Intelligent building automatic control system and method

Abstract

The invention relates to the field of intelligent control, and particularly discloses an intelligent building automatic control system and method, which adopts an artificial intelligence technology based on a deep neural network model to obtain brightness values in a building, brightness values outside the building and a monitoring video of the building at a plurality of preset time points in a preset time period, and combines time sequence change information to obtain a brightness value of the building. According to the method, data of different dimensions are fused, a time sequence feature extractor is adopted, feature changes under different time scales can be captured, and spatial information of building personnel monitoring images and brightness features are fused to generate a classification result used for representing that the illumination brightness of a building at the current time point should be increased, decreased or unchanged. According to the method, the lighting system can be guided to automatically adjust the lighting brightness, intelligent control over the lighting environment is achieved, and the energy utilization efficiency and the user experience are improved.

Description

Technical Field

[0001] This application relates to the field of intelligent control, and more specifically, to an intelligent building automation control system and method. Background Art

[0002] A smart building refers to a building that realizes the intelligent management and coordinated operation of various internal systems of the building by integrating technologies such as the Internet of Things (IoT), sensors, automation control systems, and data analysis.

[0003] Lighting is an important part of a building. Good lighting can provide a comfortable visual environment, enabling people to clearly see objects, text, and details. Appropriate lighting can reduce eye fatigue, glare, and visual discomfort, improving work efficiency and concentration. Good lighting is particularly important for places that require long-term concentration, such as offices, commercial spaces, and learning environments. A reasonable lighting design can help save energy and reduce environmental impact, can reduce energy consumption and extend the service life of lamps. Through reasonable lighting planning and control, unnecessary energy waste can be avoided, which has a positive impact on sustainable development and environmental protection.

[0004] However, since lighting control in traditional technologies usually adopts fixed brightness control, it lacks flexibility and personalization. This control method does not take into account the changes in different times and scenarios, resulting in energy waste and reduced efficiency. For example, when there is sufficient sunlight during the day, the lighting fixtures still operate at maximum brightness; when there is no one present, they still remain at high brightness, etc., all causing unnecessary energy consumption. Secondly, lighting control in traditional technologies usually can only control the brightness of the entire lighting system through manual switches or dimmers, lacking intelligence and self-adaptability, and unable to perform individual control and dynamic adjustment on different areas or places according to specific needs.

[0005] Therefore, an optimized intelligent building automation control solution is expected. Summary of the Invention

[0006] To solve the above technical problems, the present application is proposed. An embodiment of the present application provides an intelligent building automation control system and method, which uses artificial intelligence technology based on a deep neural network model to obtain the brightness values inside the building, the brightness values outside the building, and the surveillance videos of the building at multiple predetermined time points within a predetermined time period. Combining the time series change information, it fuses data in different dimensions and uses a time series feature extractor to capture feature changes at different time scales, and fuses the spatial information of the building personnel surveillance images with the brightness features to generate a classification result indicating whether the lighting brightness of the building at the current time point should be increased, decreased, or remain unchanged. This method can guide the lighting system to automatically adjust the lighting brightness, achieve intelligent control of the lighting environment, and improve energy utilization efficiency and user experience.

[0007] According to one aspect of the present application, there is provided an intelligent building automation control system, which includes: A building brightness monitoring and acquisition module, configured to obtain the brightness values inside the building, the brightness values outside the building, and the surveillance videos of the building at multiple predetermined time points within a predetermined time period; A building personnel detection module, configured to obtain a building personnel surveillance image by passing the surveillance video of the building through a personnel target detection network model; A building spatial feature acquisition module, configured to obtain a building personnel surveillance space enhanced feature matrix by passing the building personnel surveillance image through a personnel spatial attention mechanism based on a convolutional neural network model; A building brightness time series arrangement module, configured to arrange the brightness values inside the building and the brightness values outside the building at the multiple predetermined time points into an in-building brightness time series input vector and an out-building brightness time series input vector respectively according to the time dimension; A time series change calculation module, configured to calculate the difference between the in-building brightness at two adjacent time points in the in-building brightness time series input vector to obtain an in-building brightness time series change input vector, and calculate the difference between the out-building brightness at two adjacent time points in the out-building brightness time series input vector to obtain an out-building brightness time series change input vector; A building brightness concatenation module, configured to concatenate the in-building brightness time series input vector and the in-building brightness time series change input vector to obtain an in-building brightness multi-dimensional input vector, and concatenate the out-building brightness time series input vector and the out-building brightness time series change input vector to obtain an out-building brightness multi-dimensional input vector; A brightness multi-scale extraction module, configured to pass the in-building brightness multi-dimensional input vector and the out-building brightness multi-dimensional input vector through a multi-scale time series feature extractor including a first convolutional layer and a second convolutional layer to obtain an in-building brightness multi-scale feature vector and an out-building brightness multi-scale feature vector; A brightness correlation feature extraction module, which is used to perform correlation encoding on the in-building brightness multi-scale feature vector and the out-building brightness multi-scale feature vector to obtain a building brightness correlation matrix, and fuse it with the building personnel monitoring space enhancement feature matrix to obtain a building personnel brightness feature matrix; A building personnel brightness optimization module, which is used to optimize the building personnel brightness feature matrix to obtain an optimized building personnel brightness feature matrix; A building brightness adjustment module, which is used to obtain a classification result by inputting the optimized building personnel brightness feature matrix into a classifier, and the classification result is used to indicate whether the lighting brightness of the building at the current time point should be increased, decreased, or remain unchanged.

[0008] In the above intelligent building automation control system, the building personnel detection module is used for: the personnel target detection network model is an anchor window-based target detection network, and the anchor window-based target detection network is Fast R-CNN, Faster R-CNN or RetinaNet.

[0009] In the above intelligent building automation control system, the building space feature acquisition module includes: a personnel convolutional encoding unit, which is used to perform deep convolutional encoding on the building personnel monitoring image using the convolutional encoding part of the personnel space attention mechanism based on the convolutional neural network model to obtain an initial convolutional feature map of building personnel; a personnel space attention unit, which is used to input the initial convolutional feature map of building personnel into the spatial attention part of the personnel space attention mechanism to obtain a building personnel space attention map; a personnel feature activation unit, which is used to pass the building personnel space attention map through the Softmax activation function to obtain a building personnel space attention feature map; a personnel feature multiplication unit, which is used to calculate the element-wise multiplication of the building personnel space attention feature map and the initial convolutional feature map of building personnel to obtain a building personnel monitoring space enhancement feature map; and a personnel feature pooling unit, which is used to perform pooling on the building personnel monitoring space enhancement feature map along the channel dimension to obtain the building personnel monitoring space enhancement feature matrix.

[0010] In the above intelligent building automation control system, the brightness multi-scale extraction module includes: a first-scale brightness feature extraction unit, configured to input the multi-dimensional input vector of the in-building brightness into the first convolutional layer of the multi-scale time series feature extractor to obtain a first-scale in-building brightness feature vector, wherein the first convolutional layer has a first one-dimensional convolutional kernel with a first length; a second-scale brightness feature extraction unit, configured to input the multi-dimensional input vector of the in-building brightness into the second convolutional layer of the multi-scale time series feature extractor to obtain a second-scale in-building brightness feature vector, wherein the second convolutional layer has a second one-dimensional convolutional kernel with a second length, and the first length is different from the second length; and a fused brightness feature unit, configured to cascade the first-scale in-building brightness feature vector and the second-scale in-building brightness feature vector by using the cascade layer of the multi-scale time series feature extractor to obtain the in-building brightness multi-scale feature vector.

[0011] According to another aspect of the present application, there is provided an intelligent building automation control method, which includes: Obtaining the brightness values inside the building, the brightness values outside the building, and the monitoring video of the building at multiple predetermined time points within a predetermined time period; Passing the monitoring video of the building through a personnel target detection network model to obtain a building personnel monitoring image; Passing the building personnel monitoring image through a personnel spatial attention mechanism based on a convolutional neural network model to obtain a building personnel monitoring spatial enhanced feature matrix; Arranging the brightness values inside the building and the brightness values outside the building at the multiple predetermined time points in the time dimension respectively to form an in-building brightness time series input vector and an out-building brightness time series input vector; Calculating the difference between the in-building brightness at two adjacent time points in the in-building brightness time series input vector to obtain an in-building brightness time series change input vector, and calculating the difference between the out-building brightness at two adjacent time points in the out-building brightness time series input vector to obtain an out-building brightness time series change input vector; Cascading the in-building brightness time series input vector and the in-building brightness time series change input vector to obtain an in-building brightness multi-dimensional input vector, and cascading the out-building brightness time series input vector and the out-building brightness time series change input vector to obtain an out-building brightness multi-dimensional input vector; Passing the in-building brightness multi-dimensional input vector and the out-building brightness multi-dimensional input vector through a multi-scale time series feature extractor including a first convolutional layer and a second convolutional layer to obtain an in-building brightness multi-scale feature vector and an out-building brightness multi-scale feature vector; Associatively encoding the in-building brightness multi-scale feature vector and the out-building brightness multi-scale feature vector to obtain a building brightness association matrix, and fusing the building brightness association matrix with the building personnel monitoring spatial enhanced feature matrix to obtain a building personnel brightness feature matrix; Optimize the building personnel brightness feature matrix to obtain an optimized building personnel brightness feature matrix; Pass the optimized building personnel brightness feature matrix through a classifier to obtain a classification result, and the classification result is used to indicate whether the lighting brightness of the building at the current time point should be increased, decreased, or remain unchanged.

[0012] Compared with the prior art, a smart building automation control system and method provided by the present application adopt artificial intelligence technology based on a deep neural network model, obtain the brightness values inside the building, the brightness values outside the building, and the monitoring videos of the building at multiple predetermined time points within a predetermined time period, combine the time series change information, fuse data in different dimensions and adopt a time series feature extractor, which can capture the feature changes at different time scales, and fuse the spatial information of the building personnel monitoring images with the brightness features to generate a classification result indicating whether the lighting brightness of the building at the current time point should be increased, decreased, or remain unchanged. This method can guide the lighting system to automatically adjust the lighting brightness, realize intelligent control of the lighting environment, and improve energy utilization efficiency and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] By describing the embodiments of the present application in more detail with reference to the accompanying drawings, the above and other objects, features, and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation to the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0014] Figure 1 It is a block diagram of a smart building automation control system according to an embodiment of the present application.

[0015] Figure 2 It is a schematic diagram of the architecture of a smart building automation control system according to an embodiment of the present application.

[0016] Figure 3 It is a block diagram of a building space feature acquisition module in a smart building automation control system according to an embodiment of the present application.

[0017] Figure 4 It is a flowchart of a smart building automation control method according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] Hereinafter, exemplary embodiments of the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0019] Exemplary System Figure 1 It is a block diagram of an intelligent building automation control system according to an embodiment of the present application. Figure 2 It is a schematic diagram of the architecture of an intelligent building automation control system according to an embodiment of the present application. As Figure 1 and Figure 2 shown, the intelligent building automation control system 100 according to an embodiment of the present application includes: a building brightness monitoring and acquisition module 110, configured to obtain the brightness values inside the building, the brightness values outside the building, and the monitoring video of the building at multiple predetermined time points within a predetermined time period; a building personnel detection module 120, configured to obtain a building personnel monitoring image by passing the monitoring video of the building through a personnel target detection network model; a building space feature acquisition module 130, configured to obtain a building personnel monitoring space enhanced feature matrix by passing the building personnel monitoring image through a personnel space attention mechanism based on a convolutional neural network model; a building brightness time series arrangement module 140, configured to arrange the brightness values inside the building and the brightness values outside the building at the multiple predetermined time points in the time dimension respectively into an in-building brightness time series input vector and an out-building brightness time series input vector; a time series change calculation module 150, configured to calculate the difference between the in-building brightness at two adjacent time points in the in-building brightness time series input vector to obtain an in-building brightness time series change input vector and calculate the difference between the out-building brightness at two adjacent time points in the out-building brightness time series input vector to obtain an out-building brightness time series change input vector; a building brightness concatenation module 160, configured to concatenate the in-building brightness time series input vector and the in-building brightness time series change input vector to obtain an in-building brightness multi-dimensional input vector, and concatenate the out-building brightness time series input vector and the out-building brightness time series change input vector to obtain an out-building brightness multi-dimensional input vector; a brightness multi-scale extraction module 170, configured to obtain an in-building brightness multi-scale feature vector and an out-building brightness multi-scale feature vector by passing the in-building brightness multi-dimensional input vector and the out-building brightness multi-dimensional input vector through a multi-scale time series feature extractor including a first convolutional layer and a second convolutional layer; a brightness correlation feature extraction module 180, configured to perform correlation encoding on the in-building brightness multi-scale feature vector and the out-building brightness multi-scale feature vector to obtain a building brightness correlation matrix and fuse it with the building personnel monitoring space enhanced feature matrix to obtain a building personnel brightness feature matrix; a building personnel brightness optimization module 190, configured to optimize the building personnel brightness feature matrix to obtain an optimized building personnel brightness feature matrix; and, a building brightness adjustment module 200, configured to obtain a classification result by passing the optimized building personnel brightness feature matrix through a classifier, and the classification result is used to indicate whether the lighting brightness of the building at the current time point should be increased, decreased, or remain unchanged.

[0020] Smart Building refers to a building that uses advanced information technology and automated control systems to integrate various devices, sensors and networks to improve the building's operational efficiency, energy efficiency, safety and user experience. Specifically, smart buildings monitor, control and optimize various devices and systems inside the building by connecting sensors, control devices and communication networks with building facilities and systems. These devices and systems can include lighting systems, air conditioning systems, power management systems, security monitoring systems, building automation systems, communication systems, etc. Smart buildings have a wide range of applications, including commercial office buildings, hotels, hospitals, schools, shopping centers and other building types. Through the application of smart building technology, the management efficiency of buildings can be improved, operating costs can be reduced, and the quality of indoor environments can be improved. At the same time, it also provides support for sustainable development and green buildings.

[0021] In the embodiment of the present application, the building brightness monitoring and acquisition module 110 is used to obtain the brightness value inside the building, the brightness value outside the building and the monitoring video of the building at multiple predetermined time points in a predetermined time period. Considering that the monitoring video can provide real-time internal conditions of the building, including information such as personnel activities and object movement. In the technical solution of the present application, by analyzing and processing the monitoring video, the location, number and activity status of the personnel targets in the building can be obtained. This information can be used to judge the usage and personnel density in the building, so as to better understand the lighting needs of the building. For example, when there are many people in the building, it may be necessary to increase the lighting brightness to provide better lighting effects and safety. The brightness values inside and outside the building can provide the lighting conditions of the surrounding environment of the building. The brightness difference between indoors and outdoors is very important for lighting control. By monitoring the brightness values indoors and outdoors in real time, the brightness level of the lighting system can be adjusted according to the changes in light intensity. For example, when there is sufficient sunlight during the day, the lighting brightness can be reduced to save energy; while when it is dark or cloudy, the lighting brightness may need to be increased to provide sufficient lighting effects.

[0022] In detail, appropriate brightness can provide a good visual environment, allowing people inside the building to clearly see the surrounding objects and activities. Appropriate lighting conditions can improve people's work efficiency and productivity. A well-lit environment can improve attention, concentration and work quality, and reduce the occurrence of errors and mistakes. Adequate lighting can reduce eye fatigue and discomfort, and improve the comfort of work and life. In addition, in a low-brightness environment, people are prone to accidents, such as tripping and collisions. Appropriate lighting can reduce the occurrence of accidents and provide escape guidance in emergency situations. Secondly, moderate control of the lighting brightness of the building can achieve effective use of energy. By adjusting the lighting brightness according to actual needs, energy consumption can be reduced, energy costs can be reduced, and energy conservation and environmental protection can be achieved.

[0023] In an embodiment of the present application, the building personnel detection module 120 is configured to obtain a building personnel monitoring image by passing the monitoring video of the building through a personnel target detection network model. Considering that personnel position, quantity and other information can be extracted from the monitoring video of the building through personnel target detection, the position and activity status of personnel in the building can be obtained in real time, which helps to understand the usage situation inside the building, including personnel density, personnel flow, etc. Based on this information, the lighting requirements of the building can be better judged, and lighting adjustment can be made according to the actual situation. That is, precise lighting control can be achieved through personnel target detection. According to the position and distribution of personnel, the brightness and coverage of lighting devices can be intelligently adjusted to provide the best lighting effect. For example, the lighting brightness can be increased in crowded areas and decreased in sparse areas, so as to achieve energy conservation and optimization of lighting effect. In addition, the brightness of the lighting system can be adjusted according to the personnel distribution in the building. When there is no personnel activity or the personnel density is low in the building, the lighting brightness can be reduced to save energy. This intelligent lighting control can effectively reduce unnecessary energy consumption, reduce energy costs, and contribute to environmental protection.

[0024] Specifically, in an embodiment of the present application, the building personnel detection module is configured to: the personnel target detection network model is an anchor window-based target detection network, and the anchor window-based target detection network is Fast R-CNN, Faster R-CNN or RetinaNet.

[0025] In an embodiment of the present application, the building space feature acquisition module 130 is configured to obtain a building personnel monitoring space enhanced feature matrix by passing the building personnel monitoring image through a personnel space attention mechanism based on a convolutional neural network model. Considering that in the building personnel monitoring image, personnel are usually the main objects we focus on. Through the space attention mechanism, the model can pay more attention to the personnel area, highlight the features of personnel, and ignore some unnecessary background information. This can improve the model's ability to identify and analyze personnel, reduce the influence of background interference, and improve the accuracy of personnel target detection and tracking. Secondly, the space attention mechanism can adjust the importance of features in different positions by learning weights. It should be understood that for the building personnel monitoring image, the features in different regions may have different importance. Through the space attention mechanism, the feature expression of important regions can be enhanced, and the feature expression of secondary regions can be weakened, so as to better capture the key features of personnel.

[0026] Figure 3 It is a block diagram of the building space feature acquisition module in the intelligent building automation control system according to an embodiment of the present application. Specifically, in an embodiment of the present application, as Figure 3As shown, the building space feature acquisition module 130 includes: a personnel convolutional encoding unit 131, configured to perform deep convolutional encoding on the building personnel monitoring image using the convolutional encoding part of the personnel space attention mechanism based on the convolutional neural network model to obtain an initial convolutional feature map of the building personnel; a personnel space attention unit 132, configured to input the initial convolutional feature map of the building personnel into the spatial attention part of the personnel space attention mechanism to obtain a building personnel space attention map; a personnel feature activation unit 133, configured to pass the building personnel space attention map through a Softmax activation function to obtain a building personnel space attention feature map; a personnel feature multiplication unit 134, configured to calculate the element-wise multiplication of the building personnel space attention feature map and the initial convolutional feature map of the building personnel to obtain an enhanced feature map of the building personnel monitoring space; and a personnel feature pooling unit 135, configured to perform pooling on the enhanced feature map of the building personnel monitoring space along the channel dimension to obtain an enhanced feature matrix of the building personnel monitoring space.

[0027] In the embodiment of the present application, the building brightness time series arrangement module 140 is configured to arrange the brightness values inside and outside the building at the plurality of predetermined time points into an in-building brightness time series input vector and an out-building brightness time series input vector respectively according to the time dimension. Considering that the brightness values inside and outside the building may vary at different time points. By arranging according to the time dimension, the time series information of the brightness values can be retained, enabling the model to capture the change trend and periodicity of the brightness. This is very important for analyzing the lighting requirements of the building, predicting future brightness changes, and formulating corresponding dimming strategies. It should be understood that arranging the brightness values inside and outside the building according to the time dimension can establish the time correlation between them. Through the temporal adjacency, we can observe the correlation and influence relationship between the brightness inside and outside the building. This helps us understand the connection between the lighting conditions inside and outside the building and provides more accurate inputs for subsequent analysis and modeling. In addition, arranging the brightness values inside and outside the building according to the time dimension enables us to apply time series analysis methods and models. That is, time series analysis can help us explore and predict the trends, seasonal variations, periodic variations, and other time-related features of the brightness values. By performing time series analysis on the arranged time series data, we can extract useful information and patterns, providing guidance for the lighting control and optimization of the building.

[0028] In the embodiment of the present application, the timing change calculation module 150 is configured to calculate the difference between the indoor brightness values at two adjacent time points in the indoor brightness timing input vector to obtain an indoor brightness timing change input vector, and calculate the difference between the outdoor brightness values at two adjacent time points in the outdoor brightness timing input vector to obtain an outdoor brightness timing change input vector. Considering that the brightness difference can highlight the change trend of brightness. Specifically, calculating the difference between brightness values can extract the characteristics of brightness changes, such as the fluctuation degree, change speed, and periodicity of brightness. By observing the amplitude and direction of brightness changes, we can understand the change rules and characteristics of indoor and outdoor brightness, and understand the increase or decrease of indoor and outdoor brightness, so as to judge the dimming effect and change trend of the lighting system.

[0029] In the embodiment of the present application, the building brightness concatenation module 160 is configured to concatenate the indoor brightness timing input vector and the indoor brightness timing change input vector to obtain an indoor brightness multi-dimensional input vector, and concatenate the outdoor brightness timing input vector and the outdoor brightness timing change input vector to obtain an outdoor brightness multi-dimensional input vector. Considering that in this technical solution, the timing vector contains the change of indoor and outdoor brightness values over time, while the timing change vector reflects the change trend and characteristics of brightness. By concatenating these two vectors, the timing change of brightness and other relevant information can be combined to provide more comprehensive input information. It should be understood that concatenating the timing vector and the timing change vector can capture the dynamic characteristics of brightness. The timing vector reflects the absolute value of brightness, while the timing change vector reflects the change trend of brightness. Concatenating these two vectors can consider both the absolute value and the change trend of brightness, so as to better capture the dynamic change characteristics of brightness and provide more accurate input for subsequent analysis and modeling. Secondly, concatenating the timing vector and the timing change vector can comprehensively consider information from multiple dimensions. The timing vector and the timing change vector can contain different characteristics and aspects, such as absolute value, change trend, fluctuation degree, etc. Concatenating these vectors can comprehensively consider these different dimensions, so as to provide more comprehensive and multi-angle information and provide more basis for the analysis and decision-making of building lighting.

[0030] In the embodiment of the present application, the brightness multi-scale extraction module 170 is configured to obtain an in-building brightness multi-scale feature vector and an out-building brightness multi-scale feature vector by passing the in-building brightness multi-dimensional input vector and the out-building brightness multi-dimensional input vector through a multi-scale time-series feature extractor including a first convolutional layer and a second convolutional layer. Considering that different time scales may contain different feature information. Shorter time scales can capture rapid brightness changes and instantaneous features, while longer time scales can capture the trends and periodic features of brightness. By using a multi-scale time-series feature extractor, features at different time scales can be considered simultaneously, thus more comprehensively describing the brightness change situation. It should be understood that different building scenarios and applications may have different brightness change patterns and features. By using a multi-scale time-series feature extractor, different scenario and application requirements can be adapted. For rapidly changing brightness, features at shorter time scales can better capture the details of the changes; for slowly changing or periodically changing brightness, features at longer time scales can better capture the trends and periodicity of the changes. This can make the model more flexible and adaptable, and applicable to different types of building lighting problems.

[0031] Specifically, in the embodiment of the present application, the brightness multi-scale extraction module includes: a first-scale brightness feature extraction unit configured to input the in-building brightness multi-dimensional input vector into the first convolutional layer of the multi-scale time-series feature extractor to obtain a first-scale in-building brightness feature vector, where the first convolutional layer has a first one-dimensional convolutional kernel with a first length; a second-scale brightness feature extraction unit configured to input the in-building brightness multi-dimensional input vector into the second convolutional layer of the multi-scale time-series feature extractor to obtain a second-scale in-building brightness feature vector, where the second convolutional layer has a second one-dimensional convolutional kernel with a second length, and the first length is different from the second length; and a fused brightness feature unit configured to cascade the first-scale in-building brightness feature vector and the second-scale in-building brightness feature vector using the cascade layer of the multi-scale time-series feature extractor to obtain the in-building brightness multi-scale feature vector.

[0032] In the embodiment of the present application, the brightness correlation feature extraction module 180 is configured to perform correlation encoding on the in-building brightness multi-scale feature vector and the out-building brightness multi-scale feature vector to obtain a building brightness correlation matrix and fuse it with the building personnel monitoring space enhancement feature matrix to obtain a building personnel brightness feature matrix. Considering that there is often a certain correlation between the brightness inside and outside the building. For example, the lighting intensity inside the building may be affected by the degree of sunlight outside the building. Therefore, by performing correlation encoding on the in-building brightness and the out-building brightness, this correlation can be captured, the relationship and mutual influence between them can be captured, and incorporated into the feature representation.

[0033] The lighting conditions and human activities in a building are often related. By fusing human characteristics and brightness characteristics, the information of both humans and brightness can be comprehensively considered, thus more comprehensively describing the state of the building. In this technical solution, fusing human characteristics and brightness characteristics can improve the model's ability to identify and analyze humans. Human characteristics can provide information about the number, location, and behavior of people, while brightness characteristics can provide information about lighting conditions. By fusing these characteristics, the model's ability to express human characteristics can be enhanced, and the accuracy and reliability of human identification and analysis can be improved. That is, human characteristics and brightness characteristics can reflect information about the usage, energy consumption, and safety of the building. By fusing these characteristics, comprehensive building characteristics can be obtained, providing more comprehensive indicators and references for building operation and management decisions.

[0034] Specifically, in the embodiment of the present application, the brightness correlation feature extraction module is used to: calculate the product between the transposed vector of the in-building brightness multi-scale feature vector and the out-building brightness multi-scale feature vector respectively to obtain the building brightness correlation matrix.

[0035] In the embodiment of the present application, the building human brightness optimization module 190 is used to optimize the building human brightness feature matrix to obtain an optimized building human brightness feature matrix.

[0036] Particularly, in the technical solution of the present application, the in-building brightness values and out-building brightness values at multiple predetermined time points are used as inputs. The intervals between these time points may be irregular. For example, they may be sampled hourly, minute by minute, or second by second. Due to the inconsistency of the sampling time intervals, there may be significant differences in the brightness values at adjacent time points, resulting in discontinuous features. At the same time, the brightness values inside and outside the building are affected by lighting conditions, and lighting is a dynamically changing factor. At different time points, the lighting intensity inside and outside the building may change significantly. For example, the lighting conditions during the day and at night are very different. This lighting change will also lead to discontinuous features. Due to the discontinuity, the brightness difference between adjacent time points may be large, resulting in a wide range of eigenvalue ranges in the brightness time series change input vector. This range change may cause the probability distribution of eigenvalues to lack constraints, that is, the distribution of eigenvalues may not have a clear trend or pattern. The unconstrained probability distribution of eigenvalues may cause certain difficulties in subsequent classification tasks. For example, when using a classifier to train and predict the building human brightness feature matrix, the classifier may have difficulty accurately learning and inferring the relationships between different eigenvalues. This may lead to a decline in the performance of the classifier, or for specific samples, the prediction results of the classifier may be unreliable.

[0037] To solve this problem, in the technical solution of this application, parametric geometric relationship transition prior features are rigidly unified for the building personnel brightness feature matrix to perform two-way constraints on each eigenvalue sample in the building personnel brightness feature matrix in terms of information entropy dimension.

[0038] Specifically, first, by parameterizing the geometric relationship of the eigenvalue samples at each position in the building personnel brightness feature matrix, a transition prior feature is established, which can reflect the distribution and variation of the eigenvalue samples among different categories. Then, a rigid unification mechanism can be used to adjust and optimize the transition prior feature to make it more conform to the internal structure and logic of the data. In this way, two-way constraints can be performed on the eigenvalue samples in terms of information entropy dimension, that is, both maintaining high entropy of the eigenvalue samples within the category and low entropy of the eigenvalue samples among categories, thereby improving the discrimination ability and expression ability of the eigenvalue samples.

[0039] Specifically, in the embodiment of this application, the building personnel brightness optimization module includes: a probability unit for inputting the building personnel brightness feature matrix into the Sigmoid function to obtain a probability building personnel brightness feature matrix; and a rigid unification unit for performing rigid unification of parametric geometric relationship transition prior features on the probability building personnel brightness feature matrix to obtain an optimized building personnel brightness feature matrix.

[0040] More specifically, in the embodiment of this application, the rigid unification unit is used to: perform rigid unification of parametric geometric relationship transition prior features on the probability building personnel brightness feature matrix with the following optimization formula to obtain the optimized building personnel brightness feature matrix; where the optimization formula is: Where, represents the eigenvalue at the position in the probability building personnel brightness feature matrix, represents a predetermined hyperparameter, represents the logarithmic function value with base 2, represents the eigenvalue at the position in the optimized building personnel brightness feature matrix.

[0041] In the embodiment of the present application, the building brightness adjustment module 200 is configured to obtain a classification result by using a classifier for the optimized building personnel brightness feature matrix, and the classification result is used to indicate that the lighting brightness of the building at the current time point should be increased, decreased, or remain unchanged. Considering that by classifying the building personnel brightness feature matrix, the lighting state at the current time point can be classified as increased, decreased, or unchanged. In this way, automatic lighting control can be achieved, and then the brightness of the lighting device can be automatically adjusted according to the classification result. For example, if the classification result is that the lighting brightness should be increased, the system can automatically increase the brightness level of the lighting device to provide sufficient lighting; if the classification result is that the lighting brightness should be decreased, the system can automatically reduce the brightness level of the lighting device to save energy; if the classification result is unchanged, the system can maintain the current lighting brightness level. Specifically, through the classification of lighting brightness, corresponding lighting adjustments can be made according to the needs at different time points. This can improve the operating efficiency of the building. For example, during the day or when there is sufficient natural light, the lighting brightness can be reduced to save energy; while at night or when the light is insufficient, the lighting brightness can be increased to provide sufficient lighting. Secondly, according to the classification result, the brightness of the lighting device can be adjusted so that the lighting level matches the user's needs. For example, a higher lighting brightness may be required in a meeting room or an office, while a lower lighting brightness may be desired in a rest area or a recreational venue. By adjusting the lighting brightness according to the classification result, a lighting environment suitable for user activities and needs can be provided, enhancing the user's comfort and satisfaction.

[0042] Specifically, in the embodiment of the present application, the building brightness adjustment module includes: an expansion unit configured to expand the optimized building personnel brightness feature matrix into a building personnel brightness feature vector according to a row vector or a column vector; a fully connected encoding unit configured to perform fully connected encoding on the building personnel brightness feature vector by using the fully connected layer of the classifier to obtain an encoded building personnel brightness feature vector; and a classification result unit configured to input the encoded building personnel brightness feature vector into the Softmax classification function of the classifier to obtain the classification result.

[0043] In summary, the intelligent building automation control system 100 based on the embodiment of the present application is elucidated. It adopts artificial intelligence technology based on a deep neural network model, obtains the brightness values inside the building, the brightness values outside the building, and the surveillance videos of the building at multiple predetermined time points within a predetermined time period, combines the time series change information, fuses data in different dimensions, and uses a time series feature extractor to capture the feature changes at different time scales, and fuses the spatial information of the building personnel surveillance images with the brightness features to generate a classification result indicating that the lighting brightness of the building at the current time point should be increased, decreased, or remain unchanged. This method can guide the lighting system to automatically adjust the lighting brightness, achieve intelligent control of the lighting environment, and improve energy utilization efficiency and user experience.

[0044] Exemplary method Figure 4 is a flowchart of a smart building automation control method according to an embodiment of the present application. As Figure 4 shown, the smart building automation control method according to an embodiment of the present application includes: S110, obtaining the brightness values inside the building, the brightness values outside the building, and the surveillance video of the building at multiple predetermined time points within a predetermined time period; S120, passing the surveillance video of the building through a person target detection network model to obtain a building person surveillance image; S130, passing the building person surveillance image through a person spatial attention mechanism based on a convolutional neural network model to obtain a building person surveillance spatial enhanced feature matrix; S140, arranging the brightness values inside the building and the brightness values outside the building at the multiple predetermined time points along the time dimension respectively to obtain an in-building brightness time series input vector and an out-building brightness time series input vector; S150, calculating the difference between the in-building brightness at two adjacent time points in the in-building brightness time series input vector to obtain an in-building brightness time series change input vector, and calculating the difference between the out-building brightness at two adjacent time points in the out-building brightness time series input vector to obtain an out-building brightness time series change input vector; S160, concatenating the in-building brightness time series input vector and the in-building brightness time series change input vector to obtain an in-building brightness multi-dimensional input vector, and concatenating the out-building brightness time series input vector and the out-building brightness time series change input vector to obtain an out-building brightness multi-dimensional input vector; S170, passing the in-building brightness multi-dimensional input vector and the out-building brightness multi-dimensional input vector through a multi-scale time series feature extractor including a first convolutional layer and a second convolutional layer to obtain an in-building brightness multi-scale feature vector and an out-building brightness multi-scale feature vector; S180, performing correlation encoding on the in-building brightness multi-scale feature vector and the out-building brightness multi-scale feature vector to obtain a building brightness correlation matrix, and fusing it with the building person surveillance spatial enhanced feature matrix to obtain a building person brightness feature matrix; S190, optimizing the building person brightness feature matrix to obtain an optimized building person brightness feature matrix; and, S200, passing the building person brightness feature matrix through a classifier to obtain a classification result, where the classification result is used to indicate whether the lighting brightness of the building at the current time point should be increased, decreased, or remain unchanged.

[0045] Here, those skilled in the art can understand that the specific operations of each step in the above smart building automation control method have been described in detail in the description of the Figures 1 to 3 smart building automation control system above, and therefore, the repeated description thereof will be omitted.

[0046] The above are only examples of the principles of the present disclosure, and those skilled in the art can make various modifications without departing from the scope of the present disclosure. The above embodiments are presented for illustrative purposes rather than limitation. The present disclosure can also take many forms other than those explicitly described herein. Therefore, it should be emphasized that the present disclosure is not limited to the explicitly disclosed methods, systems, and devices, but is intended to include variations and modifications within the spirit scope of the appended claims.

Claims

1. An intelligent building automation control system, characterized in that Including: A building brightness monitoring and acquisition module, configured to obtain the brightness values inside the building, the brightness values outside the building, and the monitoring videos of the building at multiple predetermined time points within a predetermined time period; A building personnel detection module, configured to obtain a building personnel monitoring image by passing the monitoring video of the building through a personnel target detection network model; A building space feature acquisition module, configured to obtain a building personnel monitoring space enhanced feature matrix by passing the building personnel monitoring image through a personnel space attention mechanism based on a convolutional neural network model; A building brightness time series arrangement module, configured to arrange the brightness values inside the building and the brightness values outside the building at the multiple predetermined time points into an in-building brightness time series input vector and an out-building brightness time series input vector respectively according to the time dimension; A time series change calculation module, configured to calculate the difference between the in-building brightness at two adjacent time points in the in-building brightness time series input vector to obtain an in-building brightness time series change input vector and calculate the difference between the out-building brightness at two adjacent time points in the out-building brightness time series input vector to obtain an out-building brightness time series change input vector; A building brightness concatenation module, configured to concatenate the in-building brightness time series input vector and the in-building brightness time series change input vector to obtain an in-building brightness multi-dimensional input vector, and concatenate the out-building brightness time series input vector and the out-building brightness time series change input vector to obtain an out-building brightness multi-dimensional input vector; A brightness multi-scale extraction module, configured to obtain an in-building brightness multi-scale feature vector and an out-building brightness multi-scale feature vector by passing the in-building brightness multi-dimensional input vector and the out-building brightness multi-dimensional input vector through a multi-scale time series feature extractor including a first convolutional layer and a second convolutional layer; A brightness correlation feature extraction module, configured to perform correlation coding on the in-building brightness multi-scale feature vector and the out-building brightness multi-scale feature vector to obtain a building brightness correlation matrix and fuse it with the building personnel monitoring space enhanced feature matrix to obtain a building personnel brightness feature matrix; A building personnel brightness optimization module, configured to optimize the building personnel brightness feature matrix to obtain an optimized building personnel brightness feature matrix; A building brightness adjustment module, configured to obtain a classification result by passing the optimized building personnel brightness feature matrix through a classifier, and the classification result is used to indicate whether the lighting brightness of the building at the current time point should be increased, decreased, or remain unchanged.

2. The intelligent building automation control system according to claim 1, characterized in that The building personnel detection module is configured to: The personnel target detection network model is an anchor window-based target detection network, and the anchor window-based target detection network is Fast R-CNN, Faster R-CNN, or RetinaNet.

3. The intelligent building automation control system according to claim 2, wherein The building space feature acquisition module includes: A personnel convolutional encoding unit, configured to perform deep convolutional encoding on the building personnel monitoring image using the convolutional encoding part of the personnel space attention mechanism based on the convolutional neural network model to obtain a building personnel initial convolutional feature map; A personnel space attention unit, configured to input the building personnel initial convolutional feature map into the spatial attention part of the personnel space attention mechanism to obtain a building personnel space attention map; A personnel feature activation unit, configured to obtain a building personnel spatial attention feature map by passing the building personnel spatial attention map through a Softmax activation function; A personnel feature multiplication unit, configured to calculate the element-wise multiplication of the building personnel spatial attention feature map and the initial convolutional feature map of the building personnel to obtain a building personnel monitoring spatial enhanced feature map; A personnel feature pooling unit, configured to perform pooling on the building personnel monitoring spatial enhanced feature map along the channel dimension to obtain the building personnel monitoring spatial enhanced feature matrix.

4. The intelligent building automation control system according to claim 3, wherein The brightness multi-scale extraction module includes: A first-scale brightness feature extraction unit, configured to input the in-building brightness multi-dimensional input vector into the first convolutional layer of the multi-scale time series feature extractor to obtain a first-scale in-building brightness feature vector, where the first convolutional layer has a first one-dimensional convolutional kernel with a first length; A second-scale brightness feature extraction unit, configured to input the in-building brightness multi-dimensional input vector into the second convolutional layer of the multi-scale time series feature extractor to obtain a second-scale in-building brightness feature vector, where the second convolutional layer has a second one-dimensional convolutional kernel with a second length, and the first length is different from the second length; A fused brightness feature unit, configured to cascade the first-scale in-building brightness feature vector and the second-scale in-building brightness feature vector using the cascade layer of the multi-scale time series feature extractor to obtain the in-building brightness multi-scale feature vector.

5. The intelligent building automation control system according to claim 4, wherein The brightness correlation feature extraction module is configured to: Calculate the product between the transposed vector of the in-building brightness multi-scale feature vector and the out-building brightness multi-scale feature vector respectively to obtain the building brightness correlation matrix.

6. The intelligent building automation control system according to claim 5, wherein, The building personnel brightness optimization module is configured to: A probability unit, configured to input the building personnel brightness feature matrix into a Sigmoid function to obtain a probability building personnel brightness feature matrix; A rigid consistency unit, configured to perform rigid consistency of the parametric geometric relationship transition prior feature on the probability building personnel brightness feature matrix to obtain an optimized building personnel brightness feature matrix.

7. The intelligent building automation control system according to claim 6, characterized in that, The rigid consistency unit is configured to: Perform rigid consistency of the parametric geometric relationship transition prior feature on the probability building personnel brightness feature matrix according to the following optimization formula to obtain the optimized building personnel brightness feature matrix; Among them, the optimization formula is as follows: Among them, represents the eigenvalue at the position in the probabilistic building personnel brightness feature matrix, represents a predetermined hyperparameter, represents the logarithmic function value with base 2, represents the eigenvalue at the position in the optimized building personnel brightness feature matrix.

8. The intelligent building automation control system according to claim 7, characterized in that, The building brightness adjustment module includes: An expansion unit, configured to expand the optimized building personnel brightness feature matrix into a building personnel brightness feature vector according to a row vector or a column vector; A fully connected encoding unit, configured to perform fully connected encoding on the building personnel brightness feature vector using the fully connected layer of the classifier to obtain an encoded building personnel brightness feature vector; A classification result unit, configured to input the encoded building personnel brightness feature vector into the Softmax classification function of the classifier to obtain the classification result.

9. A method for intelligent building automation control, characterized in that, It includes: Obtain the brightness values inside the building, the brightness values outside the building, and the monitoring video of the building at multiple predetermined time points within a predetermined time period; Pass the monitoring video of the building through a personnel target detection network model to obtain a building personnel monitoring image; The building personnel monitoring image is processed through a personnel spatial attention mechanism based on a convolutional neural network model to obtain a building personnel monitoring spatial enhancement feature matrix; The brightness values inside the building and outside the building at multiple predetermined time points are respectively arranged in the time dimension to form an in-building brightness time series input vector and an out-building brightness time series input vector; Calculate the difference between the in-building brightness at two adjacent time points in the in-building brightness time series input vector to obtain an in-building brightness time series change input vector, and calculate the difference between the out-building brightness at two adjacent time points in the out-building brightness time series input vector to obtain an out-building brightness time series change input vector; Concatenate the in-building brightness time series input vector and the in-building brightness time series change input vector to obtain an in-building brightness multi-dimensional input vector, and concatenate the out-building brightness time series input vector and the out-building brightness time series change input vector to obtain an out-building brightness multi-dimensional input vector; Pass the in-building brightness multi-dimensional input vector and the out-building brightness multi-dimensional input vector through a multi-scale time series feature extractor including a first convolutional layer and a second convolutional layer to obtain an in-building brightness multi-scale feature vector and an out-building brightness multi-scale feature vector; Associate and encode the in-building brightness multi-scale feature vector and the out-building brightness multi-scale feature vector to obtain a building brightness association matrix, and fuse it with the building personnel monitoring spatial enhancement feature matrix to obtain a building personnel brightness feature matrix; Optimize the building personnel brightness feature matrix to obtain an optimized building personnel brightness feature matrix; Pass the optimized building personnel brightness feature matrix through a classifier to obtain a classification result, which is used to indicate whether the lighting brightness of the building at the current time point should be increased, decreased, or remain unchanged.

10. The intelligent building automation control method according to claim 9, wherein, Pass the monitoring video of the building through a personnel target detection network model to obtain a building personnel monitoring image, including: The personnel target detection network model is an anchor window-based target detection network, and the anchor window-based target detection network is Fast R-CNN, Faster R-CNN, or RetinaNet.

Images