Indoor photoelectric intelligent control method and system
By using camera monitoring and the U-net network segmentation model, combined with frequency division coding modulation to optimize lamp brightness, the problem of fine-grained and adaptive indoor photoelectric control in existing technologies has been solved, achieving high spatial resolution illuminance distribution and adaptive light modulation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-15
- Publication Date
- 2026-03-27
AI Technical Summary
Existing indoor photoelectric control technology is difficult to achieve precise and adaptive illuminance control of key work surfaces. In particular, it cannot ensure the illuminance and uniformity of key work surfaces when natural light changes, people move or furniture is adjusted, and it cannot identify lamp obstruction and provide targeted supplemental lighting.
By monitoring indoor images with cameras, using the U-net network segmentation model and illuminance analysis network, the illuminance image of the work surface is obtained, the average illuminance and uniformity are calculated, frequency division coding modulation is performed, the brightness modulation of the lamps is optimized, and adaptive lighting modulation control is achieved.
Without adding physical sensors, high spatial resolution illuminance distribution and adaptive lighting modulation were achieved on key working surfaces, ensuring that illuminance and uniformity meet standards and reducing redundant energy consumption.
Smart Images
Figure CN121751446A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent control technology, specifically to an indoor photoelectric intelligent control method and system. Background Technology
[0002] Existing indoor photoelectric control technologies mainly employ timed / scene-mode control, closed-loop dimming based on a few illuminance sensors, or overall brightness adjustment based on cameras. However, these methods have the following shortcomings: First, the control targets are mostly the overall illuminance of the room or the illuminance at the sensor location, making it difficult to perform timely and accurate fine closed-loop control on specific key work surfaces (such as desktops, workbenches, display surfaces, etc.). When natural light changes, people move, or furniture is rearranged, the illuminance and uniformity of key work surfaces often cannot be reliably guaranteed. Second, when people or objects partially obstruct key work surfaces, existing systems can usually only increase the overall brightness or simply turn on more lights. They cannot identify which lights are blocked, nor can they intelligently select lights that can still effectively illuminate key work surfaces for targeted supplemental lighting without increasing glare and energy consumption. Summary of the Invention
[0003] This invention obtains high spatial resolution illuminance distribution on key work surfaces such as blackboards and office desktops without adding a large number of physical illuminance sensors, achieving a stable mapping from image pixel space to physical illuminance space. Then, based on the target average illuminance and illuminance uniformity, it performs on-demand illuminance compliance judgment on different types of work surfaces. When the compliance is not met, it obtains the weight distribution of the influence of each lamp on the illuminance of the work surface by frequency division coding modulation. Then, under the constraints of illuminance and modulation change, it optimizes and solves the brightness modulation amplitude of each lamp and sends it to the lamp control chip, realizing refined and low-redundancy lighting modulation control that adapts to changes in natural light, personnel activities, and indoor layout adjustments.
[0004] This invention provides an indoor photoelectric intelligent control method, comprising: The indoor environment is continuously monitored by cameras to obtain indoor images arranged in time. The indoor image is sampled at preset intervals to obtain the target indoor image. The target indoor image is then sent to the work surface segmentation model for processing, and the target work surface area image corresponding to the work surface label selected by the user is output. The target working surface area image and camera imaging parameter data are fed into the working surface illuminance analysis network for processing to obtain the target working surface illuminance image, which includes illuminance values at different locations on the working surface. Calculate the target average illuminance and illuminance uniformity corresponding to the target working surface illuminance image. If the target average illuminance and illuminance uniformity corresponding to the target working surface illuminance image are both higher than the corresponding illuminance threshold and uniformity threshold, wait for the next indoor image sampling; otherwise, execute the working surface lighting modulation strategy. The working face lighting modulation strategy includes the following: Frequency division coding modulation is performed on all lamps, that is, different interferences are applied to all lamps, and the indoor environment after the interference is applied is monitored for N consecutive preset time periods to obtain the illuminance influence weight distribution map corresponding to each lamp. Based on the illuminance influence weight distribution diagram corresponding to all lamps, the brightness of all lamps is modulated. Under the constraints of illuminance and modulation change, the optimal lamp modulation scheme is obtained. The lamp modulation scheme includes the modulation amplitude for each lamp. All luminaires are modulated based on the lighting modulation scheme.
[0005] As a preferred aspect, all luminaires undergo frequency division coding modulation, i.e., different interferences are applied to all luminaires, and the indoor environment after the interference is applied is monitored for N consecutive preset time periods to obtain the illuminance influence weight distribution map corresponding to each luminaire. The specific steps include the following: Iterate through all the lamps, and for the selected lamp, add a sinusoidal modulation constrained by human eye perception to the current brightness of the lamp, and record the corresponding modulation frequency. All lamps perform illumination according to the added sinusoidal modulation. The indoor area is monitored continuously for N preset time periods. All indoor images are combined into a brightness analysis set. All pixels of the target working surface area image are mapped to the indoor image and recorded as target working surface pixels. The target working surface pixels are traversed. The gray values corresponding to the target working surface pixels are selected from the brightness analysis set and arranged in time order to form the time brightness signal corresponding to the target working surface pixels. Short-time Fourier transform is performed on the time brightness signal to obtain the time brightness spectrum corresponding to the target working surface pixels. Iterate through all luminaires. For the selected luminaire, obtain the temporal luminance spectrum corresponding to all target working surface pixels. Based on the modulation frequency corresponding to the luminaire, obtain the response amplitude corresponding to the target working surface pixels from the temporal luminance spectrum corresponding to all target working surface pixels. Combine the response amplitudes corresponding to all target working surface pixels into an illuminance influence weight distribution map. The scale of the illuminance influence weight distribution map is consistent with the target working surface region image.
[0006] As a preferred approach, based on the illuminance influence weight distribution diagram corresponding to all lamps, the brightness of all lamps is modulated. Under illuminance constraints and modulation change constraints, the optimal lamp modulation scheme is obtained, specifically including the following steps: Construct several simulated lighting modulation schemes, each including a randomly set modulation amplitude for each lamp. Combine all simulated lighting modulation schemes into a population set and set a maximum number of iterations. Calculate the fitness of the simulated lighting modulation scheme: Obtain the illuminance difference map. The scale of the illuminance difference map is the same as that of the target working surface illuminance image. The data corresponding to each position in the illuminance difference map is the ratio of the difference between the data at the corresponding position in the illuminance standard map and the data at the corresponding position in the target working surface illuminance image to the data at the corresponding position in the illuminance standard map. All data in the illuminance standard map are illuminance thresholds. Multiply all illuminance influence weight distribution maps with the modulation amplitudes corresponding to the lamps in the simulated lighting modulation scheme and then add the outputs to obtain the simulated modulation map. Calculate the similarity between the simulated modulation map and the illuminance difference map, and record it as the modulation coincidence value. Record the sum of all modulation amplitudes in the simulated lighting modulation scheme as the modulation coincidence value. Record the sum of the modulation coincidence value and its reciprocal as the fitness of the simulated lighting modulation scheme. Based on the fitness of the simulated lighting modulation scheme, the population set is iteratively updated using the sparrow search algorithm until the maximum number of iterations is reached. The simulated lighting modulation scheme with the highest fitness is then output as the optimal lighting modulation scheme.
[0007] As a preferred aspect, the target working surface area image and camera imaging parameter data are fed into a working surface illuminance analysis network for processing to obtain a target working surface illuminance image. Specifically, this includes the following steps: expanding each imaging parameter in the camera imaging parameter data into a parameter matrix, the parameter matrix having the same scale as the target working surface area image, and all data in the parameter matrix being the corresponding imaging parameters; then stitching all parameter matrices together with the target working surface area image according to channels and feeding them into the working surface illuminance analysis network for processing to obtain the target working surface illuminance image.
[0008] As a preferred aspect, training the work surface segmentation model specifically includes the following: acquiring several indoor images, annotating the indoor images using segmentation annotation information, which specifically includes work surface notes and work surface location information, forming a work surface segmentation training set from all the annotated indoor images, and training the work surface segmentation model using the work surface segmentation training set.
[0009] As a preferred aspect, training the working surface illuminance analysis network specifically includes the following: acquiring several working surface illuminance analysis training samples, which include images of the target working surface area and camera imaging parameter data; labeling the working surface illuminance analysis training samples using the target working surface illuminance images; forming a working surface illuminance analysis training set from all labeled working surface illuminance analysis training samples; and training the working surface illuminance analysis network using the working surface illuminance analysis training set.
[0010] As a preferred aspect, both the working surface segmentation model and the surface illuminance analysis network adopt the U-net network.
[0011] The present invention also provides an indoor photoelectric intelligent control system, comprising: The indoor image acquisition module is used to continuously monitor the indoor environment through a camera and acquire indoor images arranged in time. The work surface segmentation module is used to sample indoor images at preset time intervals to obtain target indoor images. The target indoor images are then sent to the work surface segmentation model for processing, and the output is the target work surface area image corresponding to the work surface label selected by the user. The working surface illumination analysis module is used to send the target working surface area image and camera imaging parameter data into the working surface illumination analysis network for processing to obtain the target working surface illumination image, which includes the illumination values at different locations on the working surface. The working surface illumination analysis module is used to calculate the target average illuminance and illuminance uniformity corresponding to the target working surface illuminance image. If the target average illuminance and illuminance uniformity corresponding to the target working surface illuminance image are both higher than the corresponding illuminance threshold and uniformity threshold, wait for the next indoor image sampling; otherwise, execute the working surface lighting modulation strategy. The working face lighting modulation strategy construction module is used to perform frequency division coding modulation on all lamps, that is, to apply different interferences to all lamps and monitor the indoor environment for N consecutive preset time periods after the interference is applied to obtain the illuminance influence weight distribution map corresponding to each lamp; based on the illuminance influence weight distribution map corresponding to all lamps, the brightness of all lights is modulated, and under the constraints of illuminance and modulation change amount, the optimal lighting modulation scheme is obtained, which includes the modulation amplitude for each lamp; and the lighting modulation scheme is used to modulate all lamps.
[0012] The present invention has the following advantages: This invention obtains high spatial resolution illuminance distribution on key work surfaces such as blackboards and office desktops without adding a large number of physical illuminance sensors, achieving a stable mapping from image pixel space to physical illuminance space. Then, based on the target average illuminance and illuminance uniformity, it performs on-demand illuminance compliance judgment on different types of work surfaces. When the compliance is not met, it obtains the weight distribution of the influence of each lamp on the illuminance of the work surface by frequency division coding modulation. Then, under the constraints of illuminance and modulation change, it optimizes and solves the brightness modulation amplitude of each lamp and sends it to the lamp control chip, realizing refined and low-redundancy lighting modulation control that adapts to changes in natural light, personnel activities, and indoor layout adjustments. Attached Figure Description
[0013] Figure 1 This is a schematic diagram of the indoor photoelectric intelligent control system used in an embodiment of the present invention. Detailed Implementation
[0014] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.
[0015] Example 1: An indoor photoelectric intelligent control method, comprising: Cameras are installed indoors to analyze the indoor lighting distribution. The cameras need to be positioned to observe most of the indoor area, focusing primarily on areas used by people, such as blackboards in classrooms and desks in offices. The cameras continuously monitor the indoor environment, acquiring images arranged chronologically. These images include areas corresponding to key work surfaces, which are the aforementioned areas used by people. Designers set different work surface labels, such as blackboards and office desks, which users can choose themselves. The indoor images are sampled at preset time intervals to obtain the target indoor image. This target indoor image is obtained after noise reduction and grayscale processing. The preset time is generally set to 1 second. Sampling usually involves randomly selecting one frame from all indoor images captured within the preset time. The target indoor image is then fed into the work surface segmentation model for processing, and the output is the target work surface region image corresponding to the work surface label selected by the user. It should be noted that the target work surface region image is in the form of a matrix, which is the smallest matrix that includes the target work surface region. Pixels in the target work surface region image that are not in the target work surface region are set to 0. The work surface segmentation model uses the U-net network. The target working surface area image and camera imaging parameter data are fed into the working surface illuminance analysis network for processing to obtain the target working surface illuminance image. The camera imaging parameter data here includes data such as the exposure time, ISO and resolution set in the camera. The target working surface illuminance image includes illuminance values at different locations on the working surface. The working surface illuminance analysis network can realize the mapping from the pixel values of the target working surface area image to the illuminance values of the target working surface illuminance image. Further analysis of the illuminance values of the target working surface illuminance image can determine the lighting conditions on the working surface, and then execute subsequent photoelectric control to realize illuminance modulation on the working surface. Calculate the target average illuminance and illuminance uniformity corresponding to the target work surface illuminance image. The target average illuminance is the average of all illuminance values in the target work surface illuminance image, and the illuminance uniformity is the ratio of the lowest illuminance value in the target work surface illuminance image to the average of all illuminance values. Determine whether the target average illuminance and illuminance uniformity corresponding to the target work surface illuminance image are both higher than the corresponding illuminance threshold and uniformity threshold. The illuminance threshold is generally set to 500 lx, but the corresponding illuminance threshold varies depending on the work surface selected by the user. The uniformity threshold is generally set to 0.7. If the target average illuminance and illuminance uniformity corresponding to the target work surface illuminance image are both higher than the corresponding illuminance threshold and uniformity threshold, wait for the next indoor image sampling; otherwise, execute the work surface lighting modulation strategy. The working face lighting modulation strategy includes the following: Frequency division coding modulation is performed on all lamps, that is, different interferences are applied to all lamps, and the indoor environment after interference is monitored for N consecutive preset time periods, where N is usually 2. The illuminance influence weight distribution map corresponding to each lamp is obtained. It should be noted that when the indoor work surface is affected by changes in natural light, personnel adjustments, or adjustments in the placement of indoor equipment, the illuminance of the work surface will also be affected, and the illumination of the work surface by the lamps will also be affected to varying degrees. The illuminance influence weight distribution map can be used to analyze the different effects of the lamps on the illumination of the work surface, and then the subsequent lamp modulation scheme can be executed adaptively. Based on the illuminance influence weight distribution diagram for all lamps, the brightness of all lamps is modulated. Under illuminance constraints and modulation change constraints, the optimal lamp modulation scheme is obtained. The lamp modulation scheme includes the modulation amplitude for each lamp. The modulation amplitude represents the brightness increase coefficient for the corresponding lamp. For example, if the modulation amplitude of a lamp is 5%, it means that the brightness of the lamp needs to be increased by 5%. This adjustment amplitude will be sent to the corresponding lamp, and the control chip of the corresponding lamp will adjust the corresponding PWM control command according to this modulation amplitude (if the lamp is controlled by PWM command). All luminaires are modulated based on the lighting modulation scheme.
[0016] This application obtains high spatial resolution illuminance distribution on key work surfaces such as blackboards and office desktops without adding a large number of physical illuminance sensors, achieving a stable mapping from image pixel space to physical illuminance space. Then, based on the target average illuminance and illuminance uniformity, it performs on-demand illuminance compliance judgment on different types of work surfaces. When the compliance is not met, it obtains the weight distribution of the influence of each lamp on the illuminance of the work surface by frequency division coding modulation of each lamp. Then, under the constraints of illuminance and modulation change, it optimizes and solves the brightness modulation amplitude of each lamp and sends it to the lamp control chip, realizing refined and low-redundancy lighting modulation control that adapts to changes in natural light, personnel activities, and indoor layout adjustments.
[0017] All luminaires undergo frequency division coding modulation, which involves applying different interferences to all luminaires and continuously monitoring the indoor environment for N preset time periods after the interference is applied to obtain the illuminance influence weight distribution map for each luminaire. The specific steps include the following: The process iterates through all the lamps. For each selected lamp, a sinusoidal modulation constrained by human perception is added to its current brightness, and the corresponding modulation frequency is recorded. In practice, the control chip of the corresponding lamp will adjust the control command according to the modulated brightness to achieve the change in brightness. The human perception constraint ensures that the interference modulation of the lamp is not perceived by the human eye, thus avoiding the impact of light changes on the user. The human perception constraint is set by the operator. The modulation frequency is generally set between 15-30Hz, and the modulation amplitude is set between 0.5-1%. All lamps perform illumination according to the added sinusoidal modulation. The indoor area is monitored continuously for N preset time periods. All indoor images are combined into a brightness analysis set. All pixels of the target working surface area image are mapped to the indoor image and recorded as target working surface pixels. The target working surface pixels are traversed. The gray values corresponding to the target working surface pixels are selected from the brightness analysis set and arranged in time order to form the time brightness signal corresponding to the target working surface pixels. Short-time Fourier transform is performed on the time brightness signal to obtain the time brightness spectrum corresponding to the target working surface pixels. Traverse all luminaires, and for the selected luminaire, obtain the temporal luminance spectrum corresponding to all target working surface pixels. Based on the modulation frequency corresponding to the luminaire, obtain the response amplitude corresponding to the target working surface pixels from the temporal luminance spectrum corresponding to all target working surface pixels. Combine the response amplitudes corresponding to all target working surface pixels into an illuminance influence weight distribution map. The scale of the illuminance influence weight distribution map is consistent with the target working surface region image. Under the premise of fixed exposure parameters and small dimming disturbance amplitude, the total illuminance at any point on a certain working surface can be approximately represented as the linear superposition of the illuminance contributions of each lamp. When a unique time modulation frequency (such as a sine wave or square wave) is applied to any lamp, which is different from other lamps, the illuminance signal at that point that changes with time can be written as the sum of several different frequency components. According to the orthogonality of complex exponential basis functions at different frequencies in Fourier analysis, the amplitude of the spectrum of the time-domain signal at each encoding frequency is only proportional to the modulation component of the corresponding lamp, and does not interfere with the different frequency components of other lamps. Therefore, by analyzing the time-series imaging data collected by the camera in the frequency domain and extracting the response amplitude at each encoding frequency point, the incremental contribution of each lamp to the illuminance of the working surface under the current scene and posture can be estimated.
[0018] Based on the illuminance influence weight distribution diagram for all lamps, the brightness of all lamps is modulated. Under illuminance constraints and modulation change constraints, the optimal lamp modulation scheme is obtained, which includes the following steps: Construct several simulated lighting modulation schemes, each including a randomly set modulation amplitude for each lamp. Combine all simulated lighting modulation schemes into a population set and set a maximum number of iterations. Calculate the fitness of the simulated lighting modulation scheme: Obtain the illuminance difference map. The scale of the illuminance difference map is the same as that of the target working surface illuminance image. The data corresponding to each position in the illuminance difference map is the ratio of the difference between the data at the corresponding position in the illuminance standard map and the data at the corresponding position in the target working surface illuminance image to the data at the corresponding position in the illuminance standard map. All data in the illuminance standard map are illuminance thresholds. Multiply all illuminance influence weight distribution maps with the modulation amplitudes corresponding to the lamps in the simulated lighting modulation scheme and then add the outputs to obtain the simulated modulation map. Calculate the similarity between the simulated modulation map and the illuminance difference map, and record it as the modulation coincidence value. Record the sum of all modulation amplitudes in the simulated lighting modulation scheme as the modulation coincidence value. Record the sum of the modulation coincidence value and its reciprocal as the fitness of the simulated lighting modulation scheme. In actual operation, if the modulation change is not a concern, the weighted sum of the modulation coincidence value and its reciprocal can also be used as the fitness, with the weight of the reciprocal of the modulation coincidence value set lower. Based on the fitness of the simulated lighting modulation scheme, the population set is iteratively updated using the sparrow search algorithm until the maximum number of iterations is reached. The simulated lighting modulation scheme with the highest fitness is then output as the optimal lighting modulation scheme.
[0019] The target working surface area image and camera imaging parameter data are fed into the working surface illumination analysis network for processing to obtain the target working surface illumination image. The specific steps include the following: Each imaging parameter in the camera imaging parameter data is expanded into a parameter matrix. The parameter matrix has the same scale as the target working surface area image, and all data in the parameter matrix are the corresponding imaging parameters. Then, all parameter matrices are stitched together with the target working surface area image according to channels and sent to the working surface illumination analysis network for processing to obtain the target working surface illumination image. It should be noted that the working surface illumination analysis network adopts the U-net network. In order to achieve the uniformity of the U-net network, this application expands the camera imaging parameter data into multi-channel data and integrates it into the target working surface area image.
[0020] Training the working face segmentation model includes the following: Several indoor images are acquired, which are collected by operators during the design phase of the actual scene. The indoor images are labeled with segmentation annotation information, which is the actual work surface in the indoor images labeled by the operators. Specifically, it includes work surface labels and work surface location information. All labeled indoor images are combined into a work surface segmentation training set. The work surface segmentation model is trained on the work surface segmentation training set. During training, the predicted label output of the work surface segmentation model and the labeled work surface labels are used as the type loss, and the predicted location output of the work surface segmentation model and the labeled work surface location information are used as the location loss. The specific loss value can be calculated using MSE. Based on these two losses, the parameters are backpropagated using the gradient descent method, and the training is terminated when the accuracy of the work surface segmentation model reaches the expected level.
[0021] Training a network for analyzing illuminance on the work surface includes the following: Several training samples for work surface illuminance analysis were obtained. These training samples included images of the target work surface area and camera imaging parameter data. These training samples were acquired by the operator based on actual shooting operations. The training samples were labeled using the target work surface illuminance images. These labeled target work surface illuminance images were constructed by the operator based on actual data collection by the lux meter and interpolation. All labeled training samples were combined into a work surface illuminance analysis training set. The work surface illuminance analysis network was trained using this training set. During training, the loss value was constructed using the predicted output of the work surface illuminance analysis network and the labeled target work surface illuminance images. Backpropagation of the parameters was performed using the gradient descent method. The training was terminated when the accuracy of the work surface illuminance analysis network reached the expected level.
[0022] Example 2: An indoor photoelectric intelligent control system, such as Figure 1 As shown, it includes: The indoor image acquisition module is used to continuously monitor the indoor environment through cameras and acquire indoor images arranged in time. The indoor images include areas corresponding to key work surfaces, which are the areas used by indoor personnel mentioned above. Different work surface labels are set by designers, such as blackboard and office desk, which can be selected by users themselves. The work surface segmentation module samples indoor images at preset time intervals to obtain target indoor images. These target indoor images are obtained after noise reduction and grayscale processing. The preset time is typically set to 1 second. Sampling usually involves randomly selecting one frame from all indoor images captured within the preset time. The target indoor image is then fed into the work surface segmentation model for processing, outputting a target work surface region image corresponding to the user-selected work surface label. It should be noted that the target work surface region image is in matrix form, representing the smallest matrix that includes the target work surface region. Pixels in the target work surface region image that are not part of the target work surface region are set to 0. The work surface segmentation model uses the U-net network. The working surface illuminance analysis module is used to send the target working surface area image and camera imaging parameter data into the working surface illuminance analysis network for processing to obtain the target working surface illuminance image. The camera imaging parameter data here includes data such as the exposure time, ISO and resolution set in the camera. The target working surface illuminance image includes illuminance values at different locations on the working surface. The working surface illuminance analysis network can realize the mapping from the pixel values of the target working surface area image to the illuminance values of the target working surface illuminance image. Further analysis of the illuminance values of the target working surface illuminance image can determine the lighting conditions on the working surface, and then execute subsequent photoelectric control to realize illuminance modulation on the working surface. The work surface illumination analysis module is used to calculate the target average illuminance and illuminance uniformity corresponding to the target work surface illuminance image. The target average illuminance is the average of all illuminance values in the target work surface illuminance image, and the illuminance uniformity is the ratio of the lowest illuminance value in the target work surface illuminance image to the average of all illuminance values. It determines whether the target average illuminance and illuminance uniformity corresponding to the target work surface illuminance image are both higher than the corresponding illuminance threshold and uniformity threshold. The illuminance threshold is generally set to 500 lx, and the corresponding illuminance threshold is different depending on the work surface selected by the user. The uniformity threshold is generally set to 0.7. If the target average illuminance and illuminance uniformity corresponding to the target work surface illuminance image are both higher than the corresponding illuminance threshold and uniformity threshold, it waits for the next indoor image sampling; otherwise, it executes the work surface lighting modulation strategy. The work surface lighting modulation strategy construction module is used to perform frequency division coding modulation on all lamps, that is, to apply different interferences to all lamps and monitor the indoor environment after the interference is applied for N consecutive preset time periods, where N is generally 2. The module obtains the illuminance influence weight distribution map corresponding to each lamp. It should be noted that when the indoor work surface is affected by changes in natural light, personnel adjustments, or adjustments in the placement of indoor equipment, the illuminance of the work surface will also be affected, and the illumination of the work surface by the lamps will also be affected to varying degrees. The illuminance influence weight distribution map can be used to analyze the different effects of the lamps on the illumination of the work surface, and then adaptively execute the subsequent lamp modulation scheme. Based on the illuminance influence weight distribution diagram for all lamps, the brightness of all lamps is modulated. Under illuminance constraints and modulation change constraints, the optimal lighting modulation scheme is obtained. The lighting modulation scheme includes the modulation amplitude for each lamp. The modulation amplitude represents the brightness increase coefficient for the corresponding lamp. For example, if the modulation amplitude of a lamp is 5%, it means that the brightness of the lamp needs to be increased by 5%. This adjustment amplitude will be sent to the corresponding lamp, and the control chip of the corresponding lamp will adjust the corresponding PWM control command according to this modulation amplitude (if the lamp is controlled by PWM command). All lamps are modulated based on the lighting modulation scheme.
[0023] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.
Claims
1. An indoor photoelectric intelligent control method, characterized in that, include: The indoor environment is continuously monitored by cameras to obtain indoor images arranged in time. The indoor image is sampled at preset intervals to obtain the target indoor image. The target indoor image is then sent to the work surface segmentation model for processing, and the target work surface area image corresponding to the work surface label selected by the user is output. The target working surface area image and camera imaging parameter data are fed into the working surface illuminance analysis network for processing to obtain the target working surface illuminance image, which includes illuminance values at different locations on the working surface. Calculate the target average illuminance and illuminance uniformity corresponding to the target working surface illuminance image. If the target average illuminance and illuminance uniformity corresponding to the target working surface illuminance image are both higher than the corresponding illuminance threshold and uniformity threshold, wait for the next indoor image sampling; otherwise, execute the working surface lighting modulation strategy. The working face lighting modulation strategy includes the following: Frequency division coding modulation is performed on all lamps, that is, different interferences are applied to all lamps, and the indoor environment after the interference is applied is monitored for N consecutive preset time periods to obtain the illuminance influence weight distribution map corresponding to each lamp. Based on the illuminance influence weight distribution diagram corresponding to all lamps, the brightness of all lamps is modulated. Under the constraints of illuminance and modulation change, the optimal lamp modulation scheme is obtained. The lamp modulation scheme includes the modulation amplitude for each lamp. All luminaires are modulated based on the lighting modulation scheme.
2. The indoor photoelectric intelligent control method according to claim 1, characterized in that, All luminaires undergo frequency division coding modulation, which involves applying different interferences to all luminaires and continuously monitoring the indoor environment for N preset time periods after the interference is applied to obtain the illuminance influence weight distribution map for each luminaire. The specific steps include the following: Iterate through all the lamps, and for the selected lamp, add a sinusoidal modulation constrained by human eye perception to the current brightness of the lamp, and record the corresponding modulation frequency. All lamps perform illumination according to the added sinusoidal modulation. The indoor area is monitored continuously for N preset time periods. All indoor images are combined into a brightness analysis set. All pixels of the target working surface area image are mapped to the indoor image and recorded as target working surface pixels. The target working surface pixels are traversed. The gray values corresponding to the target working surface pixels are selected from the brightness analysis set and arranged in time order to form the time brightness signal corresponding to the target working surface pixels. Short-time Fourier transform is performed on the time brightness signal to obtain the time brightness spectrum corresponding to the target working surface pixels. Iterate through all luminaires. For the selected luminaire, obtain the temporal luminance spectrum corresponding to all target working surface pixels. Based on the modulation frequency corresponding to the luminaire, obtain the response amplitude corresponding to the target working surface pixels from the temporal luminance spectrum corresponding to all target working surface pixels. Combine the response amplitudes corresponding to all target working surface pixels into an illuminance influence weight distribution map. The scale of the illuminance influence weight distribution map is consistent with the target working surface region image.
3. The indoor photoelectric intelligent control method according to claim 2, characterized in that, Based on the illuminance influence weight distribution diagram for all lamps, the brightness of all lamps is modulated. Under illuminance constraints and modulation change constraints, the optimal lamp modulation scheme is obtained, which includes the following steps: Construct several simulated lighting modulation schemes, each including a randomly set modulation amplitude for each lamp. Combine all simulated lighting modulation schemes into a population set and set a maximum number of iterations. Calculate the fitness of the simulated lighting modulation scheme: Obtain the illuminance difference map. The scale of the illuminance difference map is the same as that of the target working surface illuminance image. The data corresponding to each position in the illuminance difference map is the ratio of the difference between the data at the corresponding position in the illuminance standard map and the data at the corresponding position in the target working surface illuminance image to the data at the corresponding position in the illuminance standard map. All data in the illuminance standard map are illuminance thresholds. Multiply all illuminance influence weight distribution maps with the modulation amplitudes corresponding to the lamps in the simulated lighting modulation scheme and then add the outputs to obtain the simulated modulation map. Calculate the similarity between the simulated modulation map and the illuminance difference map, and record it as the modulation coincidence value. Record the sum of all modulation amplitudes in the simulated lighting modulation scheme as the modulation coincidence value. Record the sum of the modulation coincidence value and its reciprocal as the fitness of the simulated lighting modulation scheme. Based on the fitness of the simulated lighting modulation scheme, the population set is iteratively updated using the sparrow search algorithm until the maximum number of iterations is reached. The simulated lighting modulation scheme with the highest fitness is then output as the optimal lighting modulation scheme.
4. The indoor photoelectric intelligent control method according to claim 3, characterized in that, The target working surface area image and camera imaging parameter data are fed into the working surface illumination analysis network for processing to obtain the target working surface illumination image. The specific steps include: expanding each imaging parameter in the camera imaging parameter data into a parameter matrix. The parameter matrix has the same scale as the target working surface area image, and all data in the parameter matrix are the corresponding imaging parameters. Then, all parameter matrices are stitched together with the target working surface area image according to channels and fed into the working surface illumination analysis network for processing to obtain the target working surface illumination image.
5. The indoor photoelectric intelligent control method according to claim 4, characterized in that, Training the work surface segmentation model specifically includes the following: acquiring several indoor images, annotating the indoor images using segmentation annotation information, which specifically includes work surface notes and work surface location information, forming a work surface segmentation training set from all the annotated indoor images, and training the work surface segmentation model using the work surface segmentation training set.
6. The indoor photoelectric intelligent control method according to claim 5, characterized in that, The training of the working surface illuminance analysis network specifically includes the following: obtaining several working surface illuminance analysis training samples, which include images of the target working surface area and camera imaging parameter data; labeling the working surface illuminance analysis training samples using the target working surface illuminance images; forming a working surface illuminance analysis training set by combining all labeled working surface illuminance analysis training samples; and training the working surface illuminance analysis network using the working surface illuminance analysis training set.
7. The indoor photoelectric intelligent control method according to claim 6, characterized in that, The working surface segmentation model and the surface illuminance analysis network both use the U-net network.
8. An indoor photoelectric intelligent control system, characterized in that, The system employs an indoor photoelectric intelligent control method according to any one of claims 1-7, comprising: The indoor image acquisition module is used to continuously monitor the indoor environment through a camera and acquire indoor images arranged in time. The work surface segmentation module is used to sample indoor images at preset time intervals to obtain target indoor images. The target indoor images are then sent to the work surface segmentation model for processing, and the output is the target work surface area image corresponding to the work surface label selected by the user. The working surface illumination analysis module is used to send the target working surface area image and camera imaging parameter data into the working surface illumination analysis network for processing to obtain the target working surface illumination image, which includes the illumination values at different locations on the working surface. The working surface illumination analysis module is used to calculate the target average illuminance and illuminance uniformity corresponding to the target working surface illuminance image. If the target average illuminance and illuminance uniformity corresponding to the target working surface illuminance image are both higher than the corresponding illuminance threshold and uniformity threshold, wait for the next indoor image sampling; otherwise, execute the working surface lighting modulation strategy. The working face lighting modulation strategy construction module is used to perform frequency division coding modulation on all lamps, that is, to apply different interferences to all lamps and monitor the indoor environment for N consecutive preset time periods after the interference is applied to obtain the illuminance influence weight distribution map corresponding to each lamp; based on the illuminance influence weight distribution map corresponding to all lamps, the brightness of all lights is modulated, and under the constraints of illuminance and modulation change amount, the optimal lighting modulation scheme is obtained, which includes the modulation amplitude for each lamp; and the lighting modulation scheme is used to modulate all lamps.