images in the subsequent corresponding video frames are obtained in sequence according to the time trend until there are n+1 video frames that do not contain the driver's facial image, and then a video frame before the n+1 video frame and its corresponding timestamp are obtained, and they are marked as the last video frame and the last timestamp respectively;

[0102] Step A5, then extract the corresponding video segment from the video data according to the first time frame and the last time frame, and mark it as the driving time period;

[0103] Step A6, simultaneously obtain the timestamps of the start time node and the end time node of the driving time period in the video data, calculate the time difference between the two timestamps, and then record it as the continuous driving time of the driver;

[0104] Step 3: Video feature analysis

[0105] According to the continuous driving time, the driver is analyzed and the corresponding warning signal is obtained according to the analysis results;

[0106] Driving analysis is as follows:

[0107] Compare the same driver's consecutive driving time with the preset driving time threshold:

[0108] If the continuous driving time is greater than or equal to the driving time threshold, it means that the driver is driving fatigued, and a warning signal 1 is generated; otherwise, no corresponding warning signal is generated;

[0109] Step 4: Warning signal display

[0110] The corresponding warning signals are displayed to relevant personnel in an intuitive form.

[0111] This embodiment can extract the continuous driving time of the truck driver in real time through the cloud server, and promptly detect whether the driver is at risk of fatigue driving, which helps to ensure road traffic safety.

[0112] Embodiment 2

[0113] As the second embodiment of the present invention, when the present application is implemented, compared with the first embodiment, the technical solution of the present embodiment is different from that of the first embodiment only in that:

[0114] In this embodiment, the behavior characteristics also include the eye closure rate of the same driver;

[0115] The eye closure rate is extracted as follows:

[0116] StepB1, extract all video frames during the driving time period;

[0117] Step B2, in each video frame, identify the driver's eye area using facial recognition technology;

[0118] Then, the pre-trained eye recognition model is used to identify whether the eye area in each video frame is closed, and a closed eye mark is added to the video frame where the eye area is closed;

[0119] The eye recognition model identifies whether the eye area in the video frame is closed in the following way:

[0120] SH1. The eye recognition model uses a convolutional neural network for model training. During model training, a large amount of eye image data labeled with open and closed eyes is used to train the model so that it can learn to distinguish between open and closed eyes and create open eye templates and closed eye templates.

[0121] SH2. Calculate the similarity between the eye region in the video frame and a plurality of eye-opening templates and eye-closed templates respectively through an image similarity algorithm, and obtain the similarity between the eye region in the video frame and each eye-opening template and eye-closed template;

[0122] SH3. Then, a template corresponding to the largest similarity value is obtained. If the template is an open-eye template, it indicates that the driver's eye area in the video frame is in an open-eye state. If the template is a closed-eye template, it indicates that the driver's eye area in the video frame is in a closed-eye state. Then, a closed-eye mark is added to the video frame.

[0123] Step B3, dividing the driving time period into several predetermined standard time periods;

[0124] Wherein, each standard time period contains n video frames, and n is a fixed value;

[0125] Then, the number of video frames corresponding to the eye-closing mark is obtained from each standard time period. j , j = 1, 2, ... m, m represents the number of all standard time periods in the driving time period;

[0126] Step B4, through L j =u j / n, calculate the driver's eye closing rate L in each standard time period j ;

[0127] Driving analysis is as follows:

[0128] In each standard time period corresponding to the continuous driving time, respectively obtain the eye closing rate of the driver in each standard time period;

[0129] Then the driver's eye closure rate in each standard time period is compared with the preset eye closure rate threshold:

[0130] If the eye closing rate of any standard time period in all standard time periods is greater than or equal to the eye closing rate threshold, it means that the driver is driving fatigued and a second warning signal is generated. Otherwise, no corresponding warning signal is generated.

[0131] In addition to the continuous driving time, this embodiment can also calculate the eye closure rate of the same driver, more comprehensively evaluate the driver's fatigue state, and improve the accuracy of fatigue driving judgment.

[0132] Embodiment 3

[0133] As the third embodiment of the present invention, when the present application is implemented, compared with the first and second embodiments, the technical solution of this embodiment is to combine the solutions of the first and second embodiments, and the difference between the technical solution of this embodiment and the first and second embodiments is that:

[0134] The behavioral characteristics also include the eye-closing determination factor of the same driver;

[0135] The eye closure determination factor is extracted as follows:

[0136] Through L0 = L1 × β1 + L2 × β2 + ... L m ×β m , calculate the driver's eye closing determination factor L0 during the driving period;

[0137] Among them, β j is the preset weight factor, and β j The value of L j The value of is selected from the preset weight factor set [β a , β b , β c , β d ] is determined;

[0138] where β a <β b <β c <βd ;

[0139] When 0.75<L j ≤1, then β j =β a ;

[0140] When 0.5<L j ≤0.75, then β j =β b ;

[0141] When 0.25<L j ≤0.5, then β j =β c ;

[0142] When 0≤L j ≤0.25, then β j =β d ;

[0143] Driving analysis is as follows:

[0144] The driver's eyes-closing determination factor during continuous driving time is compared with the preset determination factor threshold:

[0145] If the eye-closing determination factor is greater than or equal to the determination factor threshold, it indicates that the driver is driving fatigued and a warning signal three is generated; otherwise, no corresponding warning signal is generated.

[0146] This embodiment introduces an eye-closing determination factor, which comprehensively considers the eye-closing conditions in multiple standard time periods, making the judgment of fatigue driving more accurate and scientific.

[0147] Embodiment 4

[0148] As the fourth embodiment of the present invention, when the present application is implemented, compared with the first embodiment, the second embodiment and the third embodiment, the difference between the present embodiment and the first embodiment, the second embodiment and the third embodiment is only that:

[0149] In this embodiment, the behavior characteristics also include the abnormal driving coefficient of the same driver;

[0150] The abnormal driving coefficient is extracted as follows:

[0151] Obtain the driver's eye-closing determination factor L0 during the driving period, and mark the continuous driving time as K;

[0152] The abnormal driving coefficient Q is calculated by Q = L0 × α1 + K × α2;

[0153] Driving analysis is as follows:

[0154] Compare the driver's abnormal driving coefficient with the preset abnormal driving threshold:

[0155] If the abnormal driving coefficient is greater than or equal to the abnormal driving threshold, it indicates that the driver is driving fatigued and a warning signal 4 is generated; otherwise, no corresponding warning signal is generated.

[0156] This embodiment calculates the abnormal driving coefficient, combines the eye-closing determination factor and the continuous driving time, further improves the evaluation index of fatigue driving, and can more effectively identify the risk of fatigue driving.

[0157] Embodiment 5

[0158] As the fifth embodiment of the present invention, when the present application is specifically implemented, compared with the first, second, third and fourth embodiments, the technical solution of this embodiment is to combine and implement the solutions of the above-mentioned first, second, third and fourth embodiments.

[0159] This embodiment combines all the above-mentioned features and evaluation methods to monitor and judge the driver's fatigue driving status from all aspects and angles, thereby ensuring the driving safety of trucks to the greatest extent and reducing the possibility of traffic accidents.

[0160] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0161] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A cloud video processing method based on a truck driving recorder, characterized in that: The following steps are involved: Step 1: Upload video data Upload the video data recorded by the truck's driving recorder to the cloud server in real time; Step 2: Video feature extraction The cloud server extracts the driver’s behavioral characteristics from the video data using facial recognition technology; Step 3: Video feature analysis According to the behavioral characteristics, the driver is analyzed and the corresponding warning signal is obtained according to the analysis results; Step 4: Warning signal display The corresponding warning signals are displayed to relevant personnel in an intuitive form.

2. The cloud video processing method based on the truck driving recorder according to claim 1 is characterized in that: in, The behavioral characteristics include the continuous driving time and eyes-closing rate, eyes-closing determination factor and abnormal driving coefficient of the same driver.

3. The cloud video processing method based on the truck driving recorder according to claim 2 is characterized in that: The behavior feature extraction method is as follows: Step 1: Decompose the video data into several video frames and extract the timestamp of each frame; Step 2: In each video frame, use facial recognition technology to identify the driver's facial image; Step 3, in the video data, obtain the time frame corresponding to the first appearance of the driver's facial image and its corresponding timestamp, and mark them as the first video frame and the first timestamp respectively; Step 4, then obtain the facial images in other corresponding video frames in sequence according to the time trend; When the driver's facial image does not appear in a video frame, a timestamp corresponding to the video frame is obtained; Then, n video frames subsequent to the video frame are acquired, and facial recognition technology is used to identify whether the n video frames contain the driver's facial image, where n is a preset value; If none of the n video frames contains the driver's facial image, then the previous video frame corresponding to the video frame and its corresponding timestamp are obtained, and they are marked as the last video frame and the last timestamp respectively; If at least one of the n video frames contains the driver's facial image, then the facial images in the subsequent corresponding video frames are obtained in sequence according to the time trend until there are n+1 video frames that do not contain the driver's facial image, and then a video frame before the n+1 video frame and its corresponding timestamp are obtained, and they are marked as the last video frame and the last timestamp respectively; Step 5, then according to the first time frame and the last time frame, the corresponding video segment is intercepted from the video data and marked as the driving time period; Step 6, simultaneously obtain the timestamps of the start time node and the end time node of the driving time period in the video data, and calculate the time difference between the two timestamps, which is then recorded as the continuous driving time of the driver; Step 7, extract all video frames during the driving period; Step 8. In each video frame, use facial recognition technology to identify the driver's eye area; Then, the pre-trained eye recognition model is used to identify whether the eye area in each video frame is closed, and a closed eye mark is added to the video frame where the eye area is closed; Step 9, dividing the driving time period into several predetermined standard time periods; Wherein, each standard time period contains n video frames, and n is a fixed value; Then, the number of video frames corresponding to the eye-closing mark is obtained from each standard time period. j , j = 1, 2, ... m, m represents the number of all standard time periods in the driving time period; Step 10, through L j =u j / n, calculate the driver's eye closing rate L in each standard time period j ; Step 11, through L0 = L1 × β1 + L2 × β2 + ... L m ×β m , calculate the driver's eye closing determination factor L0 during the driving period; Among them, β j is the preset weight factor; Step 12, obtaining the driver's eye-closing determination factor L0 during the driving period, and marking the continuous driving time as K; Then the abnormal driving coefficient Q is calculated by Q=L0×α1+K×α2; Among them, α1 and α2 are both preset weight factors.

4. The cloud video processing method based on the truck driving recorder according to claim 3 is characterized in that: in, β j The value of L j The value of is selected from the preset weight factor set [β a , β b , β c , β d ] is determined; among themb a <b b <b c <b d ; When 0.75<L j ≤1, then β j =β a ; When 0.5<L j ≤0.75, then β j =β b ; When 0.25<L j ≤0.5, then β j =β c ; When 0 ≤ L j ≤ 0.25, then β j = β d .

5. The cloud video processing method based on the truck driving recorder according to claim 3 is characterized in that: in, The eye recognition model identifies whether the eye area in a video frame is closed as follows: SH1. The eye recognition model uses a convolutional neural network for model training. During model training, a large amount of eye image data with open and closed eyes labeled is used to train the model so that it can learn to distinguish between open and closed eyes and create open eye templates and closed eye templates. SH2. Calculate the similarity between the eye region in the video frame and a plurality of eye-opening templates and eye-closed templates respectively through an image similarity algorithm, and obtain the similarity between the eye region in the video frame and each eye-opening template and eye-closed template; SH3. Then obtain the template corresponding to the similarity with the largest value. If the template is an open-eye template, it means that the driver's eye area in the video frame is in an open-eye state. If the template is a closed-eye template, it means that the driver's eye area in the video frame is in a closed-eye state, and a closed-eye mark is added to the video frame.

6. The cloud video processing method based on the truck driving recorder according to claim 3 is characterized in that: Driving analysis is as follows: Compare the same driver's consecutive driving time with the preset driving time threshold: If the continuous driving time is greater than or equal to the driving time threshold, it means that the driver is driving fatigued and a warning signal is generated; otherwise, no corresponding warning signal is generated.

7. The cloud video processing method based on the truck driving recorder according to claim 3 is characterized in that: Driving analysis is as follows: In each standard time period corresponding to the continuous driving time, respectively obtain the eye closing rate of the driver in each standard time period; Then the driver's eye closure rate in each standard time period is compared with the preset eye closure rate threshold: If the eye closing rate of any standard time period in all standard time periods is greater than or equal to the eye closing rate threshold, it means that the driver is driving fatigued and a second warning signal is generated. Otherwise, no corresponding warning signal is generated.

8. The cloud video processing method based on a truck driving recorder according to claim 3 is characterized in that: Driving analysis is as follows: The eye-closing determination factor of the driver during the driving time period is obtained, and then the eye-closing determination factor of the driver during the continuous driving time is compared with the preset determination factor threshold: If the eye-closing determination factor is greater than or equal to the determination factor threshold, it indicates that the driver is driving fatigued and a warning signal three is generated; otherwise, no corresponding warning signal is generated.

9. The cloud video processing method based on the truck driving recorder according to claim 3 is characterized in that: Driving analysis is as follows: The abnormal driving coefficient of the driver during the driving time period is obtained, and then the abnormal driving coefficient of the driver is compared with the preset abnormal driving threshold: If the abnormal driving coefficient is greater than or equal to the abnormal driving threshold, it indicates that the driver is driving fatigued and a warning signal 4 is generated; otherwise, no corresponding warning signal is generated.

Images