A method for suppressing reflection in infrared imaging glasses based on image processing

Abstract

The present invention discloses a method for suppressing reflections of infrared imaging glasses based on image processing, which relates to the technical field of automobile fatigue monitoring, and includes the following steps: using an on-board infrared camera to capture the driver's facial image under fill light conditions, ensuring that the camera setting is suitable for capturing the reflection effect of infrared light on the glasses; preprocessing the captured image to more clearly identify the light spot; and applying image processing technology to identify bright areas in the image. The present invention effectively identifies and reduces the light spots that affect the driver's facial features by identifying and classifying the features in the light spots, optimizes data processing, improves the accuracy of fatigue detection, and reduces safety risks. When a high-impact light spot is detected, the fill light mode is adjusted to a pulse mode to reduce the light spot problem, avoid the situation where the reflection of two glasses blocks the eyes, and synchronize the fill light and camera exposure time to reduce reflections, improve image quality and fatigue detection accuracy, and enhance driving safety.

Description

Technical Field

[0001] The invention relates to the technical field of automobile fatigue monitoring, and in particular to a method for suppressing reflections of infrared imaging glasses based on image processing. Background Art

[0002] Reflection suppression of infrared imaging glasses based on image processing mainly refers to the use of specific image processing methods to reduce or eliminate the reflection problem caused by the driver wearing glasses when using infrared imaging technology for driver fatigue monitoring. In the automotive field, various advanced driver assistance systems (ADAS) have been developed to improve driving safety, among which the driver fatigue detection system is a key component. This system usually uses an infrared camera to capture the driver's facial features and eye movements. The infrared camera can effectively capture images under various lighting conditions, including at night or in backlight conditions, thereby ensuring the stable operation of the system. The system evaluates the driver's attention and fatigue by analyzing parameters such as the driver's blinking frequency, eyelid closure time, pupil position, and head posture.

[0003] In addition, these systems also use advanced image processing technology and machine learning algorithms to extract facial features and behavioral patterns in order to more accurately judge the driver's status. For example, by continuously monitoring the driver's eye movement patterns and frequent yawning, the system can determine whether the driver is in a state of fatigue. Once the system detects signs of driver fatigue, it will alert the driver through visual and sound alarms, and even in some advanced systems, it can temporarily take over vehicle control to prevent the driver from getting into an accident due to excessive fatigue. The integration of these technologies not only enhances driving safety, but also greatly improves the level of vehicle intelligence, making long-distance driving safer and easier.

[0004] However, when the driver wears glasses, the surface of the glasses will reflect infrared light sources (such as LED fill lights) and produce reflected light spots on the mirror surface, which may cause the pupil and eye expression information to be obscured or distorted due to the reflected light spots, thereby affecting the accuracy of fatigue detection. Due to inaccurate fatigue detection, the system may not be able to promptly warn the driver of their fatigue state, thereby missing the critical moment to prevent potential danger. This monitoring error may cause the driver to continue driving without realizing his or her own fatigue level, increasing the incidence of traffic accidents, especially when driving for long periods of time or at night.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a method for suppressing reflections of infrared imaging glasses based on image processing. By extracting features from the identified light spots and classifying them, the light spots that seriously affect the recognition of the driver's facial features can be effectively determined and reduced. This classification optimizes the data processing process, improves the accuracy of fatigue detection, and reduces the safety risks caused by misjudgment. In addition, when high-impact light spots are detected, by adjusting the fill light mode from conventional fill light to pulse fill light, the light spot problem is significantly reduced, avoiding the situation where the eyes are blocked by the reflections of two glasses, ensuring strict synchronization of the fill light and the camera exposure time, reducing light reflection and light spot generation, thereby improving image quality and enhancing the accuracy of fatigue detection, further enhancing driving safety, and solving the problems in the above-mentioned background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a method for suppressing reflection of infrared imaging glasses based on image processing, comprising the following steps: Use the vehicle's infrared camera to capture the driver's facial image under fill-light conditions, ensuring that the camera settings are suitable for capturing the reflection of infrared light on the glasses; Pre-process the captured images to more clearly identify the light spots; Apply image processing techniques to identify bright areas in the image, i.e., light spots, and use morphological operations to precisely define the boundaries of the light spots; Extract features from the identified light spots, and divide the light spots into high-impact light spots and low-impact light spots through comprehensive analysis based on the extracted features; When a low-impact light spot exists on the mirror, the driver's facial image continues to be captured under fill-light conditions. When a high-impact light spot is detected on the mirror, the fill-light process is optimized to reduce the impact of the light spot and ensure efficient monitoring of driver fatigue driving.

[0008] Preferably, a region growing algorithm is applied to identify bright areas in the image, i.e., light spots. The specific steps are as follows: Select a seed point, which is the starting point for the region growing algorithm to start growing; Define the growth criterion, which is the rule that determines whether a pixel is added to the seed region; Start region growing, starting from the seed point, and add adjacent pixels that meet the growth criteria to the current region; Repeat the above growing process until no pixels are added to the region; The final region growing result is output, and each region is represented by a different label.

[0009] Preferably, a morphological dilation operation is used to accurately define the boundary of the light spot, and the specific steps are as follows: Convert the original color image to grayscale, simplifying the information and focusing on brightness rather than color; Select or define a structuring element that determines the refinement and direction of the dilation operation; Apply dilation operation to the binarized image; The dilated image is processed using edge detection techniques to precisely define and enhance the boundaries of the light spot.

[0010] Preferably, the spot size dynamic change information, the spot edge sharpness information, and the spot intensity gradient information are extracted from the identified light spot. After abnormal analysis and processing of the spot size dynamic change information, the spot edge sharpness information, and the spot intensity gradient information, a spot size dynamic change index, a spot edge sharpness index, and a spot intensity gradient index are generated. A spot classification and recognition model is established through the spot size dynamic change index, the spot edge sharpness index, and the spot intensity gradient index, and a spot influence coefficient is generated. The spot is divided into a high-impact spot and a low-impact spot through the spot influence coefficient.

[0011] Preferably, after obtaining the spot influence coefficient generated by the abnormal analysis processing of the mirror reflection spot, the spot influence coefficient is compared and analyzed with a pre-set spot influence coefficient reference threshold; if the spot influence coefficient is greater than or equal to the spot influence coefficient reference threshold, the identified mirror reflection spot is classified as a high-impact spot; if the spot influence coefficient is less than the spot influence coefficient reference threshold, the identified mirror reflection spot is classified as a low-impact spot.

[0012] Preferably, when a high-impact light spot is detected on the mirror surface, the steps of optimizing the fill light process are as follows: Change the conventional continuous fill light to pulse fill light; When the fill light is on, a lighting signal is generated and transmitted to the camera control system, which controls the camera exposure so that the lighting time of the fill light is completely consistent with the exposure time of the camera; When the fill light is on, the current passing through the fill light is increased in a short period of time to increase the luminous power of the fill light. When the camera exposure is completed, the fill light is turned off.

[0013] Preferably, the logic for obtaining the dynamic change index of the spot size is as follows: Image data about the spot area is collected at each time frame t, and the image data is used is represented by , where t represents the time frame, Represents the pixel position in the image; The spot area is segmented, and the specific segmentation formula is: , where Represents the set of spot pixels in time frame t, is the grayscale threshold, used to distinguish between light spot and non-light spot areas; Calculate the spot area, the calculation expression is: , where is the spot area of ​​the time frame t, that is The number of pixels in To express; Calculate the spot area change rate between two consecutive frames of images. The calculation expression is: , where is the rate of change of the spot area between two consecutive frames of images; Calculate the dynamic change index of the spot size, and the calculation expression is: ,in, is the dynamic change index of the light spot, n is the number of time frames considered, k Is a loop variable used to traverse the past n The spot area changes of the frame, For the past n The absolute value of the spot area change for each frame within a frame.

[0014] Preferably, the logic of obtaining the light spot edge sharpness index is as follows: Get image data at pixel points The brightness value of The edge strength of the image data is calculated, and the calculation expression is: , and The images are at points Chu Yan x Direction and y The gradient of the direction, The image is The gradient amplitude at , indicating the edge strength; Calculate the edge sharpness of the image data. The calculation expression is: , where F is the gradient threshold, Indicates the edge sharpness value after gradient threshold processing; Calculate the edge sharpness index of the light spot. The calculation expression is: , where Indicates the edge sharpness index of the light spot. S Indicates the light spot area, is the sum of the sharpness values ​​of all pixels in the spot area, is the number of pixels in the spot area.

[0015] Preferably, the logic of obtaining the spot intensity gradient index is as follows: Get each pixel image at point Chu Yanx Direction and y Directional gradient and , while acquiring images at The gradient magnitude at , calculated at the pixel point The gradient direction of is calculated as: , Represents pixel The gradient direction of Calculate the average gradient amplitude in the spot area. The calculation expression is: , B is the pixel set of the spot area, N Light spot area S The total number of pixels in Indicates the spot area S The average gradient magnitude within ; Calculate the spot intensity gradient index, the calculation expression is: , where It indicates the intensity gradient index of the spot. The larger the value of the intensity gradient index, the closer the gradient amplitude of each point in the spot area is to the average gradient. The greater the difference between them.

[0016] In the above technical solution, the technical effects and advantages provided by the present invention are: The present invention extracts features from the identified light spots, and divides the light spots into high-impact light spots and low-impact light spots based on a comprehensive analysis of the extracted features. By classifying and identifying the light spots, the system can effectively determine the light spots that seriously affect the recognition of the driver's facial features, so as to take effective measures in time to reduce these interferences. Such classification not only optimizes the data processing flow, but also ensures the accuracy of fatigue detection results, and reduces safety risks caused by misjudgment.

[0017] When the present invention detects the presence of high-impact light spots on the mirror surface, the conventional fill light mode is adjusted to an optimized pulse fill light mode, which can significantly reduce the light spot problem caused by continuous fill light and avoid the situation where the eyes are blocked by reflections from two glasses. This optimization ensures that the lighting of the fill light is strictly synchronized with the exposure time of the camera, thereby reducing unnecessary light reflections and light spots. This not only improves the image quality, but also makes fatigue detection more accurate, further improving driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0019] Figure 1 The present invention is a method flow chart of a method for suppressing reflections in infrared imaging glasses based on image processing. DETAILED DESCRIPTION

[0020] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0021] The present invention provides Figure 1 The method for suppressing reflection of infrared imaging glasses based on image processing shown in the figure comprises the following steps: Use the vehicle's infrared camera to capture the driver's facial image under fill-light conditions, ensuring that the camera settings are suitable for capturing the reflection of infrared light on the glasses; The infrared camera installed in the car records the driver's face with the assistance of additional light sources. The fill light here usually refers to infrared fill light, which helps the camera capture images clearly in low-light or no-light environments. Infrared fill light can improve the image quality of the driver's face, especially at night or other low-light conditions, allowing the system to accurately analyze the driver's facial expressions and fatigue status.

[0022] Make sure the camera settings are suitable for capturing the reflection of infrared light on the glasses, as follows: Adjust the camera focus and angle, make sure the camera is aimed at the driver's glasses, and adjust the camera angle and focus so that it can capture the reflected light spots on the glasses.

[0023] Adjust the exposure and gain settings to optimize the camera's exposure and gain so that the reflection on the glasses is not blown out but still visible enough for analysis.

[0024] Proper camera settings can help the system accurately identify and analyze light spots caused by reflections from glasses, which is crucial to reducing false alarms and improving the accuracy of fatigue detection.

[0025] Pre-process the captured images to more clearly identify the light spots; Preprocessing the captured image is an important step in image analysis, which aims to improve the image quality and make the subsequent feature extraction and analysis process more accurate and efficient. Preprocessing usually includes the following main steps (including but not limited to, which can be selected according to actual conditions), each of which has its specific role: Grayscale: Convert color images to grayscale images to simplify the complexity of data processing. This is because grayscale images only contain brightness information, not color information, which reduces the amount of calculation while maintaining sensitivity to lighting changes. This is done by converting RGB values ​​to grayscale values. A common method is weighted averaging, in which different weights are given according to the sensitivity of the human eye to color.

[0026] Denoising: Remove random noise from the image to improve the visual quality of the image. Noise may come from the camera sensor, ambient light fluctuations, etc. Denoising helps to highlight key features in the image, such as light spots. Use filters such as median filtering and Gaussian filtering. Median filtering is suitable for removing salt and pepper noise, while Gaussian filtering is suitable for reducing high-frequency noise in the image.

[0027] Contrast enhancement: Increase the contrast between light and dark parts of the image, making the light spots and other important features more obvious. Contrast enhancement helps in the subsequent light spot detection and feature extraction steps. This can be done by adjusting the image's histogram or applying techniques such as histogram equalization. This technique redistributes the brightness of the image so that the overall contrast of the image is improved.

[0028] Edge enhancement: Highlighting the edges in the image helps identify object boundaries and important shape features. In spot analysis, edge enhancement can help define the shape and boundaries of the spot more clearly. By applying edge detection algorithms such as Sobel, Canny, etc. These algorithms identify boundaries by calculating the intensity gradient of pixels in the image.

[0029] Through these preprocessing steps, the light spots and other features in the image will become clearer and easier to identify, thus laying a solid foundation for subsequent image processing and analysis.

[0030] Apply image processing techniques to identify bright areas in the image, i.e., light spots, and use morphological operations to precisely define the boundaries of the light spots; The region growing algorithm is used to identify the bright areas in the image, i.e., the light spots. The specific steps are as follows: Select a seed point, which is the starting point for the region growing algorithm to start growing; Usually, representative points with high brightness in the image are selected as seeds. The seed points can be determined manually or using automated methods (such as selecting the local brightest point).

[0031] Define the growth criterion, which is the rule that determines whether a pixel is added to the seed region; Usually based on the similarity between pixels, such as brightness, color or texture, a threshold is set to compare the similarity between adjacent pixels and the current region. For example, if the difference between the brightness of the adjacent pixel and the average brightness of the current region is less than the threshold, the pixel can be added to the region.

[0032] Start region growing, starting from the seed point, and add adjacent pixels that meet the growth criteria to the current region; Check each pixel around the seed point and determine whether these pixels should be added to the current region based on the growth criteria. After being added, these newly added pixels themselves become new seed points and continue to check the pixels around them.

[0033] Repeat the above growing process until no pixels are added to the region; This is done by continuing to perform the same check and add operation on all pixels surrounding the newly added seed point until the boundaries of the entire region can no longer be extended.

[0034] Output the final region growing result (usually a labeled image of the region), where each region is represented by a different label; This generates a new image in which all pixels in each region are assigned a unique identifier or color to facilitate analysis and further processing.

[0035] Use the morphological dilation operation to accurately define the boundaries of the light spot. The specific steps are as follows: Convert the original color image to grayscale, simplifying the information and focusing on brightness rather than color; The grayscale image is converted into a binary image by thresholding to more clearly distinguish the light spot from the background. This usually involves choosing an appropriate grayscale threshold so that the light spot area is clearly distinguished from other areas.

[0036] Select or define a structuring element (morphological kernel) that determines the refinement and direction of the dilation operation; Common shapes include rectangle, circle, or ellipse. The size and shape of the structuring element should be chosen based on the size of the spot and the desired boundary accuracy. Larger structuring elements can enlarge the area faster but may lose detail, while smaller ones retain more detail but operate more slowly.

[0037] Apply dilation operation to the binarized image; In dilation, the new value of each pixel will depend on the maximum value covered by the structure element shape within that pixel and its neighborhood. This will cause the bright areas (spots) in the image to expand and the boundaries to become more blurred and diffuse.

[0038] The dilated image is processed using edge detection techniques such as Sobel or Canny edge detectors to accurately define and enhance the boundaries of the light spot.

[0039] Extract features from the identified light spots, and divide the light spots into high-impact light spots and low-impact light spots through comprehensive analysis based on the extracted features; The dynamic change information of spot size, spot edge sharpness information and spot intensity gradient information are extracted from the identified light spot. After abnormal analysis and processing of the dynamic change information of spot size, spot edge sharpness information and spot intensity gradient information, the dynamic change index of spot size, the spot edge sharpness index and the spot intensity gradient index are generated. A light spot classification and recognition model is established through the dynamic change index of spot size, the spot edge sharpness index and the spot intensity gradient index, and the light spot influence coefficient is generated. The light spot is divided into a high-impact light spot and a low-impact light spot through the light spot influence coefficient.

[0040] When the dynamic change rate of the spot size of the driver's mirror-reflected light spot is fast, it will have a significant impact on the accuracy of the fatigue detection system. When the size and brightness of the light spot change rapidly in a short period of time, this rapid change may overlap or be confused with the driver's eye movements (such as blinking), making it difficult for the fatigue detection system to accurately identify and analyze the true eye state. For example, a rapidly expanding or brightening light spot may be mistakenly interpreted as the opening of the eyes, while in fact the driver may be blinking frequently or has closed his eyes. Such a misjudgment will lead to inaccurate fatigue assessment results. In addition, the rapid change of the light spot may also cause dynamic blur in the image processing process, further affecting the recognition accuracy of facial expressions and eye features, and thus affecting the system's correct assessment of the driver's fatigue state.

[0041] The logic of obtaining the dynamic change index of the spot size is as follows: At each time frame t, image data about the spot area (image data after grayscale processing) is collected and the image data is used is represented by , where t represents the time frame, Represents the pixel position in the image; The spot area is segmented, and the specific segmentation formula is: , where Represents the set of spot pixels in time frame t, is the grayscale threshold, used to distinguish between light spot and non-light spot areas; Calculate the spot area, the calculation expression is: , where is the spot area of ​​the time frame t, that is The number of pixels in To express; Calculate the spot area change rate between two consecutive frames of images. The calculation expression is: , where is the rate of change of the spot area between two consecutive frames of images; Calculate the dynamic change index of the spot size, and the calculation expression is: ,in, is the dynamic change index of the light spot, n The number of time frames considered depends on the selected time window size. k Is a loop variable used to traverse the past n The spot area changes of the frame, For the past n The absolute value of the spot area change for each frame within a frame.

[0042] From the calculation expression of the dynamic change index of the light spot size, it can be seen that the larger the performance value of the dynamic change index of the light spot size generated after the abnormal analysis and processing of the mirror reflection light spot, the greater the impact of the light spot on the driver fatigue monitoring, and vice versa.

[0043] When the edge sharpness of the driver's specular reflection light spot is high, it will have a negative impact on the accuracy of the fatigue detection system, because the sharp edge of the light spot forms obvious bright lines or bright areas in the image, which may be confused with real facial features, especially the details around the eyes (such as pupils, eyelids and corners of the eyes), causing the automatic vision system to misinterpret the edge of the light spot as part of these key features. This misinterpretation will interfere with the system's accurate recognition and analysis of the driver's eye activities. For example, the edge of the light spot may be misjudged as the position or shape of the pupil, thereby incorrectly assessing the driver's line of sight or the degree of eyelid closure. In addition, light spots with high edge sharpness are more likely to attract visual attention in dynamic environments, which may cause the system to over-focus on these false features and ignore actual fatigue signals, ultimately affecting the accuracy and reliability of fatigue detection.

[0044] The logic of obtaining the light spot edge sharpness index is as follows: Get image data at pixel points The brightness value of In MATLAB, use the imread() function to read the image file, and then use rgb2gray() to convert the color image to grayscale. Provides access to the brightness value of a specific pixel.

[0045] The edge strength of the image data is calculated, and the calculation expression is: , and The images are at points Chu Yan xDirection and y The gradient of the direction (can be calculated by Sobel, Prewitt, etc.), The image is The gradient amplitude at , indicating the edge strength; Calculate the edge sharpness of the image data. The calculation expression is: , where F is the gradient threshold, which is used to determine which gradient values ​​are high enough to be considered a sharp edge, Indicates the edge sharpness value after gradient threshold processing; Calculate the edge sharpness index of the light spot. The calculation expression is: , where Indicates the edge sharpness index of the light spot. S Indicates the light spot area, is the sum of the sharpness values ​​of all pixels in the spot area, is the number of pixels in the spot area.

[0046] From the calculation expression of the light spot edge sharpness index, it can be seen that the larger the performance value of the light spot edge sharpness index generated after the abnormal analysis of the mirror reflection light spot is, the greater the impact of the light spot on the driver fatigue monitoring is, and vice versa.

[0047] The high intensity gradient of the driver's mirror-reflected light spot will cause the pupil and eye expression information to be obscured or distorted by the reflected light spot, thus affecting the accuracy of fatigue detection, because the high intensity gradient means that the brightness of the light spot has a rapid and significant change from the center to the edge. This prominent brightness change can produce obvious visual interference in the image. Such interference will affect the facial recognition and eye tracking technologies used in fatigue detection systems, because these technologies rely on visual information in the image to accurately locate and analyze the pupil position and eye expression. For example, a light spot with a high intensity gradient may cause overexposure or bright spots in the pupil area, making it difficult for the detection algorithm to identify the true pupil boundary, or mistakenly identify the edge of the bright spot as the pupil edge, thereby misjudging the eye state. In addition, if the light spot covers or approaches the key expression muscle area, it may also lead to inaccurate analysis of expression features, affecting the judgment of fatigue status.

[0048] The logic of obtaining the spot intensity gradient index is as follows: Get each pixel image at point Chu Yan x Direction and y Directional gradient and , while acquiring images at The gradient magnitude at (It can be obtained from the step of obtaining the edge sharpness index of the light spot, which will not be described in detail here), calculate at the pixel point The gradient direction of is calculated as: , Represents pixel The gradient direction describes the direction in which the brightness changes most significantly; Calculate the average gradient amplitude in the spot area. The calculation expression is: , B is the pixel set of the spot area, N Light spot area S The total number of pixels in Indicates the spot area S The average gradient magnitude within ; Calculate the spot intensity gradient index, the calculation expression is: , where It indicates the intensity gradient index of the spot. The larger the value of the intensity gradient index, the closer the gradient amplitude of each point in the spot area is to the average gradient. The greater the difference between them.

[0049] From the calculation expression of the light spot intensity gradient index, it can be seen that the larger the performance value of the light spot intensity gradient index generated after the abnormal analysis of the mirror reflection light spot is, the greater the impact of the light spot on the driver fatigue monitoring is, and vice versa.

[0050] The specific implementation method of the spot classification and recognition model is not limited here, and it can realize the dynamic change index of the spot , Spot edge sharpness index and the spot intensity gradient index Perform comprehensive analysis and generate the spot influence coefficient The spot classification and recognition model can be used. In order to realize the technical solution of the present invention, the present invention provides a specific implementation method; Light spot influence coefficient The resulting calculation formula is:

[0051] , where , , They are the spot dynamic change index , Spot edge sharpness index , Spot intensity gradient index The preset scaling factor of , , Both are greater than 0.

[0052] It can be seen from the calculation formula that the larger the performance value of the dynamic change index of the light spot generated after the abnormal analysis of the mirror reflection light spot, the larger the performance value of the light spot edge sharpness index, and the larger the performance value of the light spot intensity gradient index, that is, the larger the performance value of the generated light spot influence coefficient, the greater the impact of the light spot on driver fatigue monitoring, and vice versa.

[0053] After obtaining the spot influence coefficient generated by the abnormal analysis of the mirror reflection spot, the spot influence coefficient is compared and analyzed with the pre-set spot influence coefficient reference threshold. If the spot influence coefficient is greater than or equal to the spot influence coefficient reference threshold, the identified mirror reflection spot is classified as a high-impact spot. If the spot influence coefficient is less than the spot influence coefficient reference threshold, the identified mirror reflection spot is classified as a low-impact spot.

[0054] When there is a low-impact light spot on the mirror, continue to capture the driver's facial image under fill-in light conditions (this is usually during the day); when a high-impact light spot is detected on the mirror, optimize the fill-in light process to reduce the impact of the light spot and ensure efficient monitoring of driver fatigue driving (this is usually at night); When a high-impact light spot is detected on the mirror surface, the steps for optimizing the fill light process are as follows: Change the conventional continuous fill light to pulse fill light; This means that instead of being on continuously, the fill light is turned on briefly in pulses, which allows the "on" time of the fill light to be strictly aligned with the camera's exposure time. In this way, the fill light only provides light at the moment when the camera captures the image, and remains off at other times.

[0055] When the fill light is on, a lighting signal is generated and transmitted to the camera control system, which controls the camera exposure so that the lighting time of the fill light is completely consistent with the exposure time of the camera; Doing so will reduce flare or light reflection issues caused by the fill light being on continuously, especially on reflective surfaces such as glasses.

[0056] When the fill light is on, increase the current passing through the fill light in a short period of time to increase the luminous power of the fill light. When the camera exposure is completed, turn off the fill light (here, adjust the luminous duration of the fill light to be equal to the exposure duration of the camera); By increasing the instantaneous power, even in a very short time, it is possible to provide enough light to ensure that the image is exposed (the luminous power of the fill light is increased to the point where the light intensity emitted meets the exposure intensity required by the camera, so the luminous power adjustment of the fill light is related to the exposure intensity required by the camera and is adjusted according to actual needs). This allows a clear image to be obtained without increasing the risk of reflection caused by long exposure. The entire fill light and exposure process is completed in a very short time. Limiting the length of the fill light helps control the effect of light spots caused by reflections, because long-term strong light can easily form obvious light spots on reflective surfaces, damaging image quality.

[0057] The present invention extracts features from the identified light spots, and divides the light spots into high-impact light spots and low-impact light spots based on a comprehensive analysis of the extracted features. By classifying and identifying the light spots, the system can effectively determine the light spots that seriously affect the recognition of the driver's facial features, so as to take effective measures in time to reduce these interferences. Such classification not only optimizes the data processing flow, but also ensures the accuracy of fatigue detection results, and reduces safety risks caused by misjudgment.

[0058] When the present invention detects the presence of a high-impact light spot on the mirror surface, the conventional fill light mode is adjusted to an optimized processing of pulsed fill light, which can significantly reduce the light spot problem caused by continuous fill light. This optimization ensures that the lighting of the fill light is strictly synchronized with the exposure time of the camera, thereby reducing unnecessary light reflections and light spots. This not only improves the image quality, but also makes fatigue detection more accurate, further improving driving safety.

[0059] 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 in the formula are set by technicians in this field according to actual conditions.

[0060] 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.

[0061] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for suppressing reflections in infrared imaging glasses based on image processing, characterized in that: The following steps are involved: Use the vehicle's infrared camera to capture the driver's facial image under fill-light conditions, ensuring that the camera settings are suitable for capturing the reflection of infrared light on the glasses; Pre-process the captured images to more clearly identify the light spots; Apply image processing techniques to identify bright areas in the image, i.e., light spots, and use morphological operations to precisely define the boundaries of the light spots; Extract features from the identified light spots, and divide the light spots into high-impact light spots and low-impact light spots through comprehensive analysis based on the extracted features; When a low-impact light spot exists on the mirror, the driver's facial image continues to be captured under fill-light conditions. When a high-impact light spot is detected on the mirror, the fill-light process is optimized to reduce the impact of the light spot and ensure efficient monitoring of driver fatigue driving.

2. The method for suppressing reflection of infrared imaging glasses based on image processing according to claim 1, characterized in that: The region growing algorithm is used to identify the bright areas in the image, i.e., the light spots. The specific steps are as follows: Select a seed point, which is the starting point for the region growing algorithm to start growing; Define the growth criterion, which is the rule that determines whether a pixel is added to the seed region; Start region growing, starting from the seed point, and add adjacent pixels that meet the growth criteria to the current region; Repeat the above growing process until no pixels are added to the region; The final region growing result is output, and each region is represented by a different label.

3. The method for suppressing reflection of infrared imaging glasses based on image processing according to claim 1, characterized in that: Use the morphological dilation operation to accurately define the boundaries of the light spot. The specific steps are as follows: Convert the original color image to grayscale, simplifying the information and focusing on brightness rather than color; Select or define a structuring element that determines the refinement and direction of the dilation operation; Apply dilation operation to the binarized image; The dilated image is processed using edge detection techniques to precisely define and enhance the boundaries of the light spot.

4. The method for suppressing reflection of infrared imaging glasses based on image processing according to claim 1, characterized in that: The dynamic change information of spot size, spot edge sharpness information and spot intensity gradient information are extracted from the identified light spot. After abnormal analysis and processing of the dynamic change information of spot size, spot edge sharpness information and spot intensity gradient information, the dynamic change index of spot size, the spot edge sharpness index and the spot intensity gradient index are generated. A light spot classification and recognition model is established through the dynamic change index of spot size, the spot edge sharpness index and the spot intensity gradient index, and the light spot influence coefficient is generated. The light spot is divided into a high-impact light spot and a low-impact light spot through the light spot influence coefficient.

5. The method for suppressing reflection of infrared imaging glasses based on image processing according to claim 4, characterized in that: After obtaining the spot influence coefficient generated by the abnormal analysis of the mirror reflection spot, the spot influence coefficient is compared and analyzed with the pre-set spot influence coefficient reference threshold. If the spot influence coefficient is greater than or equal to the spot influence coefficient reference threshold, the identified mirror reflection spot is classified as a high-impact spot. If the spot influence coefficient is less than the spot influence coefficient reference threshold, the identified mirror reflection spot is classified as a low-impact spot.

6. The method for suppressing reflection of infrared imaging glasses based on image processing according to claim 1, characterized in that: When a high-impact light spot is detected on the mirror surface, the steps for optimizing the fill light process are as follows: Change the conventional continuous fill light to pulse fill light; When the fill light is on, a lighting signal is generated and transmitted to the camera control system, which controls the camera exposure so that the lighting time of the fill light is completely consistent with the exposure time of the camera; When the fill light is on, the current passing through the fill light is increased in a short period of time to increase the luminous power of the fill light. When the camera exposure is completed, the fill light is turned off.

7. The method for suppressing reflection of infrared imaging glasses based on image processing according to claim 4, characterized in that: The logic of obtaining the dynamic change index of the spot size is as follows: Image data about the spot area is collected at each time frame t, and the image data is used is represented by , where t represents the time frame, Represents the pixel position in the image; The spot area is segmented, and the specific segmentation formula is: , where Represents the set of spot pixels in time frame t, is the grayscale threshold, used to distinguish between light spot and non-light spot areas; Calculate the spot area, the calculation expression is: , where is the spot area of ​​the time frame t, that is The number of pixels in To express; Calculate the spot area change rate between two consecutive frames of images. The calculation expression is: , where is the rate of change of the spot area between two consecutive frames of images; Calculate the dynamic change index of the spot size, and the calculation expression is: ,in, is the dynamic change index of the light spot, n is the number of time frames considered, k Is a loop variable used to traverse the past n The spot area changes of the frame, For the past n The absolute value of the spot area change for each frame within a frame.

8. The method for suppressing reflection of infrared imaging glasses based on image processing according to claim 4, characterized in that: The logic of obtaining the light spot edge sharpness index is as follows: Get image data at pixel points The brightness value of The edge strength of the image data is calculated, and the calculation expression is: , and The images are at points Chu Yan x Direction and y The gradient of the direction, The image is The gradient amplitude at , indicating the edge strength; Calculate the edge sharpness of the image data. The calculation expression is: , where F is the gradient threshold, Indicates the edge sharpness value after gradient threshold processing; Calculate the edge sharpness index of the light spot. The calculation expression is: , where Indicates the edge sharpness index of the light spot. S Indicates the light spot area, is the sum of the sharpness values ​​of all pixels in the spot area, is the number of pixels in the spot area.

9. The method for suppressing reflection of infrared imaging glasses based on image processing according to claim 4, characterized in that: The logic of obtaining the spot intensity gradient index is as follows: Get each pixel image at point Chu Yan x Direction and y Directional gradient and , while acquiring images at The gradient magnitude at , calculated at the pixel point The gradient direction of is calculated as: , Represents pixel The gradient direction of Calculate the average gradient amplitude in the spot area. The calculation expression is: , B is the pixel set of the spot area, N Light spot area S The total number of pixels in Indicates the spot area S The average gradient magnitude within ; Calculate the spot intensity gradient index, the calculation expression is: , where It indicates the intensity gradient index of the spot. The larger the value of the intensity gradient index, the closer the gradient amplitude of each point in the spot area is to the average gradient. The greater the difference between them.

Images