An office lighting adjustment method based on artificial intelligence and related devices

Abstract

The present application provides an office lighting adjustment method and related devices based on artificial intelligence. On the one hand, the office worker's concentration and / or fatigue level can be comprehensively identified based on the office worker's target facial image, the office worker's target position in the office, and the ambient light data of the target position, thereby improving the accuracy of personnel status identification. On the other hand, the present application can comprehensively consider multiple factors such as the office worker's concentration and / or fatigue level, the target position, and the ambient light data at the target position, and automatically adjust the office's lighting equipment, thereby achieving accurate and dynamic lighting adjustment, thereby improving the intelligence level and adjustment accuracy of lighting adjustment.

Description

Technical Field

[0001] The present application relates to the field of lighting control technology, and in particular to an office lighting adjustment method and related devices based on artificial intelligence. Background Art

[0002] Office lighting solutions refer to the use of lighting equipment to create a well-lit and suitable office environment in the office. It is not only related to office safety, but also to the visual comfort and work efficiency of office workers.

[0003] At present, the existing office lighting adjustment solutions mainly include manual adjustment and timed adjustment solutions. Among them, the manual adjustment solution refers to the solution in which office personnel manually operate the lighting switch or dimmer to adjust the office lighting equipment, which is less convenient and intelligent. The timed adjustment solution can adjust the office lighting equipment according to the preset timing program. Compared with manual adjustment, timed adjustment has improved convenience and intelligence, but this method cannot adapt to changes in ambient light. To solve this problem, the prior art also proposes a solution that uses a light sensor to detect the ambient light intensity in the office and automatically adjusts the office lighting equipment according to the ambient light intensity. However, the inventors have found that this solution is still insufficient in terms of adjustment accuracy.

[0004] It can be seen that the existing technology has problems of low intelligence and low adjustment accuracy. Summary of the invention

[0005] The purpose of this application is to solve at least one of the above-mentioned technical defects, especially the technical defects of low intelligence and low adjustment accuracy in the prior art.

[0006] In a first aspect, an embodiment of the present application provides an office lighting adjustment method based on artificial intelligence, comprising:

[0007] Acquire target office images of office personnel;

[0008] Determine the target position of the office worker in the office, and obtain target ambient light data of the target position;

[0009] Using a predetermined target image processing algorithm to perform image segmentation on the target office image and obtain a target face image;

[0010] Inputting the target face image, the target ambient light data and the target position into a pre-trained fatigue concentration recognition model, and obtaining a target recognition result; wherein the target recognition result is used to reflect the concentration level and / or fatigue level of the office worker;

[0011] The lighting equipment of the office is adjusted according to the target recognition result, the target ambient light data and the target position.

[0012] In one embodiment, the inputting the target face image, the target ambient light data and the target position into a pre-trained fatigue concentration recognition model and obtaining a target recognition result includes:

[0013] Inputting the target face image, the target ambient light data and the target position into the feature extraction network of the fatigue concentration recognition model, and obtaining target expression features, target ambient light features and target position features respectively;

[0014] The target expression feature, the target ambient light feature and the target position feature are fused through the feature fusion network of the fatigue concentration recognition model to obtain a target fusion feature;

[0015] Based on the target fusion feature, the target recognition result is obtained through the recognition network of the fatigue concentration recognition model, and the target recognition result is output.

[0016] In one embodiment, adjusting the lighting equipment of the office according to the target recognition result, the target ambient light data and the target position includes:

[0017] Inputting the target recognition result, the target ambient light data and the target position into a pre-built decision generation model, and obtaining a lighting adjustment decision;

[0018] Based on the lighting adjustment decision, the lighting equipment of the office is adjusted.

[0019] In one embodiment, the step of determining the target image processing algorithm includes:

[0020] Acquire a plurality of first office images of the office, and respectively determine an image acquisition time and an image acquisition position corresponding to each of the first office images;

[0021] Acquire a plurality of first ambient light data of the office, and respectively determine a light collection time and a light collection position corresponding to each of the first ambient light data;

[0022] Establishing a correspondence between each of the first office images and each of the first ambient light data according to each of the image acquisition moments, each of the image acquisition positions, each of the light acquisition moments, and each of the light acquisition positions;

[0023] Based on the corresponding relationship, the algorithm parameters of the initial image processing algorithm are optimized, and after the optimization is completed, the target image processing algorithm is obtained.

[0024] In one embodiment, the optimizing the algorithm parameters of the initial image processing algorithm based on the corresponding relationship includes:

[0025] Based on the corresponding relationship and the pre-constructed light space distribution model, the algorithm parameters of the initial image processing algorithm are optimized; wherein the light space distribution model is a spatial distribution model of light in the office.

[0026] In one embodiment, the target image processing algorithm includes a target image segmentation algorithm and a target preprocessing algorithm; the target preprocessing algorithm includes at least one of the following:

[0027] Target image denoising algorithm;

[0028] Target image enhancement algorithm;

[0029] Target image cropping algorithm;

[0030] Target image scaling algorithm;

[0031] Object edge detection algorithm;

[0032] Target texture analysis algorithm;

[0033] Target shape feature extraction algorithm.

[0034] In one embodiment, the training step of the fatigue concentration recognition model includes:

[0035] Acquire multiple sets of pre-constructed training data; wherein each set of the training data includes a training face image, second ambient light data, a training position, and pre-labeled fatigue and concentration label data; the training face image is a face image obtained based on a second office image, and the second office image is obtained by collecting images of office workers in the office;

[0036] The initial model is trained using multiple sets of the training data until a preset training end condition is met and the fatigue-concentration recognition model is obtained.

[0037] In a second aspect, an embodiment of the present application provides an office lighting adjustment device based on artificial intelligence, comprising:

[0038] A target office image acquisition module is used to acquire a target office image of an office worker;

[0039] A target position and ambient light acquisition module, used to determine the target position of the office worker in the office and to acquire target ambient light data of the target position;

[0040] An image segmentation module, used to perform image segmentation on the target office image using a predetermined target image processing algorithm, and obtain a target face image;

[0041] A fatigue and concentration recognition module, used to input the target face image, the target ambient light data and the target position into a pre-trained fatigue and concentration recognition model, and obtain a target recognition result; wherein the target recognition result is used to reflect the concentration level and / or fatigue level of the office worker;

[0042] An adjustment module is used to adjust the lighting equipment of the office according to the target recognition result, the target ambient light data and the target position.

[0043] In a third aspect, an embodiment of the present application provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the artificial intelligence-based office lighting adjustment method described in any of the above embodiments.

[0044] In a fourth aspect, an embodiment of the present application provides a lighting control device, the lighting control device comprising: one or more processors, and a memory;

[0045] The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the artificial intelligence-based office lighting adjustment method described in any of the above embodiments are executed.

[0046] In the office lighting adjustment method and related devices based on artificial intelligence provided in some embodiments of the present application, on the one hand, the office worker's concentration and / or fatigue level can be comprehensively identified based on the office worker's target face image, the office worker's target position in the office, and the ambient light data at the target position, thereby improving the accuracy of personnel status identification. On the other hand, the present application can comprehensively consider multiple factors such as the office worker's concentration and / or fatigue level, the target position, and the ambient light data at the target position, and automatically adjust the office's lighting equipment, thereby achieving accurate and dynamic lighting adjustment, thereby improving the intelligence level and adjustment accuracy of lighting adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] 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 or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0048] Figure 1 FIG. 1 is an application environment diagram of an office lighting adjustment method based on artificial intelligence in one embodiment;

[0049] Figure 2 A schematic diagram of a flow chart of an office lighting adjustment method based on artificial intelligence in one embodiment;

[0050] Figure 3 is a schematic diagram of the structure of an office lighting adjustment device based on artificial intelligence in one embodiment;

[0051] Figure 4 FIG. 4 is a diagram of the internal structure of a lighting control device in one embodiment. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0053] In one embodiment, the office lighting adjustment method based on artificial intelligence provided by the present application can be applied to Figure 1 In the application environment shown in the figure, the lighting control device 102 is connected to the image acquisition device 104 and the ambient light detection device 106 respectively. It can be understood that the "connection" described in the present application can be an electrical connection or a communication connection; it can be a wired connection or a wireless connection. This article does not make any specific restrictions on this.

[0054] Specifically, the image acquisition device 104 is installed in the office to collect facial images of office workers in the office. For example, the image acquisition device 104 may be a high-definition surveillance camera. The ambient light detection device 106 may be installed at multiple locations in the office to collect ambient light data at different locations in the office. For example, the ambient light data may include ambient light intensity data and ambient light color temperature data. The lighting control device 102 is used to process and analyze the images collected by the image acquisition device 104 and the ambient light data collected by the ambient light detection device 106, and adjust and control the lighting equipment in the office according to the results of the processing and analysis, so as to provide a more suitable lighting environment for office workers.

[0055] It should be noted that the image acquisition, storage, use, processing, transmission, provision and disclosure involved in this application are in compliance with the relevant laws and regulations and do not violate public order and good morals.

[0056] In one embodiment, the present application provides an office lighting adjustment method based on artificial intelligence. The following embodiments apply this method to Figure 1 The lighting control device shown in FIG. Figure 2 As shown, the method of the present application specifically includes the following steps:

[0057] S202: Acquire a target office image of an office worker.

[0058] The office staff refers to the staff working in the office. The target office image refers to the image obtained by capturing the face of the office staff, including the facial information of the office staff.

[0059] In this step, the lighting control device may capture images of office workers in the office through the image capture device, so as to subsequently identify the concentration and / or fatigue level of the office workers through the target office images.

[0060] S204: Determine the target position of the office worker in the office, and obtain target ambient light data of the target position.

[0061] The target position refers to the position of the office worker in the office when the target office image is collected. The target ambient light data refers to the ambient light data at the target position. For example, the target ambient light data may include ambient light intensity and ambient light color temperature.

[0062] In this step, the lighting control device can determine the position of the office worker in the office and obtain the target position. For example, the lighting control device can determine the target position according to the setting position of the target image device, where the target image device refers to an image acquisition device used to acquire the target office image. For another example, the lighting control device can perform face detection and recognition on the target office image and obtain the personnel information corresponding to the office worker, and determine the target position according to the personnel information and the preset personnel seating table.

[0063] The lighting control device may also detect the ambient light data at the target position through the ambient light detection device, and obtain the target ambient light data accordingly. In order to improve the accuracy of the target ambient light data, in one example, the lighting control device may perform data calibration, data compensation and / or data filtering on the initial ambient light data detected by the ambient light detection device, and obtain the target ambient light data after processing. Among them, the data calibration process can be used to eliminate the deviation and error of the light sensor. The data compensation process can consider the impact of environmental factors such as temperature and humidity on the ambient light measurement, and perform corresponding compensation. Data filtering processing refers to removing high-frequency noise and low-frequency drift in the initial ambient light data through filters such as low-pass filtering, band-pass filtering and / or high-pass filtering.

[0064] It can be understood that in order to improve the intelligence and accuracy of lighting adjustment, the time interval between the collection time corresponding to the target ambient light data and the collection time corresponding to the target office image should be within the preset threshold to ensure that the target ambient light data can accurately reflect the ambient light conditions of the office personnel's current location.

[0065] S206: Using a predetermined target image processing algorithm to perform image segmentation on the target office image, and obtain a target face image.

[0066] The target face image refers to the face image of the office worker, including the facial information of the office worker. Since the target office image includes the facial area of ​​the office worker, the lighting control device can use the target image processing algorithm to perform image processing on the target office image, so as to segment the facial area image corresponding to the office worker from the target office image and obtain the target face image.

[0067] It can be understood that the target image processing algorithm may include one or more image processing algorithms, and the specific algorithm type may be determined according to actual conditions. In one example, the target image processing algorithm may include a target image segmentation algorithm and a target preprocessing algorithm, and the target preprocessing algorithm includes at least one of the following: a target image denoising algorithm, a target image enhancement algorithm, a target image cropping algorithm, a target image scaling algorithm, a target edge detection algorithm, a target texture analysis algorithm, and a target shape feature extraction algorithm.

[0068] Among them, the target image denoising algorithm can use image filtering algorithms such as Gaussian filtering and median filtering to remove noise in the image to improve image quality. The target image enhancement algorithm can enhance the visual effect of the image by adjusting operations such as image contrast, brightness and / or color balance to highlight important features. The target image cropping algorithm can crop specific areas in the image according to actual needs to adjust the image size and adapt to subsequent processing. The target image scaling algorithm can scale the image according to actual needs to adjust the image size and adapt to subsequent processing. The target edge detection algorithm can use edge detection algorithms such as Sobel algorithm and Canny operator to extract edge information in the image. The target texture analysis algorithm can use texture feature extraction algorithms such as gray level co-occurrence matrix and local binary pattern to extract the texture characteristics of the image. The target shape feature extraction algorithm can use morphological operations such as corrosion, expansion, opening and closing operations to extract the shape features of objects in the image.

[0069] In the above example, the lighting control device may firstly use an image preprocessing algorithm to preprocess the target office image and obtain a preprocessed image, and then use a target image segmentation algorithm to segment the preprocessed image to obtain a target face image.

[0070] Furthermore, in one example, the target image segmentation algorithm of the present application may be a threshold-based image segmentation algorithm, an edge-based image segmentation algorithm and / or a region-based image segmentation algorithm. The threshold-based image segmentation algorithm can select a suitable threshold to segment the image into different regions according to the grayscale value distribution of the image. The edge-based image segmentation algorithm can connect continuous edges to form closed areas based on edge detection results to achieve image segmentation. The region-based image segmentation algorithm can combine pixels with similar features into regions through methods such as region growing and watershed algorithms.

[0071] S208: Input the target face image, target ambient light data and target position into the pre-trained fatigue concentration recognition model to obtain a target recognition result.

[0072] The fatigue concentration recognition model refers to an artificial intelligence model based on a deep learning algorithm, which is used to detect the concentration and / or fatigue of office workers. The target recognition result is used to reflect the concentration and / or fatigue of office workers.

[0073] In this step, the lighting control device can input the target facial image, target ambient light data and target position into the fatigue and concentration recognition model, so that the fatigue and concentration recognition model can comprehensively consider the facial expressions of the office workers, the positions of the office workers and the ambient light at the positions of the office workers, accurately identify the fatigue level and / or concentration level of the office workers, and obtain the target recognition results.

[0074] S210: Adjusting office lighting equipment according to the target recognition result, the target ambient light data, and the target position.

[0075] In this step, since an office is usually equipped with multiple lighting devices, and the lighting areas corresponding to each lighting device are different, and considering the different ambient light conditions at different locations, this application can adjust the office lighting equipment by considering the office workers' fatigue and concentration level, the office workers' positions, and the light data of the office workers' positions, and making a comprehensive decision based on this.

[0076] Among them, the target position information can be used to determine the target lighting equipment corresponding to the office personnel. By adjusting the target lighting equipment, the light can be more concentrated in the area where the office personnel are located, thereby further improving the lighting effect.

[0077] In one example, the lighting control device may calculate the lighting brightness adjustment value based on the target recognition result and the target ambient light data according to the preset adjustment rules and weight coefficients, and adjust the lighting equipment located near the office personnel to the corresponding brightness value according to the lighting brightness adjustment value, the target position and the preset device position mapping relationship. For example, when the target recognition result includes the degree of fatigue and the degree of concentration, and the target ambient light data includes the ambient light intensity and the ambient light color temperature, the weight coefficients of each parameter and the related descriptions may be shown in Table 1.

[0078]

[0079] In this case, when the fatigue level is 0.8, the concentration level is 0.6, the ambient light intensity is 0.4, and the ambient light color temperature is 0.5, the calculated lighting brightness adjustment value is 0.61. The lighting control device can adjust the lighting device at the corresponding position accordingly.

[0080] In this application, on the one hand, the office worker's concentration and / or fatigue level can be comprehensively identified based on the office worker's target face image, the office worker's target position in the office, and the ambient light data at the target position, thereby improving the accuracy of personnel status identification. On the other hand, this application can automatically adjust the office's lighting equipment by comprehensively considering multiple factors such as the office worker's concentration and / or fatigue level, the target position, and the ambient light data at the target position, thereby achieving accurate and dynamic lighting adjustment, thereby improving the intelligence and adjustment accuracy of lighting adjustment.

[0081] In one embodiment, the target face image, the target ambient light data and the target position are input into a pre-trained fatigue concentration recognition model, and a target recognition result is obtained, including:

[0082] Step A1: input the target face image, target ambient light data and target position into the feature extraction network of the fatigue concentration recognition model, and obtain the target expression feature, target ambient light feature and target position feature respectively;

[0083] Step A3: Through the feature fusion network of the fatigue concentration recognition model, the target expression feature, the target ambient light feature and the target position feature are fused to obtain the target fusion feature;

[0084] Step A5: Based on the target fusion features, the target recognition result is obtained through the recognition network of the fatigue concentration recognition model, and the target recognition result is output.

[0085] In this embodiment, the fatigue-concentration recognition model can fuse multi-source features and identify the fatigue level and / or concentration level of the office worker based on the fused features, so that the fatigue-concentration recognition model can more accurately determine the fatigue level and / or concentration level of the office worker.

[0086] Specifically, the fatigue and concentration recognition model may include a feature extraction network, a feature fusion network, and a recognition network. After the lighting control device inputs the target face image, the target ambient light data, and the target position into the fatigue and concentration recognition model, the fatigue and concentration recognition model may use the feature extraction network to extract the target expression features from the target face image, the target ambient light features from the target ambient light data, and the target position features from the target position.

[0087] After obtaining the target expression features, target ambient light features and target position features, the fatigue concentration recognition model can fuse the target expression features, target ambient light features and target position features through its feature fusion network and obtain the target fusion features. For example, feature concatenation, weighted summation and other methods can be used to achieve feature fusion.

[0088] The target fusion features can provide richer information. The recognition network of the fatigue recognition model can identify the fatigue concentration level of office workers based on the target fusion features and obtain the target recognition results.

[0089] In one embodiment, adjusting the lighting equipment of an office according to the target recognition result, the target ambient light data and the target position includes:

[0090] Input the target recognition result, target ambient light data and target position into a pre-built decision generation model and obtain the lighting adjustment decision;

[0091] Based on the lighting adjustment decision, adjust the lighting equipment in the office.

[0092] In this embodiment, the lighting control device can use a pre-built decision generation model to generate a lighting adjustment decision based on the target recognition result, the target ambient light data and the target position, and adjust the office lighting equipment accordingly to further improve the adjustment accuracy.

[0093] The decision generation model can be an artificial intelligence model based on a deep learning algorithm, or a set of preset rules. After the decision generation model is constructed, the decision generation model can be experimented and optimized according to actual conditions to further improve the performance of the decision generation model and the rationality and accuracy of the lighting adjustment decision.

[0094] In one example, the lighting adjustment decision may include a device identifier corresponding to each lighting device to be adjusted and a lighting adjustment value corresponding to each lighting device to be adjusted.

[0095] In one embodiment, the step of determining the target image processing algorithm includes:

[0096] Step B1: acquiring a plurality of first office images of an office, and determining the image acquisition time and image acquisition position corresponding to each first office image respectively;

[0097] Step B3: Acquire multiple first ambient light data of the office, and respectively determine the light collection time and light collection position corresponding to each first ambient light data;

[0098] Step B5: establishing a correspondence between each first office image and each first ambient light data according to each image acquisition time, each image acquisition position, each light acquisition time and each light acquisition position;

[0099] Step B7: Based on the corresponding relationship, the algorithm parameters of the initial image processing algorithm are optimized, and after the optimization is completed, the target image processing algorithm is obtained.

[0100] In this embodiment, multiple first office images of the office and the first ambient light data corresponding to each first office image can be obtained in advance, and the initial image processing algorithm can be optimized based on this, so that the optimized target image processing algorithm can more accurately identify and segment the face area, so as to further improve the accuracy of subsequent fatigue and concentration recognition, and then improve the accuracy of lighting adjustment.

[0101] Specifically, the lighting control device can obtain multiple first office images of the office, as well as the image acquisition time and image acquisition position corresponding to each first office image. The image acquisition time refers to the time when the first office image is acquired, and the image acquisition position refers to the office position corresponding to the first office image. Further, the relevant description of the first office image can refer to the description of the target office image above, which will not be repeated in this article.

[0102] The lighting control device can obtain multiple first ambient light data, as well as the light collection time and light collection position corresponding to each first ambient light data. The light collection time refers to the time when the first ambient light data is collected, and the light collection position refers to the office position corresponding to the first ambient light data. Further, the relevant description of the first ambient light data can refer to the relevant description of the above-mentioned target ambient light data, which will not be repeated in this article.

[0103] After obtaining each first office image and each first ambient light data, the lighting control device may associate each first office image with each first ambient light data, and establish a corresponding relationship between the first office image and the first ambient light data. Considering that the ambient light conditions at different office locations at the same time are different, and the ambient light conditions at the same office location at different times are also different, therefore, in one example, the lighting control device may associate the first ambient light data with the first office image that are collected at similar times and at adjacent locations.

[0104] After establishing the corresponding relationship, the lighting control device can optimize the algorithm parameters of the initial image processing algorithm based on the corresponding relationship to analyze the impact of light and time on image quality and feature recognition (such as face recognition), and obtain the target image processing algorithm after optimization. For example, the lighting control device can analyze the impact of light intensity on image brightness and image contrast, the impact of light color temperature on image color balance, and the impact of light distribution on target shadow and three-dimensional sense, etc., based on the corresponding relationship, and optimize the image processing algorithm accordingly.

[0105] In one example, the lighting control device may establish a correspondence table between ambient light data and image processing algorithm parameters, and automatically adjust the image processing algorithm parameters according to the target ambient light data in subsequent processing.

[0106] In another example, when the target image processing algorithm includes a target image segmentation algorithm and a target image preprocessing algorithm, the lighting control device may optimize the algorithm parameters of the initial image segmentation algorithm and the algorithm parameters of the initial image preprocessing algorithm respectively based on the corresponding relationship, and obtain the target image segmentation algorithm and the target image preprocessing algorithm after optimization.

[0107] In one embodiment, based on the correspondence, the algorithm parameters of the initial image processing algorithm are optimized, including: based on the correspondence and a pre-constructed light space distribution model, the algorithm parameters of the initial image processing algorithm are optimized; wherein the light space distribution model is a spatial distribution model of light in an office.

[0108] In this embodiment, considering that the light conditions at different locations in the office are different, the lighting control device can pre-build a light space distribution model to understand the distribution of light in different directions and positions through the light space distribution model. By optimizing the algorithm parameters according to the corresponding relationship and the light space distribution model, the impact of light on image quality can be better analyzed, and the image processing quality of the target image processing algorithm can be improved to further improve the accuracy of lighting adjustment.

[0109] In one embodiment, after acquiring multiple first office images, the lighting control device may calculate various statistical index values ​​of the first office images, and compare and analyze the features between different images based on the various statistical index values ​​to determine the commonalities and individualities between different images, thereby better understanding the features and changes of the images. For example, by comparing the image features of different people under different lighting conditions, the influence of light on the status of people can be found, providing a basis for the adjustment strategy of the lighting device.

[0110] Among them, various statistical indicators may include mean, variance and histogram, etc. Among them, the mean can reflect the overall brightness level of the image. The variance can measure the contrast of the image. The histogram can show the distribution of different grayscale values ​​in the image. The present application can understand the characteristics of the image through the various statistical indicator values ​​of the first office image, and provide a reference for subsequent image processing and analysis. For example, in the subsequent process, the brightness and contrast of the image can be adjusted according to the mean and variance of the target office graphic to make the image clearer and easier to identify.

[0111] In one embodiment, after acquiring a plurality of first ambient light data, the average intensity, peak intensity and valley intensity of the light may be calculated according to each first ambient light data, and the variation trend of the light intensity, such as periodicity and monotonicity, may be analyzed.

[0112] In one embodiment, the training step of the fatigue concentration recognition model includes:

[0113] Acquire multiple sets of pre-constructed training data; each set of training data includes training face images, second ambient light data, training positions, and pre-labeled fatigue and concentration label data; the training face images are face images obtained based on the second office images, and the second office images are obtained by collecting images of office workers in the office;

[0114] The initial model is trained using multiple sets of training data until the preset training end conditions are met and a fatigue concentration recognition model is obtained.

[0115] Specifically, before model training, multiple second office images can be collected in advance, and fatigue concentration labels corresponding to each second office image can be annotated. The expressions of people corresponding to every two second office images can be the same or different. In addition, the second ambient light data and training position information corresponding to each second office image can be collected in advance and associated with the second office image.

[0116] After obtaining multiple second office images, the lighting control device can perform operations such as cropping, scaling, random rotation, flipping, and normalization on the second office images to obtain multiple training face images. In addition, the lighting control device can also encode and convert fatigue and focus labels to obtain fatigue and focus label data that is easy for the model to understand and process. The lighting control device can generate multiple sets of training data based on the association relationship, training face images, fatigue and focus label data, second ambient light data, and training positions.

[0117] In the process of acquiring the fatigue concentration recognition model, the lighting control device can use multiple sets of training data to train the initial model, and use the back propagation algorithm to optimize the parameters of the initial model. In one example, the present application can use optimization algorithms such as stochastic gradient descent, Adagrad, Adadelta, etc. to adjust the parameters of the initial model, and can introduce regularization techniques such as L1 / L2 regularization and Dropout regularization to prevent overfitting. In another example, the initial model can be a convolutional neural network, such as ResNet network, VGG network, EfficientNet, MobileNet, etc.

[0118] In one example, during the model training process, a cross entropy loss function or a mean square error function may be used as a loss function.

[0119] In one example, the training process of the fatigue and concentration recognition model can introduce a generative adversarial network mechanism, an attention mechanism, and / or a model fusion mechanism. The generative adversarial network mechanism refers to adding an adversarial training link to the model training process, using the generator to try to generate realistic expression images and using the discriminator to distinguish between real images and generated images. The generative adversarial mechanism improves the robustness and generalization ability of the model.

[0120] The attention mechanism refers to the introduction of an attention mechanism in a convolutional neural network, which enables the model to automatically focus on key expression-related areas in facial images, such as eyes, mouth, eyebrows, etc., thereby improving the efficiency and accuracy of feature extraction.

[0121] The model fusion mechanism refers to training multiple convolutional neural network models with different structures or parameters, and fusing them through ensemble learning methods (such as averaging, weighted averaging, voting, etc.), combining the advantages of multiple models to improve the final expression recognition performance.

[0122] The following describes an artificial intelligence-based office lighting adjustment device provided in an embodiment of the present application. The artificial intelligence-based office lighting adjustment device described below and the artificial intelligence-based office lighting adjustment method described above can refer to each other.

[0123] In one embodiment, Figure 3As shown, the present application provides an office lighting adjustment device 300 based on artificial intelligence, comprising:

[0124] A target office image acquisition module 302 is used to acquire a target office image of an office worker;

[0125] The target position and ambient light acquisition module 304 is used to determine the target position of the office worker in the office and acquire the target ambient light data of the target position;

[0126] An image segmentation module 306 is used to segment the target office image using a predetermined target image processing algorithm and obtain a target face image;

[0127] The fatigue concentration recognition module 308 is used to input the target face image, the target ambient light data and the target position into a pre-trained fatigue concentration recognition model, and obtain a target recognition result; wherein the target recognition result is used to reflect the concentration level and / or fatigue level of the office worker;

[0128] The adjustment module 310 is used to adjust the lighting equipment of the office according to the target recognition result, the target ambient light data and the target position.

[0129] In one embodiment, the fatigue concentration identification module 308 of the present application includes:

[0130] A feature extraction unit, used to input the target face image, the target ambient light data and the target position into the feature extraction network of the fatigue concentration recognition model, and obtain target expression features, target ambient light features and target position features respectively;

[0131] A feature fusion unit, used to perform feature fusion on the target expression feature, the target ambient light feature and the target position feature through the feature fusion network of the fatigue concentration recognition model, and obtain a target fusion feature;

[0132] The recognition unit is used to obtain the target recognition result based on the target fusion feature through the recognition network of the fatigue concentration recognition model, and output the target recognition result.

[0133] In one embodiment, the adjustment module 310 of the present application includes:

[0134] A decision generation unit, configured to input the target recognition result, the target ambient light data and the target position into a pre-built decision generation model, and obtain a lighting adjustment decision;

[0135] The device adjustment unit is used to adjust the lighting device of the office based on the lighting adjustment decision.

[0136] In one embodiment, the artificial intelligence-based office lighting adjustment device 300 of the present application further includes:

[0137] A first office image acquisition module, used to acquire a plurality of first office images of the office, and respectively determine an image acquisition time and an image acquisition position corresponding to each of the first office images;

[0138] A first ambient light acquisition module, used for acquiring a plurality of first ambient light data of the office, and respectively determining a light collection time and a light collection position corresponding to each of the first ambient light data;

[0139] a correspondence generation module, configured to establish a correspondence between each of the first office images and each of the first ambient light data according to each of the image acquisition moments, each of the image acquisition positions, each of the light acquisition moments, and each of the light acquisition positions;

[0140] The optimization module is used to optimize the algorithm parameters of the initial image processing algorithm based on the corresponding relationship, and obtain the target image processing algorithm after the optimization is completed.

[0141] In one embodiment, the optimization module of the present application includes:

[0142] The optimization unit is used to optimize the algorithm parameters of the initial image processing algorithm based on the corresponding relationship and a pre-constructed light space distribution model; wherein the light space distribution model is a spatial distribution model of light in the office.

[0143] In one embodiment, the target image processing algorithm includes a target image segmentation algorithm and a target preprocessing algorithm; the target preprocessing algorithm includes at least one of the following:

[0144] Target image denoising algorithm;

[0145] Target image enhancement algorithm;

[0146] Target image cropping algorithm;

[0147] Target image scaling algorithm;

[0148] Object edge detection algorithm;

[0149] Target texture analysis algorithm;

[0150] Target shape feature extraction algorithm.

[0151] In one embodiment, the artificial intelligence-based office lighting adjustment device 300 of the present application further includes:

[0152] A training data acquisition module is used to acquire multiple sets of pre-constructed training data; wherein each set of the training data includes a training face image, second ambient light data, a training position, and pre-labeled fatigue and concentration label data; the training face image is a face image obtained based on a second office image, and the second office image is obtained by collecting images of office workers in the office;

[0153] The model training module is used to use multiple groups of training data to perform model training on the initial model until the preset training end conditions are met and the fatigue concentration recognition model is obtained.

[0154] In one embodiment, the present application also provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the artificial intelligence-based office lighting adjustment method in any embodiment.

[0155] In one embodiment, the present application also provides a lighting control device, in which computer-readable instructions are stored. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the artificial intelligence-based office lighting adjustment method in any embodiment.

[0156] Indicatively, Figure 4 This is a schematic diagram of the internal structure of a lighting control device provided in an embodiment of the present application. Figure 4 , the lighting control device 900 includes a processing component 902, which further includes one or more processors, and a memory resource represented by a memory 901, for storing instructions executable by the processing component 902, such as an application. The application stored in the memory 901 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 902 is configured to execute instructions to perform the steps of the office lighting adjustment method based on artificial intelligence described in any of the above embodiments.

[0157] The lighting control device 900 may further include a power supply component 903 configured to perform power management of the lighting control device 900, a wired or wireless network interface 904 configured to connect the lighting control device 900 to a network, and an input / output (I / O) interface 905. The lighting control device 900 may operate based on an operating system stored in the memory 901, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, Free BSD TM, or the like.

[0158] Those skilled in the art will understand that the internal structure of the lighting control device shown in the present application is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the lighting control device to which the scheme of the present application is applied. The specific lighting control device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0159] In one embodiment, the present application provides a lighting adjustment system, comprising:

[0160] An image acquisition device, used for acquiring target office images;

[0161] Ambient light detection equipment, used to collect target ambient light data;

[0162] The lighting control device described in any of the above embodiments is respectively connected to an image acquisition device and an ambient light detection device, and is used to execute the steps of the artificial intelligence-based office lighting adjustment method described in any of the above embodiments.

[0163] It can be understood that the "connection" described in the present application can be an electrical connection or a communication connection; it can be a wired connection or a wireless connection. This document does not impose specific restrictions on this. For example, the lighting control device can be communicatively connected to the image acquisition device and the ambient light detection device. For another example, the lighting control device can be electrically connected to the image acquisition device and the ambient light detection device. For another example, the lighting control device can be wired to the ambient light detection device and wirelessly connected to the image acquisition device.

[0164] In one example, the lighting control device may include a smart dimmer, and the lighting devices in the office are adjusted by the smart dimmer.

[0165] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not clearly listed, or also includes elements inherent to such process, method, article or equipment. In the absence of more restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or equipment including the elements. Herein, "one", "one", "said", "the" and "it" may also include plural forms, unless the context clearly indicates another way. A plurality refers to at least two cases, such as 2, 3, 5 or 8, etc. "And / or" includes any and all combinations of the relevant listed items.

[0166] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can refer to each other.

[0167] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An office lighting adjustment method based on artificial intelligence, characterized in that: include: Acquire target office images of office personnel; Determine the target position of the office worker in the office, and obtain target ambient light data of the target position; Using a predetermined target image processing algorithm to perform image segmentation on the target office image and obtain a target face image; Inputting the target face image, the target ambient light data and the target position into a pre-trained fatigue concentration recognition model, and obtaining a target recognition result; wherein the target recognition result is used to reflect the concentration level and / or fatigue level of the office worker; adjusting the lighting equipment of the office according to the target recognition result, the target ambient light data and the target position; The step of inputting the target face image, the target ambient light data and the target position into a pre-trained fatigue concentration recognition model and obtaining a target recognition result includes: Inputting the target face image, the target ambient light data and the target position into the feature extraction network of the fatigue concentration recognition model, and obtaining target expression features, target ambient light features and target position features respectively; The target expression feature, the target ambient light feature and the target position feature are fused through the feature fusion network of the fatigue concentration recognition model to obtain a target fusion feature; Based on the target fusion feature, the target recognition result is obtained through the recognition network of the fatigue concentration recognition model, and the target recognition result is output.

2. The method according to claim 1, characterized in that The step of adjusting the lighting equipment of the office according to the target recognition result, the target ambient light data and the target position includes: Inputting the target recognition result, the target ambient light data and the target position into a pre-built decision generation model, and obtaining a lighting adjustment decision; Based on the lighting adjustment decision, the lighting equipment of the office is adjusted.

3. The method according to any one of claims 1 to 2, characterized in that: The step of determining the target image processing algorithm comprises: Acquire a plurality of first office images of the office, and respectively determine an image acquisition time and an image acquisition position corresponding to each of the first office images; Acquire a plurality of first ambient light data of the office, and respectively determine a light collection time and a light collection position corresponding to each of the first ambient light data; Establishing a correspondence between each of the first office images and each of the first ambient light data according to each of the image acquisition moments, each of the image acquisition positions, each of the light acquisition moments, and each of the light acquisition positions; Based on the corresponding relationship, the algorithm parameters of the initial image processing algorithm are optimized, and after the optimization is completed, the target image processing algorithm is obtained.

4. The method according to claim 3, characterized in that The step of optimizing the algorithm parameters of the initial image processing algorithm based on the corresponding relationship includes: Based on the corresponding relationship and the pre-constructed light space distribution model, the algorithm parameters of the initial image processing algorithm are optimized; wherein the light space distribution model is a spatial distribution model of light in the office.

5. The method according to any one of claims 1 to 2, characterized in that: The target image processing algorithm includes a target image segmentation algorithm and a target preprocessing algorithm; the target preprocessing algorithm includes at least one of the following: Target image denoising algorithm; Target image enhancement algorithm; Target image cropping algorithm; Target image scaling algorithm; Object edge detection algorithm; Target texture analysis algorithm; Target shape feature extraction algorithm.

6. The method according to any one of claims 1 to 2, characterized in that: The training steps of the fatigue concentration recognition model include: Acquire multiple sets of pre-constructed training data; wherein each set of the training data includes a training face image, second ambient light data, a training position, and pre-labeled fatigue and concentration label data; the training face image is a face image obtained based on a second office image, and the second office image is obtained by collecting images of office workers in the office; The initial model is trained using multiple sets of the training data until a preset training end condition is met and the fatigue-concentration recognition model is obtained.

7. An office lighting adjustment device based on artificial intelligence, characterized in that: include: A target office image acquisition module is used to acquire a target office image of an office worker; A target position and ambient light acquisition module, used to determine the target position of the office worker in the office and to acquire target ambient light data of the target position; An image segmentation module, used to perform image segmentation on the target office image using a predetermined target image processing algorithm, and obtain a target face image; A fatigue and concentration recognition module, used to input the target face image, the target ambient light data and the target position into a pre-trained fatigue and concentration recognition model, and obtain a target recognition result; wherein the target recognition result is used to reflect the concentration level and / or fatigue level of the office worker; An adjustment module, configured to adjust the lighting equipment of the office according to the target recognition result, the target ambient light data and the target position; Wherein, the fatigue concentration recognition module includes: A feature extraction unit, used to input the target face image, the target ambient light data and the target position into the feature extraction network of the fatigue concentration recognition model, and obtain target expression features, target ambient light features and target position features respectively; A feature fusion unit, used to perform feature fusion on the target expression feature, the target ambient light feature and the target position feature through the feature fusion network of the fatigue concentration recognition model, and obtain a target fusion feature; The recognition unit is used to obtain the target recognition result based on the target fusion feature through the recognition network of the fatigue concentration recognition model, and output the target recognition result.

8. A storage medium, characterized in that: The storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the artificial intelligence-based office lighting adjustment method as described in any one of claims 1 to 6.

9. A lighting control device, characterized in that: include: one or more processors, and memory; The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the artificial intelligence-based office lighting adjustment method as described in any one of claims 1 to 6 are performed.

Images