Driving assistance light control system and method based on multispectral vision enhancement

Abstract

The invention is applicable to the technical field of safe driving assistance, and provides a driving assistance light control system and method based on multispectral vision enhancement, and the driving assistance light control system comprises a camera, a T-box, a vehicle-mounted intelligent computing platform, a microcontroller, a near-infrared LED control module, a near-infrared LED array and an RGB programmable LED lamp strip. On the basis of keeping the original lighting function of the vehicle-mounted atmosphere lamp, the system can effectively improve the fatigue degree during driving, reduce the driving fatigue value, delay the arrival of a fatigue surge moment, prolong the waking time and prolong the safe driving duration by adjusting lamplight on the basis of a physiological phototherapy mechanism.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of safe driving assistance, and particularly relates to a driving assistance light control system and method based on multispectral vision enhancement. BACKGROUND

[0002] Under the situation of continuous growth of the current automobile population and increasing demand for transportation volume, the occurrence of driving accidents increases. According to data statistics, accidents caused by driving fatigue account for a large proportion. At present, the field of safe driving assistance has put forward many auxiliary driving devices, which mainly rely on analyzing external information and then providing the driver with reminders of the vehicle or road conditions; or reminding the driver of the current mental state so that they can rest in time. For example, the augmented reality head-up display (AR HUD) enables the driver to more clearly understand the speed, oil quantity and other information at night, the vehicle-mounted radar can let the driver know the distance from the obstacle when the visibility is poor, and the fatigue detector can inform the driver of the degree of fatigue. But a very obvious defect is that these devices can only provide information, and cannot actually help the driver

[0003] Although the field of safe driving assistance has put forward many auxiliary driving devices, the core problem of these devices is that when fatigue occurs, they only serve as a prompt device to remind the driver, and lack feedback, and cannot actually improve the fatigue state of the driver. The research on stimulating the driver himself has different shortcomings, such as the low-current skin stimulation contacts the skin, which may affect the driving safety; the olfactory stimulation using odor has individual differences, and different people have different sensitivity and acceptance to odor; the visual stimulation improving the road sign is also affected by individual differences; the auditory stimulation playing vehicle-mounted music will aggravate fatigue for a long time, and is greatly affected by personal preference. SUMMARY

[0004] The purpose of the embodiment of the present application is to provide a driving assistance light control system and method based on multispectral vision enhancement, which aims to solve the problems proposed in the above background.

[0005] The embodiment of the present application is implemented in the following way, a driving assistance light control system based on multispectral vision enhancement, comprising a camera, a T-box, a vehicle-mounted intelligent computing platform, a microcontroller, a near-infrared LED control module, a near-infrared LED array and an RGB programmable LED light strip;

[0006] The camera is used to collect image information of the driver's face and send it to the T-box for storage;

[0007] The T-box is used to receive the image information collected by the camera and send it to the vehicle-mounted intelligent computing platform;

[0008] The vehicle-mounted intelligent computing platform is configured to receive image information sent by the T-box, pre-process the image, extract feature values, and calculate fatigue degree through a neural network deep learning algorithm, perform a trend test on the fatigue degree change every certain time period, and generate a control instruction according to the test result and send the control instruction to the microcontroller;

[0009] The microcontroller is configured to receive the control instruction sent by the vehicle-mounted intelligent computing platform, and send the control instruction to the RGB programmable LED light strip and the near-infrared LED control module;

[0010] The near-infrared LED control module is configured to receive the control instruction sent by the microcontroller, and send a control signal for controlling the on-off time and frequency of the near-infrared LED array to the near-infrared LED array according to the control instruction;

[0011] The near-infrared LED array is configured to receive the control signal sent by the near-infrared LED control module, and adjust the on-off state according to the control signal;

[0012] The RGB programmable LED light strip is configured to receive the control instruction sent by the microcontroller, and adjust the color and brightness according to the control instruction.

[0013] In a further technical solution, the camera is an infrared sensing camera.

[0014] In a further technical solution, the vehicle-mounted intelligent computing platform includes an image processing module, a perclos calculation module, a perclos trend test module, and a command issuing control module.

[0015] The image processing module is configured to improve the gray value of the image information sent by the T-box, identify the positions of the eyes and mouth in the image, and determine the contours.

[0016] The perclos calculation module is configured to qualitatively analyze the contours of the eyes and mouth, and calculate the perclos value according to a formula.

[0017] The perclos trend test module is configured to test whether the calculated perclos value has an upward trend.

[0018] The command issuing control module is configured to convert the test result into a control instruction, and send the control instruction to the microcontroller.

[0019] In a further technical solution, the RGB programmable LED light strip adopts single-wire serial communication, and the microcontroller's PWM (pulse width modulation) and DMA (direct memory access) functions are used to realize precise control of the lamp beads. All the lamp beads on the RGB programmable LED light strip are connected in series on a signal line, and each lamp bead is internally provided with a control circuit and an RGB chip.

[0020] Another purpose of the embodiment of the present application is to provide a driving assistance light control method based on multispectral vision enhancement, based on the driving assistance light control system described above, comprising the following steps:

[0021] Step 1: The camera collects image information of the driver's face in frames and sends it to the T-box for storage;

[0022] Step 2: The T-box receives the image information of the driver's face and sends it to the vehicle-mounted intelligent computing platform for processing;

[0023] Step 3: The vehicle-mounted intelligent computing platform receives the image information sent by the T-box, pre-processes the image, improves the image gray scale, and identifies the processed image to extract the feature values of the eyes and mouth, calculates the fatigue degree of multiple images through a neural network deep learning algorithm, uses a trend test algorithm to test the trend of fatigue degree every certain period of time, and generates a control instruction according to the test result and sends it to the microcontroller;

[0024] Step 4: The microcontroller receives the control instruction sent by the vehicle-mounted intelligent computing platform and sends the control instruction to the RGB programmable LED light strip and near-infrared LED control module;

[0025] Step 5: The near-infrared LED control module controls the on-off time and frequency of the near-infrared LED array according to the control instruction, and the RGB programmable LED light strip adjusts different colors and brightness according to the control instruction.

[0026] Further technical solutions, in the step 3, the pre-processing of image information is realized by python with opencv library, and the gray scale of the image is improved for subsequent recognition;

[0027] The recognition of feature values is realized by Yolov5 algorithm in deep learning, which can locate the positions of eyes and mouth in the image and identify the distraction behavior when driving.

[0028] Further technical solutions, in the step 3, the fatigue degree is calculated by using the improved perclos algorithm, and the specific calculation method is as follows:

[0029] If the degree of eye opening is less than the set threshold, two and eye-related counters are added by 1: COUNTER+=1, Roll eye+=1; if it is less than the threshold for 2 times in succession, it indicates that an eye blinking activity is performed, and the eye frame counter is reset: COUNTER=0; if the degree of mouth opening is less than the set threshold, two and mouth-related counters are added by 1: mCOUNTER+=1, Roll mouth+=1; if it is less than the threshold for 3 times in succession, it indicates that a yawn is performed, and the mouth frame counter is reset: mCOUNTER=0; the fatigue model takes 150 frames as a cycle, and Roll is added by 1 every frame, and when 150 frames are detected, the perclos value of the model is calculated;

[0030] perclos=(Rolleye / Roll)+(Rollmouth / Roll) x 0.2

[0031] The distraction detection takes 15 frames as a cycle, and if a distraction behavior is detected, the cycle time is extended; if no distraction behavior is detected for more than 15 frames, the calculation of the perclos value is restored.

[0032] Further technical solutions, in the step 3, the trend test of the fatigue degree adopts the Mann-Kendall test algorithm, and the calculated perclos data is sequentially taken according to the calculation time, that is, X={x1, x2,..., x n}, the difference function f(x i -x j )(n>=i>j>=1) of all is determined, and the calculation formula is as follows:

[0033]

[0034] The value of the calculation formula S is that the number of positive differences is subtracted from the number of negative differences:

[0035]

[0036] If S is a positive number, the observation value of the latter part tends to be larger than the observation value of the former part; if S is a negative number, the observation value of the latter part tends to be smaller than the observation value of the former part; the variance VAR(S) of S is calculated by using the following formula:

[0037]

[0038] Wherein, p is the number of repeated numbers, g is the number of unique numbers, t p is the number of repetitions of each repeated number; hypothesis testing is performed on the existence of the trend to determine whether the trend exists, and the MK test statistic is calculated:

[0039]

[0040] That is: Z MK Standard normal distribution. Given the significance test level alpha, using the standard normal distribution table to determine the standard normal variance If There is a significant upward or downward trend.

[0041] Further technical solutions, in the step 5, if the first time the perclos value trend test has a significant upward trend, control the RGB programmable LED lamp belt to open the blue light, and open the near-infrared light LED array; if the first time the perclos value trend test has no significant trend, control the RGB programmable LED lamp belt to open the red light, and open the near-infrared light LED array; if the subsequent detection obtains the upward trend of the perclos value, control the RGB programmable LED lamp belt to change to the blue light.

[0042] The driving auxiliary light control system and method based on multispectral vision enhancement provided by the embodiment of the application can effectively improve the fatigue degree during driving, reduce the driving fatigue value, delay the arrival of the fatigue surge moment, prolong the wake-up time, and prolong the safe driving time on the basis of retaining the original illumination function of the vehicle-mounted atmosphere lamp based on the physiological phototherapy mechanism through the adjustment of the light. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 The structural schematic diagram of the driving auxiliary light control system based on multispectral vision enhancement provided by the embodiment of the application is shown in the figure.

[0044] Figure 2 The flowchart of the driving auxiliary light control method based on multispectral vision enhancement provided by the embodiment of the application is shown in the figure.

[0045] In the figure: camera 1; T-box 2; vehicle-mounted intelligent computing platform 3; microcontroller 4; near-infrared LED control module 5; near-infrared LED array 6; RGB programmable LED lamp belt 7; image processing module 8; perclos calculation module 9; perclos trend test module 10; command issuing control module 11. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical scheme and advantages of the application more clear, the application will be further described in detail below in combination with the figures and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application and do not limit the application.

[0047] The specific implementation of the application is described in detail below in combination with specific embodiments.

[0048] As shown in Figure 1 A driving assistance light control system based on multispectral vision enhancement is provided for an embodiment of the present application, which comprises a camera 1, a T-box 2, an in-vehicle intelligent computing platform 3, a microcontroller 4, a near-infrared LED control module 5, a near-infrared LED array 6 and an RGB programmable LED light strip 7.

[0049] The camera 1 is used to collect image information of the driver's face and send it to the T-box 2 for storage.

[0050] The T-box 2 is used to receive the image information collected by the camera 1 and send it to the in-vehicle intelligent computing platform 3.

[0051] The in-vehicle intelligent computing platform 3 is used to receive the image information sent by the T-box 2, pre-process the image, extract feature values, and calculate fatigue degree through a neural network deep learning algorithm. Trend test is performed on the fatigue degree change of a certain time period every certain time period, and control instructions are generated according to the test results and sent to the microcontroller 4.

[0052] The microcontroller 4 is used to receive the control instructions sent by the in-vehicle intelligent computing platform 3 and send the control instructions to the RGB programmable LED light strip 7 and the near-infrared LED control module 5.

[0053] The near-infrared LED control module 5 is used to receive the control instructions sent by the microcontroller 4 and send control signals to the near-infrared LED array 6 to control the on-off time and frequency of the near-infrared LED array 6 according to the control instructions.

[0054] The near-infrared LED array 6 is used to receive the control signals sent by the near-infrared LED control module 5 and adjust the on-off state according to the control signals.

[0055] The RGB programmable LED light strip 7 is used to receive the control instructions sent by the microcontroller and adjust the color and brightness according to the control instructions.

[0056] As a preferred embodiment of the present application, the camera 1 is an infrared sensing camera.

[0057] As shown in Figure 1 As a preferred embodiment of the present application, the in-vehicle intelligent computing platform 3 comprises an image processing module 8, a perclos calculation module 9, a perclos trend test module 10 and a command issuing control module 11.

[0058] The image processing module 8 is used to improve the gray value of the image information sent by the T-box 2, and simultaneously identify the positions of eyes and mouth in the image and determine the outline.

[0059] The perclos calculation module 9 is used to qualitatively determine the eye mouth contour and calculate the perclos value according to the formula;

[0060] The perclos trend test module 10 is used to test whether the calculated perclos value has an upward trend;

[0061] The command issuing control module 11 is used to convert the test result into a control instruction and send it to the microcontroller 4.

[0062] In the embodiment of the present application, when the driver enters the vehicle, the system is started, the camera 1 is turned on to collect the image information of the driver's face in frames. The image information is sent to the T-box 2 for storage, and the T-box 2 uploads the image information to the vehicle intelligent computing platform 3 for algorithm processing. After image processing, perclos calculation and perclos trend test, the results are sent to the microcontroller 4 through the command issuing control module 11. The microcontroller 4 transmits the control instruction to the near-infrared LED control module 5 and the RGB programmable LED light strip 7 through the PWM current signal and DMA mode, adjusts the time frequency of the near-infrared LED array 6, and adjusts the color and brightness of the RGB programmable LED light strip 7.

[0063] After the system is started, the trend of the calculated perclos value is tested every 5 minutes, and at most the previous 20 minutes of the time point is calculated each time. If the result of the first trend test of the vehicle intelligent computing platform 3 is that the perclos value has a significant upward trend, the command issuing control module 11 issues a fatigue instruction to the microcontroller 4, the microcontroller 4 issues a control instruction to turn on the blue light to the RGB programmable LED light strip 7, and turns on the blue light. The near-infrared LED control module 5 issues an opening control instruction to turn on the near-infrared light LED array 6, so as to reduce fatigue in a short time.

[0064] If the result of the first trend test of the vehicle intelligent computing platform 3 is that the perclos value has no significant upward trend, the command issuing control module 11 issues a wake-up instruction to the microcontroller 4, the microcontroller 4 issues a control instruction to turn on the red light to the RGB programmable LED light strip 7, and turns on the red light. The near-infrared LED control module 5 issues an opening control instruction to turn on the near-infrared light LED array 6, so as to prolong the initial wake-up time. If the subsequent detection obtains a perclos value with an upward trend, the command issuing control module 11 issues a fatigue instruction to the microcontroller 4, the microcontroller 4 issues a control instruction to convert to blue light to the RGB programmable LED light strip 7, and controls the RGB programmable LED light strip 7 to change to blue light. The blue light can temporarily reduce the degree of fatigue to achieve a longer safe driving time.

[0065] If the vehicle-mounted intelligent computing platform 3 detects that the driver has a distraction behavior such as drinking water, smoking, playing mobile phones, etc., the cycle time is extended; if no distraction behavior is detected for more than 15 frames, the calculation of the perclos value is resumed.

[0066] If the vehicle-mounted intelligent computing platform 3 detects that the brightness of the collected image is low, it commands the control module 11 to issue a control instruction to increase the brightness to the microcontroller 4, which issues a control instruction to increase the brightness to the RGB programmable LED light strip 7, and controls the atmosphere lamp to increase the brightness.

[0067] If the vehicle-mounted intelligent computing platform 3 detects that the brightness of the collected image is high, it commands the control module 11 to issue a control instruction to reduce the brightness to the microcontroller 4, which issues a control instruction to reduce the brightness to the RGB programmable LED light strip 7, and controls the atmosphere lamp to reduce the brightness.

[0068] As a preferred embodiment of the present application, the RGB programmable LED light strip 7 is a smart LED light source integrating control circuit and light-emitting circuit, with control circuit and RGB chip built-in each lamp bead.

[0069] The RGB programmable LED light strip 7 adopts single-wire serial communication, and realizes precise control of the lamp beads through the PWM (Pulse Width Modulation) and DMA (Direct Memory Access) functions of the microcontroller 4. All lamp beads on the RGB programmable LED light strip 7 are connected in series on one signal line.

[0070] As shown in Figure 2 , a driving assistance light control method based on multi-spectral vision enhancement is provided for another embodiment of the present application, based on the above driving assistance light control system, comprising the following steps:

[0071] Step 1: The camera 1 collects image information of the driver's face in frames and sends it to the T-box 2 for storage;

[0072] Step 2: The T-box 2 receives the image information of the driver's face and sends it to the vehicle-mounted intelligent computing platform 3 for processing;

[0073] Step 3: The vehicle-mounted intelligent computing platform 3 receives the image information sent by the T-box 2, pre-processes the image, improves the image grayscale, and identifies the features of the eyes and mouth according to the processed image, calculates the fatigue degree of multiple frames of images through a neural network deep learning algorithm, uses a trend test algorithm to perform a trend test on the fatigue degree change every certain period of time, performs a trend test on the calculated perclos value every 5 minutes since the system is started, calculates the previous twenty minutes at each time point at most, and generates a control instruction according to the test result and sends it to the microcontroller 4;

[0074] Step 4: The microcontroller 4 receives the control instructions sent by the in-vehicle intelligent computing platform 3 and sends the control instructions to the RGB programmable LED light strip 7 and the near-infrared LED control module 5;

[0075] Step 5: The near-infrared LED control module 5 controls the on-off time and frequency of the near-infrared LED array 6 according to the control instructions, and the RGB programmable LED light strip 7 adjusts different colors and brightness according to the control instructions.

[0076] As a preferred embodiment of the present application, in the step 3, the pre-processing of image information is realized by python with opencv library, and the gray scale of the image is improved for subsequent recognition.

[0077] The feature value recognition is mainly realized by the Yolov5 algorithm in deep learning, which locates the positions of eyes and mouth in the image, and can also recognize distracted behaviors such as drinking, smoking and playing mobile phone while driving.

[0078] As a preferred embodiment of the present application, in the step 3, the fatigue degree is calculated by using the improved perclos (eye closure time ratio) algorithm, and the specific calculation method is as follows:

[0079] If the opening degree of the eyes is less than the set threshold value, two eye-related counters are added by 1: COUNTER+=1, Roll eye+=1; if it is less than the threshold value for two consecutive times, it means that an eye blinking activity is performed, and the eye frame counter COUNTER=0 is reset; if the opening degree of the mouth is less than the set threshold value, two mouth-related counters are added by 1: mCOUNTER+=1, Rollmouth+=1; if it is less than the threshold value for three consecutive times, it means that a yawn is made, and the mouth frame counter mCOUNTER=0 is reset; the fatigue model takes 150 frames as a cycle, and Roll is added by 1 every frame, and when 150 frames are detected, the perclos value of the model is calculated;

[0080] perclos=(Rolleye / Roll)+(Rollmouth / Roll)×0.2

[0081] The distraction detection takes 15 frames as a cycle, and if a distracted behavior (drinking, smoking, playing mobile phone) is detected, the cycle time is extended; if no distracted behavior is detected for more than 15 frames, the calculation of the perclos value is restored.

[0082] As a preferred embodiment of the present application, in the step 3, the trend test of the fatigue degree adopts the Mann-Kendall test algorithm, and the calculated perclos data is taken out in sequence according to the calculation time, that is, X={x1,x2,......,xn}, determine all the difference function f(x i - x j )(n>=i>j>=1) is calculated as follows:

[0083]

[0084] The value of S is calculated as the number of positive differences minus the number of negative differences:

[0085]

[0086] If S is a positive number, then the observations in the latter part tend to be larger than the observations in the former part; if S is a negative number, then the observations in the latter part tend to be smaller than the observations in the former part. Since the perclos data is greater than 10, the variance VAR(S) of S is calculated as follows:

[0087]

[0088] where p is the number of replicates, g is the number of unique groups (treatments), t p is the number of times each replicate is repeated; a hypothesis test is performed to determine if a trend exists, and the MK test statistic is calculated:

[0089]

[0090] That is, Z MK follows a standard normal distribution. Given a significance test level a, the standard normal variance If then a significant upward or downward trend exists.

[0091] As a preferred embodiment of the present application, in the step 5, if the first time the trend test of the perclos value has a significant upward trend, the RGB programmable LED light strip 7 is controlled to turn on blue light, and the near-infrared light LED array 6 is turned on; if the first time the trend test of the perclos value has no significant trend, the RGB programmable LED light strip 7 is controlled to turn on red light, and the near-infrared light LED array 6 is turned on; if the subsequent detection obtains an upward trend of the perclos value, the RGB programmable LED light strip 7 is controlled to change to blue light.

[0092] The above only describes preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, and improvement within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A driving assistance light control system based on multispectral vision enhancement, characterized by, The camera, the T-box, the vehicle-mounted intelligent computing platform, the microcontroller, the near-infrared LED control module, the near-infrared LED array and the RGB programmable LED lamp strip are included. The camera is used for collecting image information of the driver's face and sending the image information to the T-box for storage. The T-box is used for receiving the image information collected by the camera and sending the image information to the vehicle-mounted intelligent computing platform. The vehicle-mounted intelligent computing platform is used for receiving the image information sent by the T-box, pre-processing the image, extracting a characteristic value, calculating fatigue degree through a neural network deep learning algorithm, performing a trend test on the fatigue degree change every certain time period, and generating a control instruction according to the test result and sending the control instruction to the microcontroller. The microcontroller is used for receiving the control instruction sent by the vehicle-mounted intelligent computing platform and sending the control instruction to the RGB programmable LED lamp strip and the near-infrared LED control module. The near-infrared LED control module is used for receiving the control instruction sent by the microcontroller and sending a control signal for controlling the opening and closing time and frequency of the near-infrared LED array to the near-infrared LED array according to the control instruction. The near-infrared LED array is used for receiving the control signal sent by the near-infrared LED control module and adjusting the opening and closing state according to the control signal. The RGB programmable LED lamp strip is used for receiving the control instruction sent by the microcontroller and adjusting the color and brightness according to the control instruction.

2. The multispectral vision enhancement based driving assistance light control system of claim 1, wherein, The camera is an infrared sensing camera.

3. The multispectral vision enhancement based driving assistance light control system of claim 1, wherein, The vehicle-mounted intelligent computing platform includes an image processing module, a perclos calculation module, a perclos trend test module and a command issuing control module. The image processing module is used for improving the gray value of the image information sent by the T-box, identifying the positions of eyes and mouth in the image and determining the contours. The perclos calculation module is used for qualitatively analyzing the contours of eyes and mouth and calculating the perclos value according to a formula. The perclos trend test module is used for testing whether the calculated perclos value has an upward trend. The command issuing control module is used for converting the test result into a control instruction and sending the control instruction to the microcontroller.

4. The multispectral vision enhancement based driving assistance light control system of claim 1, wherein, The RGB programmable LED lamp strip adopts single-wire serial communication and realizes the control of the lamp beads through the PWM and DMA functions of the microcontroller. All the lamp beads on the RGB programmable LED lamp strip are connected in series on a signal line, and each lamp bead is internally provided with a control circuit and an RGB chip.

5. A method of controlling the light of a driving assistance based on multispectral vision enhancement, based on the system of controlling the light of a driving assistance based on multispectral vision enhancement according to any one of claims 1 to 4, characterized in that, The steps include the following steps: Step 1: The camera collects image information of the driver's face in frames and sends the image information to the T-box for storage. Step 2: The T-box receives the image information of the driver's face and sends the image information to the vehicle-mounted intelligent computing platform for processing. Step 3: The vehicle-mounted intelligent computing platform receives the image information sent by the T-box, pre-processes the image, improves the image gray scale, and identifies the image according to the processed image, extracts the feature values of the eyes and mouth, calculates the fatigue degree of multiple images through a neural network deep learning algorithm, performs a trend test on the fatigue degree change every certain period of time using a trend test algorithm, and generates a control instruction according to the test result and sends the control instruction to the microcontroller; Step 4: The microcontroller receives the control instruction sent by the vehicle-mounted intelligent computing platform, and sends the control instruction to the RGB programmable LED light strip and the near-infrared LED control module; Step 5: The near-infrared LED control module controls the on-off time and frequency of the near-infrared LED array according to the control instruction, and the RGB programmable LED light strip adjusts different colors and brightness according to the control instruction.

6. The multispectral vision enhancement-based driving assistance light control method according to claim 5, characterized by, In step 3, the pre-processing of image information is realized by python with opencv library, and the gray scale of the image is improved for subsequent recognition; The feature value recognition is realized by Yolov5 algorithm in deep learning, which locates the positions of eyes and mouth in the image and recognizes the distraction behavior during driving.

7. The multispectral vision enhancement-based driving assistance light control method according to claim 6, characterized by, In step 3, the fatigue degree is calculated by improved perclos algorithm, and the specific calculation method is as follows: If the degree of eye opening and closing is less than the set threshold, two eye-related counters are added by 1: COUNTER+=1, Roll eye+=1; If it is less than the threshold for two consecutive times, it means that an eye blinking activity is performed, and the eye frame counter COUNTER is reset to 0; if the degree of mouth opening and closing is less than the set threshold, two mouth-related counters are added by 1: mCOUNTER+=1, Rollmouth+=1; if it is less than the threshold for three consecutive times, it means that a yawn is performed, and the mouth frame counter mCOUNTER is reset to 0; the fatigue model takes 150 frames as a cycle, and Roll is added by 1 every frame; when 150 frames are detected, the perclos value of the model is calculated; perclos=(Rolleye / Roll)+(Rollmouth / Roll)×0.2 The distraction detection takes 15 frames as a cycle, and if a distraction behavior is detected, the cycle time is extended; If no distraction behavior is detected for more than 15 frames, the calculation of the perclos value is restored.

8. The multispectral vision enhancement-based driving assistance light control method according to claim 7, characterized by, In step 3, the fatigue trend is tested using the Mann-Kendall test algorithm. The calculated perclos data are extracted sequentially according to the calculation time, as X = {x1, x2, ..., x...} n }, determine all x i -x j The difference function f(x) i -x j (n>=i>j>=1), its calculation formula is as follows: The value of the calculation formula S is the difference between the number of positive differences and the number of negative differences: If S is a positive number, the observation value of the latter part will tend to be larger than the observation value of the former part; if S is a negative number, the observation value of the latter part will tend to be smaller than the observation value of the former part; the variance VAR(S) of S is calculated by the following formula: where p is the number of repetitions, g is the number of unique numbers, t p is the number of repetitions for each repetition; a hypothesis test is performed to determine the presence of a trend, and the MK test statistic is calculated: That is: Z MK Fits the standard normal distribution; given the significance test level a, use the standard normal distribution table to determine the standard normal variance If There is a significant upward or downward trend.

9. The multispectral vision enhancement-based driving assistance light control method according to claim 6, characterized by, In step 5, if the first time the perclos value trend test has a significant upward trend, the RGB programmable LED light strip is turned on with blue light, and the near-infrared LED array is turned on. If the first time the perclos value trend test is not significant trend, control the RGB programmable LED light belt to open red light, and open the near infrared light LED array; if the subsequent detection gets the rising trend of the perclos value, control the RGB programmable LED light belt to change to blue light.

Images