Methods, devices, electronic equipment, and storage media for adaptive adjustment of fill lights
By dividing the monitoring equipment into supplementary lighting areas and using a supplementary lighting prediction model to adaptively adjust the intensity of the supplementary lights, the problems of light pollution and overexposure of faces caused by fixed supplementary light brightness are solved, thus improving the quality of monitoring images.
Patent Information
- Application Number
- CN202411966573.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-30
- Publication Date
- 2026-06-30
AI Technical Summary
In existing technologies, the brightness of fill lights is fixed and relatively strong, which has not effectively solved the problems of light pollution and overexposure when people's faces are close.
By acquiring facial monitoring images from surveillance equipment, and using a supplementary lighting prediction model trained based on calibrated sample images and label data, the radiation angle and field of view of each group of supplementary lights are determined, the monitoring area is divided into zones, and feature fusion is performed to adaptively adjust the intensity and combination of supplementary lights to achieve compatibility with bright and dark areas.
It reduces light pollution, avoids overexposure when faces are close together, and improves the image quality of monitoring equipment.
Smart Images

Figure CN122317430A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of video surveillance technology, and in particular to a method, apparatus, electronic device, and storage medium for adaptive adjustment of supplementary lighting. Background Technology
[0002] As the application scenarios for cameras continue to expand, video technology is becoming increasingly comprehensive. To meet the requirements for nighttime video quality, most cameras are now equipped with supplementary lighting such as white and warm lights. Supplementary lights are widely used because they can provide a similar lighting effect to daytime at night, thus providing users with a better video experience. However, because the brightness of supplementary lights is fixed and relatively strong, it can cause light pollution to some extent, and when faces are close, it can lead to overexposure, affecting the usage effect.
[0003] In related technologies, facial brightness is typically calculated by acquiring white light and infrared facial images using a binocular camera, thereby determining the intensity of the fill light and enabling adaptive adjustment of the fill light intensity according to environmental changes. However, the above methods do not solve the overexposure problem when the face is close to the camera. Therefore, how to solve the adaptive adjustment of fill light brightness and the overexposure problem when the face is close to the camera are currently urgent technical problems to be solved. Summary of the Invention
[0004] This invention provides a method, apparatus, electronic device, and storage medium for adaptive adjustment of fill lights, in order to solve the defects of related technologies that cannot address light pollution caused by fill lights and overexposure when faces are close together.
[0005] This invention provides an adaptive adjustment method for supplementary lighting, comprising: Acquire facial surveillance images captured by monitoring equipment; The face monitoring image is input into the supplementary lighting prediction model to obtain the target supplementary lighting combination result corresponding to the monitoring device. The supplementary lighting prediction model is trained based on calibration sample images and label data. The calibration sample images are collected when the face object appears in different supplementary lighting areas corresponding to different groups of supplementary lighting variable parameters. Each supplementary lighting area is determined based on the radiation angle corresponding to each group of supplementary lighting in the monitoring device. The label data is used to characterize the supplementary lighting variable parameters corresponding to the calibration sample images. The target supplementary lighting combination result is obtained by feature fusion of the region images of the supplementary lighting areas corresponding to each group of supplementary lighting in the face monitoring image by the supplementary lighting prediction model. Based on the target fill light combination results, the intensity of each target fill light in the monitoring equipment is adaptively adjusted.
[0006] According to the adaptive adjustment method for supplementary lighting provided by the present invention, the supplementary lighting area corresponding to the monitoring device is determined based on the following steps: Obtain the radiation angles corresponding to at least two sets of supplementary lights in the monitoring equipment; Based on the focal length of the camera corresponding to the monitoring equipment and the radiation angle corresponding to each group of supplementary lights, the field of view width corresponding to each group of supplementary lights is determined; The monitoring area of the monitoring device is divided into partitions based on the field of view width corresponding to each group of supplementary lights, thereby obtaining the supplementary lighting area corresponding to the monitoring device.
[0007] According to the adaptive adjustment method for supplementary lighting provided by the present invention, the supplementary lighting prediction model is trained based on the following steps: For each group of fill lights, the calibration sample images and the label data are input into the initial fill light prediction model. The initial fill light prediction model is used to fuse the calibration sample images of the same group of fill lights that are turned on and the face object is in different fill light areas to obtain the initial fused sample image corresponding to the fill light. The initial fused sample images corresponding to each group of fill lights are enhanced to obtain the first fused sample features corresponding to each group of fill lights. Based on all the first fused sample features and the label data, the trained fill light prediction model is determined.
[0008] According to the adaptive adjustment method for fill lights provided by the present invention, the step of enhancing the initial fusion sample images corresponding to each group of fill lights to obtain the first fusion sample features corresponding to each group of fill lights includes: For each group of fill lights, the mean feature value corresponding to the initial fused sample image is determined, whereby the mean feature value includes the pixel mean or the feature mean. Based on the initial fused sample features corresponding to the initial fused sample image and the mean feature value, edge detail sample features are determined. Based on the edge detail sample features and the initial fused sample features, detail enhancement sample features are determined. Based on the detail enhancement sample features and the initial fused sample features, the first fused sample feature corresponding to the fill light is determined.
[0009] According to the adaptive adjustment method for fill lights provided by the present invention, the step of fusing calibration sample images of the same group of fill lights that are turned on and the face object is located in different fill light areas to obtain an initial fused sample image corresponding to the fill lights includes: Based on the region weights corresponding to each supplementary lighting region and each of the calibration sample images, the initial fusion sample image corresponding to the supplementary lighting is determined.
[0010] According to the adaptive adjustment method for supplementary lighting provided by the present invention, determining the trained supplementary lighting prediction model based on all first fused sample features and the label data includes: S1. Based on the features of all first fusion samples and the fusion weights corresponding to each fill light, the target fusion sample features corresponding to each fill light sample combination are obtained by fusing. S2. Based on the image contrast, spatial frequency, and label data corresponding to the features of each target fusion sample, determine the loss function of the initial supplementary lighting prediction model; the image contrast is used to characterize the brightness difference corresponding to the features of the target fusion sample, and the spatial frequency is used to characterize the edge effect of the location of the face object; S3. Based on the loss function, adjust the model parameters of the initial supplementary lighting prediction model; S4. Repeat steps S1 to S4 until the initial supplementary lighting prediction model converges, and obtain the trained supplementary lighting prediction model.
[0011] According to the adaptive adjustment method for supplementary lighting provided by the present invention, the target supplementary lighting combination result includes the intensity prediction weight corresponding to each of the target supplementary lighting, and the intensity prediction weight is used to characterize the ratio of the intensity corresponding to each of the target supplementary lighting after adjustment to the maximum intensity; The adaptive adjustment of the intensity of each target supplementary light in the monitoring device based on the target supplementary light combination result includes: Based on the light color temperature type corresponding to each of the target fill lights, or the light color temperature type corresponding to each of the target fill lights, and the distance between the target face in the face monitoring image and the monitoring device, the adjustment coefficient corresponding to each of the target fill lights is determined; Based on the adjustment coefficients, the intensity prediction weights corresponding to the target fill lights are corrected to obtain the target intensity weights corresponding to the target fill lights. The intensity of each target fill light is adaptively adjusted based on the target intensity weight corresponding to each target fill light.
[0012] The present invention also provides an adaptive adjustment device for supplementary lighting, comprising: The acquisition module is used to acquire facial surveillance images collected by monitoring equipment; The input module is used to input the face monitoring image into the supplementary lighting prediction model to obtain the target supplementary lighting combination result corresponding to the monitoring device. The supplementary lighting prediction model is trained based on calibration sample images and label data. The calibration sample images are collected when the face object appears in different supplementary lighting areas corresponding to different groups of supplementary lighting variable parameters. Each supplementary lighting area is determined based on the radiation angle corresponding to each group of supplementary lighting in the monitoring device. The label data is used to characterize the supplementary lighting variable parameters corresponding to the calibration sample images. The target supplementary lighting combination result is obtained by feature fusion of the region images of the supplementary lighting areas corresponding to each group of supplementary lighting in the face monitoring image by the supplementary lighting prediction model. The correction module is used to adaptively adjust the intensity of each target fill light in the monitoring device based on the result of the target fill light combination.
[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the adaptive adjustment method for fill light as described above.
[0014] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the adaptive adjustment method for fill light as described above.
[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the adaptive adjustment method for fill light as described above.
[0016] The adaptive adjustment method, device, electronic device, and storage medium for supplementary lighting provided by this invention input real-time acquired face monitoring images into a supplementary lighting prediction model trained based on calibration sample images and label data. The calibration sample images are acquired when a face appears in different supplementary lighting regions corresponding to different groups of supplementary lighting variable parameters after dividing the supplementary lighting region according to the radiation angles corresponding to each group of supplementary lighting in the monitoring device. The label data are the supplementary lighting variable parameters of the calibration sample images. The supplementary lighting prediction model performs feature fusion on the regional images of the supplementary lighting regions corresponding to each group of supplementary lighting in the face monitoring image, outputs the target supplementary lighting combination result corresponding to the monitoring device, and adaptively adjusts the target supplementary lighting corresponding to the monitoring device based on the target supplementary lighting combination result. In this invention, the supplementary lighting area is divided by the radiation angle of each group of supplementary lights. Based on the energy field of each group of supplementary lights, more precise supplementary lighting intensity is provided to face objects appearing in different areas. At the same time, the regional images of supplementary lighting areas with different intensities in the face monitoring image are fused according to the supplementary lighting prediction model, which is compatible with bright and dark areas in the face monitoring image and reduces the overexposure problem when the face is close. Furthermore, based on the target supplementary lighting combination result obtained after fusion, the intensity of each target supplementary light is adaptively adjusted, thereby improving the image effect of the monitoring equipment. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in this invention or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1This is a flowchart illustrating the adaptive adjustment method for supplementary lighting provided in an embodiment of the present invention.
[0019] Figure 2 This is one of the schematic diagrams of the supplementary lighting areas of each group of supplementary lights provided in the embodiments of the present invention.
[0020] Figure 3 This is the second schematic diagram of the supplementary lighting area of each group of supplementary lights provided in the embodiments of the present invention.
[0021] Figure 4 This is a schematic diagram of the process for determining the features of the first fused sample provided in an embodiment of the present invention.
[0022] Figure 5 This is a schematic diagram of the adaptive adjustment device for supplementary lighting provided in an embodiment of the present invention.
[0023] Figure 6 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0025] To address the issues of light pollution and overexposure when faces are close to the subject that cannot be resolved by related technologies, this invention provides an adaptive adjustment method for fill lights. Figure 1 This is a flowchart illustrating the adaptive adjustment method for supplementary lighting provided in an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes steps 110 to 130.
[0026] Step 110: Obtain facial monitoring images captured by the monitoring equipment.
[0027] It should be noted that the facial surveillance image is captured in real time by the monitoring equipment when a face appears in the camera lens, and there is at least one face.
[0028] Optionally, the monitoring device can be a bullet camera, a PTZ camera, or a pan-tilt camera, etc., and the embodiments of the present invention do not limit this.
[0029] The monitoring equipment includes at least two sets of supplementary lights, each with a different radiation angle, resulting in varying radiation range and distance. Specifically, a smaller radiation angle allows for a longer radiation distance but a smaller radiation range, resulting in a bright area with large dark areas on either side. Conversely, a larger radiation angle allows for a shorter radiation distance but a larger radiation range with smaller dark areas on either side. Therefore, if a face appears within the radiation range of any of the supplementary lights in the face monitoring image and is close to the monitoring equipment, the face will be overexposed. If a face appears outside the radiation range of any of the supplementary lights, it will be in a dark area and may not be visible. Therefore, before performing step 120, the monitoring area of the monitoring equipment needs to be divided into zones.
[0030] Furthermore, the supplementary lighting area corresponding to the monitoring device is determined based on the following steps: Obtain the radiation angles corresponding to at least two sets of supplementary lights in the monitoring equipment; Based on the focal length of the camera corresponding to the monitoring equipment and the radiation angle corresponding to each group of supplementary lights, the field of view width corresponding to each group of supplementary lights is determined; The monitoring area of the monitoring device is divided into partitions based on the field of view width corresponding to each group of supplementary lights, thereby obtaining the supplementary lighting area corresponding to the monitoring device.
[0031] The radiation angle of each group of supplementary lights in the monitoring equipment is the same as the angle of the LED in each group of supplementary lights. This LED angle can be... , or However, this embodiment of the invention does not limit the scope of the invention. Each group of supplementary lights may include at least one supplementary light. Furthermore, the radiation range of each group of supplementary lights may be independent or may have overlapping areas.
[0032] For example, consider a monitoring device that includes three sets of supplementary lights, where the radiation ranges of the three sets of supplementary lights overlap. Figure 2 This is one of the schematic diagrams of the supplementary lighting areas of each group of supplementary lights provided in the embodiments of the present invention, such as... Figure 2 As shown, the third group of supplementary lights has the largest radiation angle, followed by the second group, and the first group has the smallest radiation angle. Then, using equation (1), based on the focal length of the camera in the monitoring equipment and the radiation angles of each group of supplementary lights, the field of view width corresponding to each group of supplementary lights is calculated. This field of view width can be considered as the radiation range of each group of supplementary lights. Equation (1) is: , Among them, W iThis represents the field of view or radiation range corresponding to the i-th group of supplementary lights. This represents the radiation angle or LED angle corresponding to the i-th group of supplementary lights, and f represents the focal length of the camera on the monitoring device. The radiation angle and camera focal length can be determined by viewing the configuration parameters of the monitoring device. .
[0033] Since the radiation ranges of the three sets of supplementary lights overlap, i.e., the field of view W1 corresponding to the first set of supplementary lights is included in the field of view W2 corresponding to the second set of supplementary lights, and the field of view W2 corresponding to the second set of supplementary lights is included in the field of view W3 corresponding to the third set of supplementary lights, therefore, according to equation (2), the supplementary lighting areas in the monitoring equipment are determined by using the field of view corresponding to each set of supplementary lights. Equation (2) is: N i =W i -W i-1 , Where, N i This represents the illumination area corresponding to the i-th group of supplementary lights. When i=1, since there is no zero-th group of supplementary lights in the monitoring equipment, therefore, W i-1 When i=0, the supplementary lighting area N1 corresponding to the first group of supplementary lights is the radiation range corresponding to the first group of supplementary lights. When i=2, the supplementary lighting area N2 corresponding to the second group of supplementary lights is the overlapping area between the dark areas on both sides of the radiation range corresponding to the first group of supplementary lights and the radiation range corresponding to the second group of supplementary lights. Similarly, when i=3, the supplementary lighting area N3 corresponding to the third group of supplementary lights is the overlapping area between the dark areas on both sides of the radiation range corresponding to the second group of supplementary lights and the radiation range corresponding to the third group of supplementary lights.
[0034] For example, Figure 3 This is a second schematic diagram of the supplementary lighting area of each group of supplementary lights provided in the embodiments of the present invention, as shown in the figure. Figure 3 As shown, taking a monitoring device that includes three sets of supplementary lights and whose radiation ranges do not overlap as an example, the field width of each set of supplementary lights can be calculated according to formula (1). Since the radiation ranges of the three sets of supplementary lights do not overlap, the field width W1 of the first set of supplementary lights can be determined as the supplementary lighting area N1 of the first set of supplementary lights, the field width W2 of the second set of supplementary lights can be determined as the supplementary lighting area N2 of the second set of supplementary lights, and the field width W3 of the third set of supplementary lights can be determined as the supplementary lighting area N3 of the third set of supplementary lights.
[0035] It should be noted that the number of supplementary lighting units in the monitoring equipment and the corresponding radiation angle of each unit depend on the equipment model and its purpose. If the configuration of the supplementary lighting units changes, the monitoring area needs to be re-divided according to the updated supplementary lighting units.
[0036] like Figure 2 As shown, after dividing the lighting areas, if a face appears in both lighting area N1 and lighting area N3, turning on only the first set of lights will cause the face in lighting area N1 to be overexposed, while the face in lighting area N3 will be too dark, failing to guarantee image quality. Furthermore, the different distances from the monitoring device or lights in different areas will result in varying brightness. Therefore, it is necessary to simultaneously select appropriate target lights and adjust their intensity to balance the image quality of both bright and dark areas, avoiding overexposure or underexposure. Thus, before executing step 120, an image fusion method can be used to train the initial lighting prediction model. After training the lighting prediction model, the best combination of target lights can be selected to fuse bright and dark areas, adaptively adjusting the brightness of the lights to improve image quality and avoid overexposure or underexposure.
[0037] Before training the initial supplementary lighting prediction model, it is necessary to collect calibration sample images and corresponding label data. For each group of supplementary lights in the monitoring equipment, the controlled variable method is used. Based on the variable parameters of different groups of supplementary lights in Table 1, a large number of face calibration sample images under different scenarios are collected, with each group of supplementary light variable parameters corresponding to one scenario. For example, taking a monitoring equipment with three groups of supplementary lights, the supplementary light variable parameters may include: environment category distinguished by ambient brightness, image brightness, intensity of the first group of supplementary lights, intensity of the second group of supplementary lights, intensity of the third group of supplementary lights, the supplementary lighting area where the face is located, and the brightness of the face, etc. When acquiring calibration sample images, only one set of supplementary lights can be turned on in each environment. Images are acquired when the face is positioned in supplementary lighting region N1, N2, or N3. For example, in a brighter environment, only the first set of supplementary lights is turned on, acquiring three calibration sample images when the face is in supplementary lighting region N1, N2, or N3; in a brighter environment, only the second set of supplementary lights is turned on, acquiring three more images; and in a brighter environment, only the third set of supplementary lights is turned on, acquiring three more images. Repeating this process yields all calibration sample images. The supplementary light variable parameters corresponding to each calibration sample image constitute the label data.
[0038] Table 1
[0039] It should be noted that in Table 1, E ij This represents the image brightness at the j-th sample corresponding to the activation of the i-th group of fill lights. , j is an integer and j>0, Dij This represents the face brightness of the j-th sample corresponding to the i-th group of fill lights being turned on.
[0040] It should be noted that when each group of supplementary lights is turned on for data acquisition, calibration sample images corresponding to different supplementary light intensities can be acquired. For example, when the intensity of the first group of supplementary lights is A1, three calibration sample images can be acquired when the face is located in supplementary light area N1, supplementary light area N2, or supplementary light area N3, respectively; when the intensity of the first group of supplementary lights is A2, three more calibration sample images can be acquired when the face is located in supplementary light area N1, supplementary light area N2, or supplementary light area N3, etc.
[0041] Furthermore, the supplemental lighting prediction model is trained based on the following steps: For each group of fill lights, the calibration sample images and the label data are input into the initial fill light prediction model. The initial fill light prediction model is used to fuse the features of the calibration sample images where the same group of fill lights is turned on and the face object is in different fill light areas to obtain the initial fused sample image corresponding to the fill light. The initial fused sample images corresponding to each group of fill lights are enhanced to obtain the first fused sample features corresponding to each group of fill lights. Based on all the first fused sample features and the label data, the trained fill light prediction model is determined.
[0042] Specifically, after acquiring the calibration sample images and label data, the calibration sample images and corresponding label data are input into the initial supplementary lighting prediction model. Within the initial supplementary lighting prediction model, feature fusion is performed on calibration sample images of faces appearing in different supplementary lighting regions when the same group of supplementary lights is turned on. For example, when the first group of supplementary lights is turned on, the calibration sample image p1 of a face appearing in supplementary lighting region N1, the calibration sample image p2 of a face appearing in supplementary lighting region N2, and the calibration sample image p3 of a face appearing in supplementary lighting region N3 are fused to obtain the initial fused sample image p'1 corresponding to the first group of supplementary lights being turned on. Repeating the above steps, the initial fused sample images corresponding to each supplementary light are obtained. Then, since there are at least two face objects, each initial fused sample image can be enhanced to obtain the features of each first fused sample, making the edge details corresponding to each face object more prominent. Finally, the model parameters of the initial supplementary lighting prediction model are adjusted based on all the features of the first fused samples and the label data until the model converges, resulting in a trained supplementary lighting prediction model.
[0043] It should be noted that the initial supplemental lighting prediction model can be constructed using methods such as convolutional neural networks, graph neural networks, or neural networks based on self-attention mechanisms, and the embodiments of the present invention do not limit this.
[0044] Further, the step of fusing features from calibration sample images of the same group of fill lights and the face object located in different fill light areas to obtain an initial fused sample image corresponding to the fill light includes: Based on the region weights corresponding to each supplementary lighting region and each of the calibration sample images, the initial fusion sample image corresponding to the supplementary lighting is determined.
[0045] For example, when the first set of fill lights is turned on, the region weights w1, w2, and w3 corresponding to fill light region N1, N2, and N3 are obtained. Then, key features, such as edge features, texture features, and brightness features, are extracted from the calibration sample images p1, p2, and p3 where the face appears in fill light region N1, N2, and N3, respectively. These features are then weighted and fused with their corresponding region weights. For instance, the product of texture feature F1 and region weight w1 corresponding to calibration sample image p1, the product of texture feature F2 and region weight w2 corresponding to fill light region N2, and the product of texture feature F3 and region weight w3 corresponding to fill light region N3 are calculated. Finally, the sum of these three products, F'1, is calculated. The sum is the fused texture feature after fusing the three calibration sample images. By repeating the above operation, other fused features after fusing the three calibration sample images can be obtained. By integrating the above fused features, the initial fused sample image p'1 corresponding to only turning on the first set of fill lights can be obtained.
[0046] Furthermore, Figure 4 This is a schematic diagram of the process for determining the first fused sample features provided in an embodiment of the present invention, as shown below. Figure 4 As shown, the step of enhancing the initial fused sample images corresponding to each group of fill lights to obtain the first fused sample features corresponding to each group of fill lights includes: For each group of fill lights, the mean feature value corresponding to the initial fused sample image is determined. Based on the initial fused sample features and the mean feature value, edge detail sample features are determined. Based on the edge detail sample features and the initial fused sample features, detail enhancement sample features are determined. Based on the detail enhancement sample features and the initial fused sample features, the first fused sample features corresponding to the fill light are determined.
[0047] Specifically, for the initial fusion sample images corresponding to each group of fill lights, the initial fusion sample features can be extracted from the initial fusion sample images, and the feature mean of the initial fusion sample features can be calculated. Then, the feature mean is subtracted from the initial fusion sample features in the initial fusion sample images to obtain the edge detail sample features. Next, the edge detail sample features can be normalized to the range of (0,1). Then, the normalized edge detail sample features are multiplied with the initial fusion sample features to obtain the detail enhancement sample features. Finally, the detail enhancement sample features are added to the initial fusion sample features to obtain the first fusion sample features that highlight the edge details, thus highlighting the edge details corresponding to each fill light area while retaining the original features.
[0048] Further, determining the trained supplemental lighting prediction model based on all the features of the first fused samples and the label data includes: S1. Based on the features of all first fusion samples and the fusion weights corresponding to each fill light, the target fusion sample features corresponding to each fill light sample combination are obtained by fusing. S2. Based on the image contrast, spatial frequency, and label data corresponding to the features of each target fusion sample, determine the loss function of the initial supplementary lighting prediction model; the image contrast is used to characterize the brightness difference corresponding to the features of the target fusion sample, and the spatial frequency is used to characterize the edge effect of the location of the face object; S3. Based on the loss function, adjust the model parameters of the initial supplementary lighting prediction model; S4. Repeat steps S1 to S4 until the initial supplementary lighting prediction model converges, and obtain the trained supplementary lighting prediction model.
[0049] Specifically, after determining all the first fusion sample features and label data, the initial supplementary lighting prediction model can further perform weighted fusion of the first fusion sample features according to the preset initial fusion weights corresponding to each supplementary light, to obtain the target fusion sample features corresponding to different supplementary light sample combinations. For example, the first fusion sample feature G1 corresponding to only the first group of supplementary lights being turned on can be multiplied by the initial fusion weight w'1, and the first fusion sample feature G3 corresponding to only the third group of supplementary lights being turned on can be multiplied by the initial fusion weight w'3. The two products are then added together to achieve weighted fusion of the first fusion sample features G1 and G3, resulting in the target fusion sample feature H corresponding to the supplementary light sample combination. .
[0050] Subsequently, due to the strong reflectivity of the face object to the supplementary lighting, its reflectivity differs significantly from that of other objects in the environment. Furthermore, as the face object approaches the monitoring equipment, its reflectivity increases, leading to overexposure and making it impossible to obtain face information. Therefore, different supplementary lighting intensities need to be set according to different locations. Thus, after determining the characteristics of each target fusion sample, equation (3) can be used to calculate the image contrast of the target fusion sample feature based on the maximum and minimum brightness features. This image contrast is used to characterize the brightness difference in the target fusion sample image corresponding to the target fusion sample feature, i.e., it reflects the brightness level in the target fusion sample image. When the face object is in different supplementary lighting areas, the image contrast will vary significantly. Therefore, image contrast can be used to constrain the distance from the face object to the monitoring equipment. Equation (3) is: Con=(Lmax-Lmin) / (Lmax+Lmin).
[0051] Where Con represents image contrast, Lmax represents the maximum brightness in the target fused sample features, and Lmin represents the minimum brightness in the target fused sample features.
[0052] Simultaneously, spatial frequency can be calculated based on the features of the target fusion sample. This spatial frequency characterizes the edge effect of a face object appearing in different lighting regions. Image quality is evaluated using image contrast and spatial frequency, and these serve as constraints in the initial lighting prediction model training process. During training, the true weights are determined based on the intensity of each lighting lamp in the label data. These true weights can be the ratio of the intensity of each lighting lamp in the lighting lamp sample combination corresponding to the target fusion sample features to the maximum intensity. A loss function is calculated based on image contrast and spatial frequency, and the model parameters are adjusted through backpropagation based on the loss function to minimize the difference between the output predicted weights and the true weights in the label data. This process continues until the model converges or reaches a preset number of iterations, resulting in a well-trained lighting prediction model.
[0053] Step 120: Input the face monitoring image into the supplementary lighting prediction model to obtain the target supplementary lighting combination result corresponding to the monitoring device; the supplementary lighting prediction model is trained based on calibration sample images and label data; the calibration sample images are collected when the face object appears in different supplementary lighting areas corresponding to different groups of supplementary lighting variable parameters; each supplementary lighting area is determined based on the radiation angle corresponding to each group of supplementary lighting in the monitoring device; the label data is used to characterize the supplementary lighting variable parameters corresponding to the calibration sample images; the target supplementary lighting combination result is obtained by feature fusion of the region images of the supplementary lighting areas corresponding to each group of supplementary lighting in the face monitoring image by the supplementary lighting prediction model.
[0054] Specifically, after determining the face surveillance image, the face surveillance image can be input into a trained supplementary lighting prediction model. The supplementary lighting prediction model performs feature fusion on the regional images of the supplementary lighting areas corresponding to each group of supplementary lights to obtain the target supplementary lighting combination result. The target supplementary lighting combination result includes the intensity prediction weight corresponding to each target supplementary light, which is used to characterize the ratio of the intensity of each target supplementary light after adjustment to the maximum intensity.
[0055] Step 130: Based on the target fill light combination result, adaptively adjust the intensity of each target fill light in the monitoring device.
[0056] Specifically, after determining the target supplementary lighting combination result, the target intensity of each target supplementary lighting can be obtained by calculating the product of the intensity prediction weight and the maximum intensity of each target supplementary lighting in the target supplementary lighting combination result. The target intensity corresponding to the intensity adjustment value of each target supplementary lighting in the monitoring equipment is then used to avoid light pollution caused by excessive intensity of each target supplementary lighting.
[0057] Furthermore, the adaptive adjustment of the intensity of each target supplementary light in the monitoring device based on the target supplementary light combination result includes: Based on the light color temperature type corresponding to each of the target fill lights, or the light color temperature type corresponding to each of the target fill lights, and the distance between the target face in the face monitoring image and the monitoring device, the adjustment coefficient corresponding to each of the target fill lights is determined; Based on the adjustment coefficients, the intensity prediction weights corresponding to the target fill lights are corrected to obtain the target intensity weights corresponding to the target fill lights. The intensity of each target fill light is adaptively adjusted based on the target intensity weight corresponding to each target fill light.
[0058] Specifically, for each target supplementary light, due to the different color temperatures of each light, and the fact that the wavelength of the light source in the supplementary light becomes weaker as the distance between the face and the monitoring equipment increases, color differences occur. Therefore, in face monitoring images, the color of the supplementary lighting area corresponding to different target supplementary lights may differ at different distances. The light color temperature types of supplementary lights in monitoring equipment include warm light and white light. Warm light has a lower color temperature (e.g., a color temperature range of 2700K to 3200K), emitting yellow or warm white light. When the distance between the face and the monitoring equipment is close and the intensity of the warm light is too high, in addition to causing face overexposure, it will also cause color cast in the image. On the other hand, white light has a higher color temperature (e.g., a color temperature range of 4000K to 5000K), emitting white or cool white light. When the distance between the face and the monitoring equipment is close and the intensity of the white light is too high, it will only cause face overexposure and will not cause color difference in the image. Therefore, after determining the target supplementary lighting combination result output by the supplementary lighting prediction model, the adjustment coefficient of each target supplementary lighting can be determined according to the light color temperature type corresponding to each target supplementary lighting included in the target supplementary lighting combination result. The adjustment coefficient is shown in Equation (4), which is: , in, d represents the adjustment coefficient corresponding to the target supplement light of the i-th group, and d represents the distance between the face object and the monitoring device. d can be determined by determining the proportion of the pixels of the face object in the face monitoring image to the total pixels in the face monitoring image.
[0059] Then, according to equation (5), the intensity prediction weight of each target supplementary light is corrected, that is, the product of the intensity prediction weight of each target supplementary light and the corresponding adjustment coefficient is calculated, and this product is the target intensity weight corresponding to each target supplementary light. Equation (5) is: .
[0060] Where Str represents the corrected target fill light combination weight, and w_k represents the intensity prediction weight corresponding to the k-th group of target fill lights. For example, if the target fill light combination result includes the intensity prediction weight w_1 corresponding to the first group of fill lights and the intensity prediction weight w_3 corresponding to the third group of fill lights, then... Then, by calculating the product of the target intensity weight and the maximum intensity of each target fill light, the adjusted target intensity of each target fill light can be obtained. The target intensity corresponding to the intensity adjustment value of each target fill light in the monitoring equipment is then set to avoid light pollution caused by excessive intensity of each target fill light. At the same time, the problem of overexposure of faces when objects are close by and the problem of uneven color in the overall image caused by turning on the fill lights are solved, thereby improving the image effect.
[0061] The adaptive adjustment method for supplementary lighting provided in this invention involves inputting real-time captured face monitoring images into a supplementary lighting prediction model trained based on calibration sample images and label data. The calibration sample images are obtained by dividing the supplementary lighting area according to the radiation angles of each group of supplementary lights in the monitoring device, and then capturing images when a face appears in different supplementary lighting areas corresponding to different groups of supplementary light variable parameters. The label data consists of the supplementary light variable parameters of the calibration sample images. The supplementary lighting prediction model performs feature fusion on the regional images of the supplementary lighting areas corresponding to each group of supplementary lights in the face monitoring image, outputs the target supplementary lighting combination result corresponding to the monitoring device, and adaptively adjusts the target supplementary lighting corresponding to the monitoring device based on the target supplementary lighting combination result. In this invention, the supplementary lighting area is divided by the radiation angle of each group of supplementary lights. Based on the energy field of each group of supplementary lights, more precise supplementary lighting intensity is provided to face objects appearing in different areas. At the same time, the regional images of supplementary lighting areas with different intensities in the face monitoring image are fused according to the supplementary lighting prediction model, which is compatible with bright and dark areas in the face monitoring image and reduces the overexposure problem when the face is close. Furthermore, based on the target supplementary lighting combination result obtained after fusion, the intensity of each target supplementary light is adaptively adjusted, thereby improving the image effect of the monitoring equipment.
[0062] The adaptive adjustment device for supplementary lighting provided by the present invention is described below. The adaptive adjustment device for supplementary lighting described below can be referred to in correspondence with the adaptive adjustment method for supplementary lighting described above.
[0063] This invention also provides an adaptive adjustment device for supplementary lighting. Figure 5 This is a schematic diagram of the adaptive adjustment device for supplementary lighting provided in an embodiment of the present invention, as shown below. Figure 5 As shown, the adaptive adjustment device 500 for the supplementary lighting includes: an acquisition module 510, an input module 520, and a correction module 530.
[0064] The acquisition module 510 is used to acquire facial monitoring images collected by the monitoring equipment.
[0065] The input module 520 is used to input the face monitoring image into the supplementary lighting prediction model to obtain the target supplementary lighting combination result corresponding to the monitoring device; the supplementary lighting prediction model is trained based on calibration sample images and label data; the calibration sample images are collected when the face object appears in different supplementary lighting areas corresponding to different groups of supplementary lighting variable parameters; each supplementary lighting area is determined based on the radiation angle corresponding to each group of supplementary lighting in the monitoring device; the label data is used to characterize the supplementary lighting variable parameters corresponding to the calibration sample images; the target supplementary lighting combination result is obtained by feature fusion of the region images of the supplementary lighting areas corresponding to each group of supplementary lighting in the face monitoring image by the supplementary lighting prediction model.
[0066] The correction module 530 is used to adaptively adjust the target fill light corresponding to the monitoring device based on the target fill light combination result.
[0067] The adaptive adjustment device for supplementary lighting provided by this invention inputs real-time acquired face monitoring images into a supplementary lighting prediction model trained based on calibration sample images and label data. The calibration sample images are acquired when a face appears in different supplementary lighting areas corresponding to different groups of supplementary lighting variable parameters after dividing the supplementary lighting area according to the radiation angles of each group of supplementary lighting in the monitoring device. The label data are the supplementary lighting variable parameters of the calibration sample images. The supplementary lighting prediction model performs feature fusion on the regional images of the supplementary lighting areas corresponding to each group of supplementary lighting in the face monitoring image, outputs the target supplementary lighting combination result corresponding to the monitoring device, and adaptively adjusts the target supplementary lighting corresponding to the monitoring device based on the target supplementary lighting combination result. In this invention, the supplementary lighting area is divided by the radiation angle of each group of supplementary lights. Based on the energy field of each group of supplementary lights, more precise supplementary lighting intensity is provided to face objects appearing in different areas. At the same time, the regional images of supplementary lighting areas with different intensities in the face monitoring image are fused according to the supplementary lighting prediction model, which is compatible with bright and dark areas in the face monitoring image and reduces the overexposure problem when the face is close. Furthermore, based on the target supplementary lighting combination result obtained after fusion, the intensity of each target supplementary light is adaptively adjusted, thereby improving the image effect of the monitoring equipment.
[0068] Optionally, the adaptive adjustment device 500 for the supplementary lighting also includes a training module, which is specifically used for: Obtain the radiation angles corresponding to at least two sets of supplementary lights in the monitoring equipment; Based on the focal length of the camera corresponding to the monitoring equipment and the radiation angle corresponding to each group of supplementary lights, the field of view width corresponding to each group of supplementary lights is determined; The monitoring area of the monitoring device is divided into partitions based on the field of view width corresponding to each group of supplementary lights, thereby obtaining the supplementary lighting area corresponding to the monitoring device.
[0069] Optionally, this training module is specifically used for: For each group of fill lights, the calibration sample images and the label data are input into the initial fill light prediction model. The initial fill light prediction model is used to fuse the features of the calibration sample images where the same group of fill lights is turned on and the face object is in different fill light areas to obtain the initial fused sample image corresponding to the fill light. The initial fused sample images corresponding to each group of fill lights are enhanced to obtain the first fused sample features corresponding to each group of fill lights. Based on all the first fused sample features and the label data, the trained fill light prediction model is determined.
[0070] Optionally, this training module is specifically used for: For each group of fill lights, the mean feature value corresponding to the initial fused sample image is determined; based on the initial fused sample features and the mean feature value, edge detail sample features are determined; based on the edge detail sample features and the initial fused sample features, detail enhancement sample features are determined; based on the detail enhancement sample features and the initial fused sample features, the first fused sample feature corresponding to the fill light is determined.
[0071] Optionally, this training module is specifically used for: Based on the region weights corresponding to each supplementary lighting region and each of the calibration sample images, the initial fusion sample image corresponding to the supplementary lighting is determined.
[0072] Optionally, this training module is specifically used for: S1. Based on all the features of the first fusion sample and the fusion weights corresponding to each fill light, the target fusion sample features corresponding to each fill light sample combination in the label data are obtained by fusing them together. S2. Based on the image contrast, spatial frequency, and label data corresponding to the features of each target fusion sample, determine the loss function of the initial supplementary lighting prediction model; the image contrast is used to characterize the brightness difference corresponding to the features of the target fusion sample, and the spatial frequency is used to characterize the edge effect of the location of the face object; S3. Based on the loss function, adjust the model parameters of the initial supplementary lighting prediction model; S4. Repeat steps S1 to S4 until the initial supplementary lighting prediction model converges, and obtain the trained supplementary lighting prediction model.
[0073] Optionally, the target fill light combination result includes the intensity prediction weight corresponding to each target fill light, and the intensity prediction weight is used to characterize the ratio of the intensity corresponding to each target fill light after adjustment to the maximum intensity.
[0074] Optionally, the correction module 530 is specifically used for: Based on the light color temperature type corresponding to each of the target fill lights, or the light color temperature type corresponding to each of the target fill lights, and the distance between the target face in the face monitoring image and the monitoring device, the adjustment coefficient corresponding to each of the target fill lights is determined; Based on the adjustment coefficients, the intensity prediction weights corresponding to the target fill lights are corrected to obtain the target intensity weights corresponding to the target fill lights. The intensity of each target fill light is adaptively adjusted based on the target intensity weight corresponding to each target fill light.
[0075] Figure 6This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present invention, such as... Figure 6 As shown, the electronic device may include: a processor 610, a communications interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communications interface 620, and the memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute an adaptive adjustment method for the supplementary lighting, the method including: Acquire facial surveillance images captured by monitoring equipment; The face monitoring image is input into the supplementary lighting prediction model to obtain the target supplementary lighting combination result corresponding to the monitoring device. The supplementary lighting prediction model is trained based on calibration sample images and label data. The calibration sample images are collected when the face object appears in different supplementary lighting areas corresponding to different groups of supplementary lighting variable parameters. Each supplementary lighting area is determined based on the radiation angle corresponding to each group of supplementary lighting in the monitoring device. The label data is used to characterize the supplementary lighting variable parameters corresponding to the calibration sample images. The target supplementary lighting combination result is obtained by feature fusion of the region images of the supplementary lighting areas corresponding to each group of supplementary lighting in the face monitoring image by the supplementary lighting prediction model. Based on the target fill light combination results, the intensity of each target fill light in the monitoring equipment is adaptively adjusted.
[0076] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0077] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the adaptive adjustment method for supplementary lighting provided by the above methods, the method comprising: Acquire facial surveillance images captured by monitoring equipment; The face monitoring image is input into the supplementary lighting prediction model to obtain the target supplementary lighting combination result corresponding to the monitoring device. The supplementary lighting prediction model is trained based on calibration sample images and label data. The calibration sample images are collected when the face object appears in different supplementary lighting areas corresponding to different groups of supplementary lighting variable parameters. Each supplementary lighting area is determined based on the radiation angle corresponding to each group of supplementary lighting in the monitoring device. The label data is used to characterize the supplementary lighting variable parameters corresponding to the calibration sample images. The target supplementary lighting combination result is obtained by feature fusion of the region images of the supplementary lighting areas corresponding to each group of supplementary lighting in the face monitoring image by the supplementary lighting prediction model. Based on the target fill light combination results, the intensity of each target fill light in the monitoring equipment is adaptively adjusted.
[0078] In another aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the adaptive adjustment method for supplementary lighting provided by the methods described above, the method comprising: Acquire facial surveillance images captured by monitoring equipment; The face monitoring image is input into the supplementary lighting prediction model to obtain the target supplementary lighting combination result corresponding to the monitoring device. The supplementary lighting prediction model is trained based on calibration sample images and label data. The calibration sample images are collected when the face object appears in different supplementary lighting areas corresponding to different groups of supplementary lighting variable parameters. Each supplementary lighting area is determined based on the radiation angle corresponding to each group of supplementary lighting in the monitoring device. The label data is used to characterize the supplementary lighting variable parameters corresponding to the calibration sample images. The target supplementary lighting combination result is obtained by feature fusion of the region images of the supplementary lighting areas corresponding to each group of supplementary lighting in the face monitoring image by the supplementary lighting prediction model. Based on the target fill light combination results, the intensity of each target fill light in the monitoring equipment is adaptively adjusted.
[0079] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0080] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for adaptive adjustment of a supplementary light, characterized in that, include: Acquire facial surveillance images captured by monitoring equipment; The face monitoring image is input into the supplementary lighting prediction model to obtain the target supplementary lighting combination result corresponding to the monitoring device. The supplementary lighting prediction model is trained based on calibration sample images and label data. The calibration sample images are collected when the face object appears in different supplementary lighting areas corresponding to different groups of supplementary lighting variable parameters. Each supplementary lighting area is determined based on the radiation angle corresponding to each group of supplementary lighting in the monitoring device. The label data is used to characterize the supplementary lighting variable parameters corresponding to the calibration sample images. The target supplementary lighting combination result is obtained by feature fusion of the region images of the supplementary lighting areas corresponding to each group of supplementary lighting in the face monitoring image by the supplementary lighting prediction model. Based on the target fill light combination results, the intensity of each target fill light in the monitoring equipment is adaptively adjusted.
2. The adaptive adjustment method for supplementary lighting according to claim 1, characterized in that, The supplementary lighting area corresponding to the monitoring equipment is determined based on the following steps: Obtain the radiation angles corresponding to at least two sets of supplementary lights in the monitoring equipment; Based on the focal length of the camera corresponding to the monitoring equipment and the radiation angle corresponding to each group of supplementary lights, the field of view width corresponding to each group of supplementary lights is determined; The monitoring area of the monitoring device is divided into partitions based on the field of view width corresponding to each group of supplementary lights, thereby obtaining the supplementary lighting area corresponding to the monitoring device.
3. The adaptive adjustment method for supplementary lighting according to claim 1, characterized in that, The supplemental lighting prediction model was trained based on the following steps: For each group of fill lights, the calibration sample images and the label data are input into the initial fill light prediction model. The initial fill light prediction model is used to fuse the features of the calibration sample images where the same group of fill lights is turned on and the face object is in different fill light areas to obtain the initial fused sample image corresponding to the fill light. The initial fused sample images corresponding to each group of fill lights are enhanced to obtain the first fused sample features corresponding to each group of fill lights. Based on all the first fused sample features and the label data, the trained fill light prediction model is determined.
4. The adaptive adjustment method for supplementary lighting according to claim 3, characterized in that, The step of enhancing the initial fused sample images corresponding to each group of supplementary lights to obtain the first fused sample features corresponding to each group of supplementary lights includes: For each group of fill lights, the mean feature value corresponding to the initial fused sample image is determined; based on the initial fused sample features and the mean feature value, edge detail sample features are determined; based on the edge detail sample features and the initial fused sample features, detail enhancement sample features are determined; based on the detail enhancement sample features and the initial fused sample features, the first fused sample feature corresponding to the fill light is determined.
5. The adaptive adjustment method for supplementary lighting according to claim 3, characterized in that, The process of fusing calibration sample images of faces located in different lighting regions with the same group of fill lights turned on to obtain an initial fused sample image corresponding to the fill lights includes: Based on the region weights corresponding to each supplementary lighting region and each of the calibration sample images, the initial fusion sample image corresponding to the supplementary lighting is determined.
6. The adaptive adjustment method for supplementary lighting according to claim 3, characterized in that, The step of determining the trained supplemental lighting prediction model based on all the features of the first fused samples and the label data includes: S1. Based on the features of all first fusion samples and the fusion weights corresponding to each fill light, the target fusion sample features corresponding to each fill light sample combination are obtained by fusing. S2. Based on the image contrast, spatial frequency, and label data corresponding to the features of each target fusion sample, determine the loss function of the initial supplementary lighting prediction model; the image contrast is used to characterize the brightness difference corresponding to the features of the target fusion sample, and the spatial frequency is used to characterize the edge effect of the location of the face object; S3. Based on the loss function, adjust the model parameters of the initial supplementary lighting prediction model; S4. Repeat steps S1 to S4 until the initial supplementary lighting prediction model converges, and obtain the trained supplementary lighting prediction model.
7. The adaptive adjustment method for supplementary lighting according to any one of claims 1-6, characterized in that, The target fill light combination result includes the intensity prediction weight corresponding to each target fill light, and the intensity prediction weight is used to characterize the ratio of the intensity corresponding to each target fill light after adjustment to the maximum intensity; The adaptive adjustment of the intensity of each target supplementary light in the monitoring device based on the target supplementary light combination result includes: Based on the light color temperature type corresponding to each of the target fill lights, or the light color temperature type corresponding to each of the target fill lights, and the distance between the target face in the face monitoring image and the monitoring device, the adjustment coefficient corresponding to each of the target fill lights is determined; Based on the adjustment coefficients, the intensity prediction weights corresponding to the target fill lights are corrected to obtain the target intensity weights corresponding to the target fill lights. The intensity of each target fill light is adaptively adjusted based on the target intensity weight corresponding to each target fill light.
8. A supplementary lighting adaptive adjustment device, characterized in that, include: The acquisition module is used to acquire facial surveillance images collected by the monitoring equipment; The input module is used to input the face monitoring image into the supplementary lighting prediction model to obtain the target supplementary lighting combination result corresponding to the monitoring device. The supplementary lighting prediction model is trained based on calibration sample images and label data. The calibration sample images are collected when the face object appears in different supplementary lighting areas corresponding to different groups of supplementary lighting variable parameters. Each supplementary lighting area is determined based on the radiation angle corresponding to each group of supplementary lighting in the monitoring device. The label data is used to characterize the supplementary lighting variable parameters corresponding to the calibration sample images. The target supplementary lighting combination result is obtained by feature fusion of the region images of the supplementary lighting areas corresponding to each group of supplementary lighting in the face monitoring image by the supplementary lighting prediction model. The correction module is used to adaptively adjust the intensity of each target fill light in the monitoring device based on the result of the target fill light combination.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the adaptive adjustment method for fill light as described in any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the adaptive adjustment method for fill light as described in any one of claims 1 to 7.
