Fresh air purification energy-saving method and system based on indoor environment monitoring

Abstract

This application provides a fresh air purification and energy-saving method and system based on indoor environmental monitoring, relating to the field of ventilation system control technology. The method involves first acquiring a video stream, ambient sound signals, and static pressure feedback signals; then generating a sound intensity distribution; next, performing moving target segmentation and morphological recognition on the video stream to generate a personnel distribution map; then, sequentially associating the sound intensity distribution with the personnel distribution map to form a predictive personnel density distribution map; subsequently, inputting the predictive personnel density distribution map into a trained network model to output an initial airflow command; then, dynamically adjusting the power amplification factor of the drive circuit based on the static pressure feedback signal; and finally, controlling the operation of the fresh air system's fan based on the initial airflow command and the adjusted power amplification factor. The technical solution provided in this application, based on audiovisual fusion for personnel trend prediction and pressure closed-loop regulation, achieves predictive on-demand air supply from the fresh air system, balancing energy saving and air quality.

Description

Technical Field

[0001] This application relates to the field of ventilation system control technology, and in particular to a fresh air purification and energy-saving method and system based on indoor environmental monitoring. Background Technology

[0002] As large public buildings such as office buildings and shopping malls continue to raise their requirements for indoor air quality and energy conservation, fresh air purification systems need to make precise and forward-looking air volume adjustments based on the indoor activity of people. This requires the system to be able to detect the trend of people gathering in advance in scenarios with dynamic changes and uneven distribution of people density, and adjust the air supply volume of each area accordingly, so as to ensure air quality while avoiding energy waste.

[0003] The current solution involves installing carbon dioxide concentration sensors at fixed locations indoors. When the carbon dioxide concentration in a certain area exceeds a set threshold, the fresh air volume in that area is increased. This solution uses CO2 concentration as an indirect indicator of the presence and density of people, thus achieving a certain degree of on-demand ventilation.

[0004] However, changes in carbon dioxide concentration lag significantly behind the occurrence of gatherings, resulting in a delayed response from the control system. This makes it impossible to make proactive adjustments when people begin to gather. At the same time, fixed-point sensors cannot accurately reflect changes in the distribution of people in the entire area, especially in areas not covered by sensors, which can easily lead to excessive or insufficient ventilation in certain areas. Summary of the Invention

[0005] This application provides a fresh air purification and energy-saving method and system based on indoor environmental monitoring, which solves the problems of inaccurate control and high energy consumption of fresh air systems in the prior art due to the lag in environmental perception and incomplete spatial coverage.

[0006] Firstly, this application provides a fresh air purification and energy-saving method based on indoor environmental monitoring, including:

[0007] Acquire video streams captured by image acquisition devices within the target area, as well as ambient sound signals captured by sound acquisition arrays and static pressure feedback signals within air supply ducts within the target area;

[0008] The ambient sound signal acquired by the sound acquisition array is subjected to time delay analysis to locate the spatial position of the sound source within the target area, and the sound intensity distribution of the target area is generated by combining the amplitude of the sound signal.

[0009] The video stream is segmented into moving targets and its shape is recognized to generate a personnel distribution map reflecting the real-time location of people within the target area;

[0010] By correlating the sound intensity distribution with the population distribution map in a time sequence, the aggregation trend of each zone in the target area within a preset time period is deduced, and a predictive population density distribution map is formed.

[0011] The predictive population density distribution map is input into the trained network model to output the initial air volume command corresponding to each fresh air outlet in the target area;

[0012] The power amplification factor of the drive circuit is dynamically adjusted based on the static pressure feedback signal to obtain the adjusted power amplification factor.

[0013] The operation of the fresh air system's fan is controlled based on the initial air volume command and the adjusted power amplification factor.

[0014] Optionally, time delay analysis is performed on the ambient sound signals acquired by the sound acquisition array to locate the spatial position of the sound source within the target area, and the sound intensity distribution of the target area is generated by combining the amplitude of the sound signals, including:

[0015] Receives multiple audio signals collected by audio acquisition units located at different physical locations within the target area;

[0016] For the multiple sound signals, based on the time difference between the arrival of the sound signals at any two sound acquisition units and the speed of sound propagation in the air, the distance difference from the potential sound source point to any two sound acquisition units is determined.

[0017] Based on the multiple distance differences, a system of equations is constructed regarding the spatial location of potential sound source points, and the system of equations is solved to obtain the spatial coordinates of at least one sound source within the target area.

[0018] For each sound source obtained by the solution, select a sound acquisition unit that is closest to the sound source, and obtain the signal amplitude corresponding to the sound source in the sound signal acquired by the sound acquisition unit.

[0019] Based on the spatial coordinates of each sound source and the corresponding signal amplitude, the energy contribution of the sound source to each preset grid point in the target area is calculated;

[0020] The energy contributions of all sound sources to the same grid point are superimposed to obtain the sound intensity distribution of the target area.

[0021] Optionally, the video stream is subjected to moving target segmentation and morphological recognition to generate a personnel distribution map reflecting the real-time location of personnel within the target area, including:

[0022] Extract any two consecutive video frames from the video stream, compare the difference in brightness values ​​of corresponding pixels in the two video frames, and mark pixels whose brightness value changes exceed a preset change threshold as changed pixels.

[0023] Spatial connectivity analysis is performed on the changed pixels to group multiple spatially adjacent changed pixels into an independent motion region.

[0024] For each independent motion region, the shape contour features representing the independent motion region are extracted from the current video frame. The shape contour features of the independent motion region are compared with the pre-stored human body contour shape feature set to determine whether the shape contour of the independent motion region conforms to human features. Independent motion regions whose shape contours conform to human features are determined as valid human body regions.

[0025] Calculate the geometric center coordinates of the effective personnel area in the current video frame, and use the geometric center coordinates as the real-time position of the personnel within the effective personnel area;

[0026] Based on the real-time locations of all valid personnel within the target area, a personnel distribution map reflecting the real-time locations of personnel within the target area is generated.

[0027] Optionally, the sound intensity distribution is correlated with the population distribution map in a time sequence to deduce the aggregation trend of each zone of the target area within a preset time period, and a predictive population density distribution map is formed, including:

[0028] Obtain sound intensity distribution and personnel distribution maps corresponding to the current moment and multiple consecutive historical moments to form a historical sequence;

[0029] The historical sequence was analyzed to identify the temporal and spatial correspondence between high-value areas of sound intensity distribution and densely populated areas on the population distribution map, and to establish a correlation mapping between the sound activity distribution pattern of the high-value areas and the population location distribution pattern of the densely populated areas.

[0030] Based on the aforementioned association mapping, sound feature regions related to personnel gathering activities are extracted;

[0031] Based on the historical sequence, calculate the number of people in each partition within the target area at the current moment, and combine this with the sound intensity value of the sound feature area at the current moment to infer the intensity level of gathering activity in each partition within the target area at the current moment.

[0032] The intensity level of gathering activity in each partition at the current moment, the number of people in each partition at the current moment, and the rate of change of the number of people in each partition in the historical sequence are used as a comprehensive feature vector to describe the current gathering state.

[0033] The comprehensive feature vector is input into a pre-trained state inference model. Based on the changes in the number of people in subsequent periods corresponding to similar comprehensive feature vectors in the historical sequence, the predicted number of people in each partition of the target area after a preset period is inferred.

[0034] Based on the predicted number of people in each zone, a predictive population density distribution map is generated.

[0035] Secondly, this application provides a fresh air purification and energy-saving system based on indoor environmental monitoring, including:

[0036] The acquisition module is used to acquire video streams captured by the image acquisition device, ambient sound signals captured by the sound acquisition array, and static pressure feedback signals in the air supply duct within the target area.

[0037] The analysis module is used to perform time delay analysis on the environmental sound signals acquired by the sound acquisition array, locate the spatial position of the sound source in the target area, and generate the sound intensity distribution of the target area by combining the amplitude of the sound signal.

[0038] The recognition module is used to perform moving target segmentation and morphological recognition on the video stream and generate a personnel distribution map reflecting the real-time location of personnel within the target area;

[0039] The extrapolation module is used to correlate the sound intensity distribution with the personnel distribution map in a time sequence, extrapolate the aggregation trend of each partition of the target area within a preset time period, and form a predictive personnel density distribution map;

[0040] The output module is used to input the predictive population density distribution map into the trained network model to output the initial air volume command corresponding to each fresh air outlet in the target area.

[0041] The adjustment module is used to dynamically adjust the power amplification factor of the drive circuit according to the static pressure feedback signal, so as to obtain the adjusted power amplification factor.

[0042] The control module is used to control the operation of the fresh air system's fan according to the initial air volume command and the adjusted power amplification factor.

[0043] Thirdly, this application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a fresh air purification and energy-saving method based on indoor environmental monitoring as described in the first aspect above.

[0044] Fourthly, this application provides a computer storage medium storing a computer program, which, when executed by a computer, implements a fresh air purification and energy-saving method based on indoor environmental monitoring as described in the first aspect.

[0045] This application first constructs a spatiotemporal correlation between personnel distribution and sound intensity distribution by integrating video and audio information, which can proactively deduce personnel gathering trends and generate predictive personnel density distribution. It overcomes the shortcomings of traditional carbon dioxide sensors with lag in response, enabling the fresh air system to adjust air volume commands in advance before the actual personnel density changes, realizing the transformation from delayed response to advance prediction, and effectively solving the problem of inaccurate control caused by perception delay.

[0046] Furthermore, the air volume command generated based on personnel density prediction is combined with the power amplification factor dynamically adjusted according to pipeline static pressure to jointly control the fan operation. This allows the system to compensate for pipeline pressure disturbances caused by changes in damper opening in real time when responding to predicted changes in air volume demand, maintaining stable system static pressure. Thus, while achieving on-demand ventilation, it ensures that the fan always operates near its high-efficiency operating point. This solves the problem of system pressure fluctuations and increased overall energy consumption caused by local air volume adjustment in traditional methods, achieving energy-saving effects while improving control accuracy.

[0047] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 A flowchart of a fresh air purification and energy-saving method based on indoor environmental monitoring provided in this application is shown;

[0050] Figure 2 This application provides a schematic diagram of the structure of a fresh air purification and energy-saving system based on indoor environmental monitoring.

[0051] Figure 3 A schematic diagram of the structure of a computing device provided in this application is shown. Detailed Implementation

[0052] To enable those skilled in the art to better understand the present application, the technical solution of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0053] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0054] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0055] Figure 1 This application provides a flowchart of a fresh air purification and energy-saving method based on indoor environmental monitoring, such as... Figure 1 As shown, the method includes:

[0056] Step 101: Acquire the video stream captured by the image acquisition device, the ambient sound signal captured by the sound acquisition array, and the static pressure feedback signal in the air supply duct within the target area.

[0057] In this step, the video stream refers to the sequence of images continuously captured and output by cameras deployed in the target area, used to record and analyze personnel activities and location changes within the area.

[0058] Ambient sound signals refer to raw sound waveform data collected by a set of microphones placed in a target area, used to analyze the source and intensity of the sound.

[0059] Static pressure feedback signal refers to the static pressure value inside the duct measured by a pressure sensor installed inside the fresh air system's air supply duct. It is used to reflect the pressure state of the airflow inside the duct and is obtained by measuring the pressure in real time through the pressure sensor and converting it into an electrical signal.

[0060] Image acquisition devices refer to network cameras installed on indoor ceilings or walls to continuously capture images of a target area, converting light signals into digital video streams through their optical lenses and image sensors.

[0061] A sound acquisition array refers to a acquisition device consisting of multiple microphones arranged in a specific geometric structure, used to synchronously acquire sound from different locations in space, obtained through the collaborative work of multiple microphone units.

[0062] In this step, the process begins with an image acquisition device, the core of which is an image sensor. This sensor uses the photoelectric effect to convert light entering the lens into continuous electrical signals. A built-in video encoder then encodes and encapsulates these electrical signals in real time according to H.264 or a similar compression standard, generating a transmittable video stream. Next, a sound acquisition array is used. Each microphone unit in this array contains a diaphragm that converts sound wave vibrations into analog electrical signals. These analog signals are fed into a multi-channel analog-to-digital converter (ADC). This ADC digitizes each signal with a uniform sampling frequency and quantization precision, ultimately outputting synchronized multi-channel digital audio. The system collects ambient sound signals. Simultaneously, static pressure is acquired via a piezoresistive or capacitive pressure sensor. The sensor's internal sensing element changes resistance or capacitance when subjected to gas pressure within the pipe. A signal conditioning circuit converts this change into a standard analog current signal. Finally, a data acquisition card samples and digitizes this current signal to obtain the static pressure feedback signal. Ultimately, a central processing unit (CPU) simultaneously receives network data packets from the video encoder, audio data streams from the analog-to-digital converter chip, and digital pressure values ​​from the data acquisition card via Ethernet or fieldbus communication interfaces, completing the unified timestamp marking and buffering of these three types of signals.

[0063] For example, in a large open-plan office area A, a data acquisition layer was deployed to implement energy-saving controls. First, five high-definition network cameras were evenly installed on the ceiling of area A as image acquisition devices. The CMOS sensors inside these cameras continuously sense light, and their processors perform H.265 encoding to generate a continuous video stream, which is then transmitted via the local area network. Simultaneously, an omnidirectional microphone was installed at each of the four corners and the center of area A. These five microphones together form a sound acquisition array. The captured sound waves are converted into analog signals and connected to an external eight-channel... The audio acquisition card performs synchronous digitization, generating five synchronous ambient sound signals. In addition, a piezoresistive pressure transmitter is installed on the main air supply duct of the fresh air system serving Area A, which converts the detected duct static pressure into a 4-20mA current signal in real time. This current signal is connected to the analog input module in the control cabinet and converted into a digital value, namely the static pressure feedback signal. Finally, a server deployed in the computer room receives RTSP video streams from five cameras, network audio streams from the audio acquisition card, and static pressure data from the PLC simultaneously through a switch, completing the synchronous acquisition of multi-source data in Area A.

[0064] Step 102: Perform time delay analysis on the ambient sound signal acquired by the sound acquisition array to locate the spatial position of the sound source within the target area, and generate the sound intensity distribution of the target area by combining the amplitude of the sound signal.

[0065] Optionally, step 102 may specifically include:

[0066] Step 1021: Receive multiple audio signals collected by audio acquisition units located at different physical locations within the target area.

[0067] Step 1022: For the multiple sound signals, based on the time difference of the sound signals arriving at any two sound acquisition units and the speed of sound propagation in the air, determine the distance difference from the potential sound source point to any two sound acquisition units.

[0068] Step 1023: Construct a system of equations about the spatial location of potential sound source points based on the multiple distance differences, and solve the system of equations to obtain the spatial coordinates of at least one sound source within the target area.

[0069] Step 1024: For each sound source obtained by the solution, select a sound acquisition unit that is closest to the sound source, and obtain the signal amplitude corresponding to the sound source in the sound signal acquired by the sound acquisition unit.

[0070] Step 1025: Calculate the energy contribution of each sound source to each preset grid point within the target area based on the spatial coordinates of each sound source and the corresponding signal amplitude.

[0071] Step 1026: Superimpose the energy contributions of all sound sources to the same grid point to obtain the sound intensity distribution of the target area.

[0072] In this step, the sound intensity distribution of the target area refers to a two-dimensional or three-dimensional map that describes the intensity of sound energy at each point within the target area, used to visually reflect the distribution of sound activity in space.

[0073] Multiple audio signals refer to multiple digital audio streams synchronously acquired from each independent audio acquisition unit in the audio acquisition array, used to provide synchronous audio data from different observation points in space.

[0074] The time difference refers to the time difference between the arrival of the same sound event at any two different sound acquisition units in the array. It is used to calculate the distance difference between the sound source and the two acquisition units, and is obtained by comparing the offset of the waveforms of the two sound signals on the time axis.

[0075] Distance difference refers to the difference in the distance traveled by a sound from the same sound source to two different sound acquisition units. It is used to construct a mathematical relationship that constrains the location of the sound source and is calculated by multiplying the time difference by the speed of sound in the air.

[0076] Potential sound source points refer to spatial locations that are assumed to be capable of emitting sound during the localization calculation process. They are used as unknown variables to solve for the actual sound source location and are defined by geometric equations based on distance differences.

[0077] Spatial location coordinates refer to a set of values ​​in a predefined spatial coordinate system used to uniquely represent the location of a sound source, and are used to accurately describe the physical location of the sound source within the target area.

[0078] Signal amplitude refers to the intensity or response value of a sound signal waveform within a specific time period. It is used to characterize the strength of the sound emitted by the sound source. It is obtained by extracting the signal envelope or root mean square value of the corresponding time period from the signal recorded by the sound acquisition unit closest to the sound source.

[0079] The energy contribution of a preset grid point refers to the sound energy value contributed by a located sound source to a pre-divided fixed grid point within the target area after the sound emitted by the source has attenuated during propagation. It is used to quantify the contribution of the sound source to the total sound intensity at that point.

[0080] In this step, firstly, a multi-channel audio acquisition card simultaneously receives digital recordings from all microphones in the array, obtaining multiple audio signals. Secondly, cross-correlation calculation technology is used to compare any two audio signals. By sliding the waveform of one signal, the sliding time when it is most aligned with the waveform of the other signal is found. This time is the time difference. Then, this time difference is multiplied by the fixed speed of sound in the air to obtain the distance difference between the two microphones corresponding to the two signals and the sound source. Next, the multiple distance differences obtained for different microphone pairs are combined to establish a hyperbolic equation for each pair of distance differences. All equations form a system of equations. Then, the least squares method is used to solve this system of equations, and finally, the coordinates of one or more points that make all equations as true as possible at the same time are obtained. These coordinates are the spatial coordinates of the sound source.

[0081] Next, for each sound source coordinate found, the straight-line distance between the known installation positions of all microphones in the array and that coordinate is calculated. The microphone with the shortest distance is selected, and then the segment of sound corresponding to the previous cross-correlation calculation is extracted from the recording signal of this nearest microphone. The root mean square value of this sound waveform is taken, and this calculation result is the signal amplitude. Then, a uniform virtual grid is laid on the map of the target area, and for each sound source, the straight-line distance from its spatial coordinates to each virtual grid point is calculated. Then, a simplified physical attenuation model based on the inverse square law is applied to divide the signal amplitude of the sound source coordinate by the square of its distance to the grid point, thereby calculating the energy contribution of the sound source to the grid point. Finally, a blank distribution map of the same size as the virtual grid is created, and then all the located sound sources are traversed. The energy contribution value calculated by each sound source to the same grid point is accumulated and added to the blank map position corresponding to that grid point. When all the contributions of all sound sources are superimposed, the map filled with the accumulated value is the final sound intensity distribution of the target area.

[0082] For example, following the specific implementation of the previous step, five ambient sounds were first recorded synchronously in office area A and analyzed. Then, after the server read these five recordings, it detected a continuous laugh and selected the recordings from microphones 1 and 2. Through cross-correlation calculation, it was found that the waveform of microphone 2 was delayed by 1 millisecond compared to microphone 1, thus obtaining the time difference. Next, this time difference was multiplied by the speed of sound (340 meters per second) to obtain a distance difference of 0.34 meters. The distance differences between other combinations, such as microphones 1 and 3, 2 and 4, were obtained using a similar method. Then, using these distance differences as constraints, a system of equations was established, and a least-squares solver was used to calculate the coordinates of a sound source located near the sofa in rest area B. Afterwards, it was determined that microphone 4, installed in the corner of rest area B, was closest to this coordinate. Therefore, the waveform of the laugh segment was extracted from the recording of microphone 4, and its root mean square amplitude was calculated as the signal amplitude of the laugh. Subsequently, the floor plan of office area A was divided into squares with sides of 1 meter. For the sound source at the sofa, the distance from it to the center of each square was calculated. Assuming that sound energy decreases with the square of distance, divide the signal amplitude value obtained earlier by the square of each distance to get the energy contribution of the laughter to each square. Finally, initialize a blank grid diagram and fill all the calculated contribution values ​​into the corresponding squares. If there is only one sound source, this diagram is the current sound intensity distribution. If there are other sound sources, such as distant keyboard sounds, calculate their contribution values ​​and add them to the corresponding squares to form the final distribution diagram.

[0083] Step 103: Perform moving target segmentation and morphological recognition on the video stream to generate a personnel distribution map reflecting the real-time location of personnel within the target area.

[0084] Optionally, step 103 may specifically include:

[0085] Step 1031: Extract any two consecutive video frames from the video stream, compare the brightness value difference of corresponding pixels in the two video frames, and mark the pixels whose brightness value changes by more than a preset change threshold as changed pixels.

[0086] Step 1032: Perform spatial connectivity analysis on the changed pixels and merge multiple changed pixels that are spatially adjacent into an independent motion region.

[0087] Step 1033: For each independent motion region, extract the shape contour features representing the independent motion region from the current video frame, compare the shape contour features of the independent motion region with the pre-stored human body contour shape feature set, determine whether the shape contour of the independent motion region conforms to human body features, and determine the independent motion region whose shape contour conforms to human body features as a valid human body region.

[0088] Step 1034: Calculate the geometric center coordinates of the effective personnel area in the current video frame, and use the geometric center coordinates as the real-time position of the personnel within the effective personnel area.

[0089] Step 1035: Based on the real-time locations of personnel within all valid personnel areas, generate a personnel distribution map reflecting the real-time locations of personnel within the target area.

[0090] In this step, the personnel distribution map refers to a schematic diagram or coordinate list that marks the location of all identified personnel within the target area, used to visually demonstrate the real-time distribution of personnel in space.

[0091] Brightness value refers to the lightness or darkness of each pixel in a video image. It is used to determine which parts of the image have changed and is obtained by measuring and digitizing light intensity through an image sensor.

[0092] The preset change threshold refers to a pre-set numerical threshold used to determine whether a change in brightness is significant, and is used to filter out minor brightness fluctuations that may be caused by noise.

[0093] Variable pixels refer to those pixels whose brightness values ​​differ by more than a preset threshold in the comparison of two consecutive frames of images. They are used to initially indicate the location of motion in the image and are obtained through the inter-frame difference algorithm.

[0094] An independent motion region refers to a complete block formed by connecting a group of changing pixels that are spatially adjacent to each other. It is used to aggregate scattered changing points into a whole that may represent an independent moving object.

[0095] Shape profile features refer to a series of geometric parameters used to describe the external shape of an independent moving area. They are used to determine whether the shape of the area resembles a person and are obtained by calculating the spatial distribution characteristics of the boundary points of the area.

[0096] The pre-stored human body contour shape feature set refers to a pre-established database that contains the shape parameter range for various typical human body postures and proportions, and is used as a reference standard to determine whether a shape is humanoid.

[0097] The effective personnel area refers to the independent motion areas that are identified as conforming to human features after shape comparison, and is used to ultimately determine the precise image range in which the person is located in the video frame.

[0098] Geometric center point coordinates refer to the position of the center point of an effective personnel area within its image plane. They are represented by pixel coordinates or mapped physical coordinates and are used to represent the approximate location of a person within the area using a single point. They are obtained by calculating the average of the coordinates of all pixels in the area.

[0099] In this step, the inter-frame difference method is first used to calculate the absolute value of the difference in brightness values ​​of pixels at the same position in two consecutive frames of images extracted from the video stream. Then, this absolute value difference is compared with a pre-set value, namely a preset change threshold. All pixels with differences greater than the threshold are marked as changed pixels, resulting in a binary change mask. Next, an algorithm called connected component labeling is used on this change mask to scan the image and group all adjacent changed pixels in the horizontal, vertical, or diagonal directions into the same group. Each marked pixel group forms an independent motion region.

[0100] Next, for each independent motion region, its outer contour is first extracted using an edge detection algorithm. Then, a set of quantitative features that can describe the shape of the contour are calculated as its shape contour features, such as the aspect ratio, area and perimeter ratio of the contour, etc. These calculated features are then matched with a pre-stored set of human contour shape features, which is a standard feature range database obtained from a large number of human samples. If the feature matching is successful, the region is determined to conform to human shape features and is identified as a valid human region.

[0101] Then, for each valid personnel area, the geometric center coordinates of the valid personnel area are obtained by calculating the arithmetic mean of the coordinates of all pixels within it, and these coordinates are regarded as the real-time location of the personnel. Finally, all the calculated geometric center coordinates are drawn or mapped onto a base map corresponding to the target area space, thereby generating the final personnel distribution map.

[0102] For example, following the specific implementation of the previous step, the video stream is first processed based on the footage captured by the camera in office area A. Then, two consecutive frames are extracted, and by comparing pixel brightness changes, pixels showing significant changes at the doorway and several workstations are marked as changed pixels. Next, connectivity analysis is performed on these changed pixels, aggregating adjacent changed pixels at the doorway into a large independent motion region A, and aggregating the points at the workstations into several smaller regions B, C, etc. Subsequently, the shape and contour features of the independent motion region A are extracted and calculated. It is found that its aspect ratio and contour features match a pre-stored human shape library, thus determining it as a valid personnel region. Region B also matches successfully. Region C is excluded due to shape mismatch. Then, the average value of all pixel coordinates within valid regions A and B is calculated to obtain the coordinates of their respective geometric center points. For example, the center coordinates of region A are approximately 320, 180. Finally, these two coordinate points are marked on a blank floor plan to generate a personnel distribution map, visually showing that two people are currently located at the doorway and a certain workstation, respectively.

[0103] Step 104: Associate the sound intensity distribution with the personnel distribution map in chronological order, deduce the aggregation trend of each partition of the target area within a preset time period, and form a predictive personnel density distribution map.

[0104] Optionally, step 104 may specifically include:

[0105] Step 1041: Obtain the sound intensity distribution and personnel distribution maps corresponding to the current moment and multiple consecutive historical moments to form a historical sequence.

[0106] Step 1042: Analyze the historical sequence to identify the temporal and spatial correspondence between high-value areas of sound intensity distribution and densely populated areas on the population distribution map, and establish a correlation mapping between the sound activity distribution pattern of the high-value areas and the population location distribution pattern of the densely populated areas.

[0107] Step 1043: Based on the association mapping, extract the sound feature regions related to the gathering of people.

[0108] Step 1044: Based on the historical sequence, calculate the number of people in each partition within the target area at the current moment, and combine the sound intensity value of the sound feature area at the current moment to infer the intensity level of gathering activities in each partition within the target area at the current moment.

[0109] Step 1045: The current gathering activity intensity level of each partition, the number of people in each partition at the current time, and the rate of change of the number of people in each partition in the historical sequence are used as a comprehensive feature vector to describe the current gathering state.

[0110] Step 1046: Input the comprehensive feature vector into the pre-trained state inference model, and infer the predicted number of people in each partition of the target area after the preset time period based on the changes in the number of people in subsequent periods corresponding to similar comprehensive feature vectors in the historical sequence.

[0111] Step 1047: Generate a predictive population density distribution map based on the predicted population numbers for each zone.

[0112] In this step, the aggregation trend refers to the predicted direction of change in the number of people in each zone within the target area over a future period, which may increase, decrease, or remain stable. It is used to predict the dynamic development of population aggregation.

[0113] A predictive population density distribution map is a map that depicts the estimated number or density of people in each zone of a target area at a predetermined time in the future, and is used to proactively guide the control of the fresh air system.

[0114] Historical sequences refer to a set of data containing sound intensity distribution maps and population distribution maps arranged in chronological order, which are used to provide context and patterns for analyzing changes.

[0115] High-value areas refer to local spatial ranges in a sound intensity distribution map where the sound intensity values ​​are significantly higher than the surrounding areas, and are used to identify hotspots of active sound.

[0116] Sound activity distribution patterns refer to the summaries of the temporal regularity, spatial location, and intensity changes of high-value areas in historical sequences, used to describe the typical behavior of sound events.

[0117] The population location distribution pattern refers to the temporal regularity and spatial location characteristics of densely populated areas in a historical sequence. It is used to describe typical behaviors of population gathering and is obtained by clustering and trajectory analysis of population distribution maps in historical sequences.

[0118] Association mapping refers to the mathematical or logical relationship established between sound activity distribution patterns and personnel location distribution patterns, describing how the two correspond to each other in time and space. It is used to reveal the intrinsic connection between sound and personnel activities and is obtained through statistical correlation analysis or pattern matching algorithms.

[0119] Sound feature regions refer to the spatial regions that are determined, based on correlation mapping, to be primarily caused by sounds generated by human gatherings and activities, and are used to more accurately focus on human-caused sound sources.

[0120] Sound intensity value refers to a specific numerical value that represents the amount of sound energy in a sound characteristic area, measured or calculated. It is used to quantify the current acoustic activity of the area and is obtained by reading the value of the corresponding area in the sound intensity distribution map at the current moment.

[0121] The intensity level of gathering activities refers to a qualitative or semi-quantitative indicator used to classify the intensity of current activities of people in a certain area, and is used to comprehensively measure static number of people and dynamic noise.

[0122] The rate of change of the number of people refers to how fast the number of people in a certain area increases or decreases per unit time. It is used to describe the speed of people gathering or dispersing. It is obtained by calculating the time derivative of the number of people in that area at the most recent time in the historical sequence.

[0123] A comprehensive feature vector is a one-dimensional array that combines multiple features such as the intensity level of clustered activities, the current number of people, and the rate of change of the number of people in sequence. It is used to comprehensively and structurally describe the current clustering state of a partition.

[0124] A pre-trained state projection model refers to a mathematical model that is trained using a large amount of historical data and can predict future state changes based on the current state input, used to project the future number of people.

[0125] The predicted number of people refers to the estimated number of people in each zone of the target area at a specific future time, output by the state extrapolation model. It is used to generate a predictive distribution map and is obtained by inputting the comprehensive feature vector into the trained model and performing forward computation.

[0126] In this step, firstly, sound intensity distribution maps and personnel distribution maps from multiple consecutive moments before and after the current moment are extracted from storage using data caching and retrieval technologies, and arranged chronologically to form a historical sequence. Secondly, time series clustering and spatial statistical techniques are applied to analyze the historical sequence. First, spatial clustering is performed on each sound intensity distribution map to identify high-value areas, and their changes are tracked through time alignment to summarize the sound activity distribution pattern. At the same time, density clustering is performed on each personnel distribution map to identify densely populated areas and track their evolution to summarize the personnel location distribution pattern. Then, a correlation analysis algorithm is used to calculate the statistical correlation between the two patterns in time and space, thereby establishing an association mapping.

[0127] Next, based on the association mapping, the high-value sound areas that are highly correlated with the gathering activities are extracted as sound feature areas by setting a correlation threshold for filtering. Then, spatial statistical methods are used to calculate the number of people in each partition from the current personnel distribution map and read the average sound intensity value of the sound feature area from the current sound intensity distribution map. These values ​​are then input into a predefined hierarchical rule engine, which outputs the intensity level of the gathering activities corresponding to each partition through multi-condition judgment.

[0128] Then, through feature vector concatenation, the intensity level, number of people, and calculated rate of change of the number of people in each partition are combined into a comprehensive feature vector by performing time difference calculation on the number of people in that partition in the historical sequence. This vector is then input into a pre-trained state extrapolation model, which can be a sequence prediction model based on a recurrent neural network. The model performs nonlinear transformation and time-series extrapolation on the input vector through the parameters learned internally, and finally outputs the predicted number of people in each partition after a preset time period. Finally, through data mapping and visualization technology, the predicted number of people in each partition is used as attribute values ​​and filled into a grid base map corresponding to the physical partition, thereby generating a predictive population density distribution map.

[0129] For example, following the specific implementation of the previous step, the sound intensity map and personnel map of office area A over the past 5 minutes have been continuously recorded, forming a historical sequence for analysis. Secondly, analysis of the historical sequence reveals that high-intensity areas have repeatedly appeared near rest area B in the past, and the timing of these areas is always slightly later than or simultaneous with the personnel distribution map showing people moving towards rest area B, forming a densely populated area. A correlation mapping is then established, determining that rest area B is a typical sound characteristic area. At the current moment, the personnel distribution map shows 3 people in rest area B, and the current sound intensity value of the sound characteristic area in rest area B is relatively high. Combining these two points, it is inferred that... The activity intensity level of rest area B is high. Simultaneously, calculations show that the number of people in rest area B increased from 1 to 3 within the last 2 minutes, indicating a positive rate of change. The high intensity level, the current 3 people, and the positive rate of change are then combined to form a comprehensive feature vector for this area, which is input into a pre-trained LSTM prediction model. Based on historical similar patterns, the prediction model then predicts that the number of people in this area may increase to 5 in 2 minutes. Finally, similar operations are performed on other areas of the office area to obtain the predicted number of people in all areas after 2 minutes. Based on this, a predictive population density distribution map is generated, clearly showing that rest area B will become the most densely populated area.

[0130] Step 105: Input the predictive population density distribution map into the trained network model to output the initial air volume command corresponding to each fresh air outlet in the target area.

[0131] Optionally, step 105 may specifically include:

[0132] Step 1051: Map the predicted number of people in each zone represented by the predictive population density distribution map onto a grid map corresponding to the physical space of the target area to generate a gridded density data map.

[0133] Step 1052: Perform channelization processing on the gridded density data map, taking the predicted number of people at each grid point as an independent numerical channel and the relative position coordinates of the grid point in its respective partition as an independent spatial channel, so as to generate a multi-channel feature map containing numerical and spatial channels.

[0134] Step 1053: Input the multi-channel feature map into the first processing branch of the trained network model, extract multi-scale neighborhood information from the spatial channels in the multi-channel feature map, and generate a first intermediate feature map.

[0135] Step 1054: Input the gridded density data map into the second processing branch of the trained network model, perform global statistical feature extraction on the gridded density data map, and generate a second intermediate feature map that reflects the overall population distribution trend in the target area.

[0136] Step 1055: The first intermediate feature map and the second intermediate feature map are spliced ​​together along the feature dimension, and the spliced ​​combined feature map is subjected to nonlinear transformation and information compression to obtain a weight distribution map containing the wind volume demand weights of each location in the target area.

[0137] Step 1056: Based on the weight distribution diagram, the preset total system air volume requirement is allocated according to the weight of each location, and the theoretical air volume value corresponding to each fresh air outlet is calculated.

[0138] Step 1057: Based on the theoretical air volume value corresponding to each fresh air outlet, and combined with the duct resistance characteristics corresponding to each air outlet, the theoretical air volume value is adjusted to generate an initial air volume command.

[0139] In this step, the initial air volume command refers to a set of control commands that indicate how much air volume each fresh air outlet in the target area should provide, which is used to directly drive the fan and damper actuators.

[0140] A gridded density data map is a digital map that divides a target area into uniform small grids, each filled with the predicted number of people at the corresponding location. It is used to transform the density prediction of a region into fine spatial grid data.

[0141] Independent numerical channels refer to image channels in multi-channel feature maps that are specifically used to carry the one-dimensional data of the predicted number of people at grid points. They are used to input density information into the network model and are obtained by treating each value in the gridded density data map as an independent two-dimensional matrix element.

[0142] Independent spatial channels refer to image channels in multi-channel feature maps that are specifically used to carry the relative position coordinates of each grid point within its respective partition. They are used to input spatial structure information into the network model and are obtained by calculating the normalized coordinates of each grid point relative to the center of the partition.

[0143] Multichannel feature maps refer to a composite data structure containing multiple independent channels, used to simultaneously and structurally input numerical density information and spatial location information into a network model.

[0144] The first processing branch refers to a part of the trained network model specifically designed to process multi-channel feature maps to extract local spatial correlation features. It is used to generate a first intermediate feature map containing local context information and is implemented through a neural network submodule containing multiple convolutional and pooling layers.

[0145] The first intermediate feature map refers to the feature data map output by the first processing branch after local feature extraction and transformation, which is used to characterize the air volume demand correlation features of each location and its surrounding area.

[0146] The second processing branch refers to a portion of the trained network model specifically designed to process gridded density data maps to extract global distribution features. It is used to generate a second intermediate feature map that reflects the overall situation and is implemented through a neural network submodule containing global pooling layers and fully connected layers.

[0147] The second intermediate feature map refers to the feature data output by the second processing branch after global feature extraction and transformation, which is used to characterize the general features of the distribution of people in the entire target area.

[0148] A combined feature map refers to a larger feature data map formed by connecting the first intermediate feature map and the second intermediate feature map along the feature dimension, which is used to fuse local details and global situational information.

[0149] A weighted distribution map is a map with the same size as the grid map of the target area, where the value of each grid point represents the proportion of the required air volume at that location to the total air volume. It is used to quantify the relative air volume demand intensity at each point in the space.

[0150] The preset total system air volume requirement refers to the total fresh air volume required for the entire target area at the current moment, calculated in advance based on factors such as hygiene standards and space volume. It serves as the benchmark for calculating the specific air volume of each air outlet.

[0151] The theoretical air volume value refers to the calculated air volume value allocated to each fresh air outlet under ideal air duct conditions, based on the weight distribution diagram, without correction for actual air duct characteristics. It is used to initially determine the air volume supply target of each air outlet, and is obtained by multiplying the preset total system air volume demand by the sum of the weights of the areas covered by each air outlet.

[0152] Duct resistance characteristics refer to physical parameters that describe the degree of airflow obstruction by the duct network from the fan to each fresh air outlet. They are used to correct the deviation between theoretical calculations and actual air delivery capacity and are obtained by fitting the duct system design drawings, measured data, or historical operating data.

[0153] In this step, spatial interpolation is first used to map the predicted number of people in each zone of the predictive population density distribution map to a high-resolution uniform grid based on the geographical boundaries of the zones, generating a gridded density data map. Secondly, channelization is performed, where each value in the gridded density data map is directly copied into a new matrix to form an independent numerical channel. At the same time, the normalized coordinates of each grid point are calculated to generate two independent spatial channels. Finally, these three matrices are superimposed in the depth direction to generate a multi-channel feature map.

[0154] Next, the multi-channel feature map is input into the first processing branch of the network model. This first processing branch consists of multiple convolutional layers, which use convolutional kernels of different sizes, such as 3x3 and 5x5, to perform convolution operations on the input to extract multi-scale neighborhood features and output the first intermediate feature map. At the same time, the gridded density data map is input into the second processing branch. This second processing branch first compresses the entire map into a feature vector through a global average pooling layer, and then processes it through a fully connected layer to output the second intermediate feature map that represents the global distribution.

[0155] Then, the first and second intermediate feature maps are connected in the channel dimension through feature concatenation to form a combined feature map. The combined feature map is then processed with convolutional layers and ReLU activation function to perform nonlinear transformation and compression, and finally output a weight distribution map. Then, the preset total air volume requirement of the system is read. For each air outlet, the weight values ​​of all grid points in the weight distribution map of its service area are summed first. Then, this weight sum is multiplied by the total air volume requirement to calculate the theoretical air volume value of each air outlet.

[0156] Finally, based on the pre-stored duct resistance characteristic data, resistance compensation calculations are performed on the theoretical air volume value of each air outlet. For example, the theoretical value is appropriately increased for long duct outlets, thereby generating the final initial air volume command.

[0157] Specifically, following the concrete implementation of the previous step, a predictive personnel density distribution map of office area A has first been generated, showing that people will gather in rest area B, and based on this, an airflow command is generated. Next, this predictive personnel density distribution map data is mapped onto a fine grid covering the office area, generating a gridded density data map. Then, three data layers of the same size are created: a numerical layer directly carries the grid data, and two spatial layers carry the normalized coordinates of the grid points. These three layers are superimposed to form a multi-channel feature map. This multi-channel feature map is then input into the first branch of the network model, passing through multiple... The convolutional layer outputs the first intermediate feature map; simultaneously, the gridded density data map is input into the second branch, processed by global pooling and fully connected layers, and outputs the second intermediate feature map representing the overall situation; then the two are concatenated and processed by convolution and activation to generate a weight distribution map, in which the rest area B has a high weight; then, based on the preset total air volume, the theoretical air volume value is calculated according to the weight and allocation of the area under the jurisdiction of each air outlet; finally, compensation and adjustment are made in combination with the air duct resistance characteristics of each air outlet, for example, appropriately increasing the air volume for air outlets with long ducts, generating the final initial air volume command and issuing it.

[0158] This step leverages the model's ability to comprehensively analyze the spatial details and overall situation of density, generate reasonable demand weights, and make compensation adjustments based on actual duct resistance. Ultimately, it outputs precise initial airflow commands, providing a direct decision-making basis for the fresh air system to achieve on-demand, efficient, and precise air delivery.

[0159] Step 106: Dynamically adjust the power amplification factor of the drive circuit according to the static pressure feedback signal to obtain the adjusted power amplification factor.

[0160] Optionally, step 106 may specifically include:

[0161] Step 1061: Compare the real-time static pressure value with the preset static pressure target value and calculate the static pressure deviation value.

[0162] Step 1062: Perform trend analysis on the static pressure deviation value to determine the direction and magnitude of change of the static pressure deviation value in multiple consecutive sampling periods, and generate trend analysis results.

[0163] Step 1063: Based on the absolute value of the static pressure deviation and the trend analysis results, determine the expected adjustment range and direction for adjusting the power amplification factor of the drive circuit.

[0164] Step 1064: Calculate the target power amplification factor for this adjustment by combining the power amplification factor of the drive circuit after the previous adjustment, as well as the expected adjustment range and direction.

[0165] Step 1065: Generate a corresponding pulse width modulation signal based on the target power amplification factor.

[0166] Step 1066: Output the pulse width modulation signal to the driving circuit, and control the driving circuit to amplify the input power according to the target power amplification factor to obtain the adjusted power amplification factor.

[0167] In this step, the power amplification factor refers to the gain factor when the drive circuit amplifies the input standard control signal. It is used to determine the final power applied to the wind turbine motor and is obtained by adjusting the gain of the power amplifier in the control circuit.

[0168] The adjusted power amplification factor refers to the latest gain factor obtained after calculation in the current control cycle, which is used to actually drive the wind turbine and match the wind turbine power with the current system requirements.

[0169] The real-time static pressure value refers to the specific value of the static pressure in the air supply duct that is read from the pressure sensor at the current moment. It is used to compare with the target value and is obtained through real-time sampling and analog-to-digital conversion of the sensor signal.

[0170] The preset static pressure target value refers to the ideal static pressure value that is desired to be maintained in the air supply duct according to the design and operation requirements of the fresh air system, and is used as a reference benchmark for the control system adjustment.

[0171] Static pressure deviation refers to the algebraic difference between the real-time static pressure value and the preset static pressure target value, which is used to quantify the gap between the current pressure state and the ideal state.

[0172] Trend analysis results refer to qualitative or semi-quantitative judgments on the direction and speed of change of static pressure deviation over a continuous period of time. They are used to predict future pressure change trends and are obtained through statistical analysis of historical static pressure deviation values.

[0173] The expected adjustment magnitude and direction refer to the magnitude and direction of the change in the power amplification factor calculated by the control algorithm, and whether it should be increased or decreased. This is used to guide specific control actions and is calculated by combining the magnitude and trend of the static pressure deviation value.

[0174] The target power amplification factor refers to the specific gain value that the drive circuit is expected to achieve after control calculation, which is used to generate the final power control signal.

[0175] Pulse width modulation (PWM) signal refers to an electronic signal that controls the average output power by adjusting the duty cycle of the pulse signal conduction time within a fixed period. It is used to precisely control the conversion of digital control quantities into analog power and is generated by a microcontroller or dedicated chip according to the target power amplification factor.

[0176] In this step, firstly, an analog-to-digital comparator or software subtractor is used to subtract the read real-time static pressure value from the preset static pressure target value to calculate the static pressure deviation value; secondly, a time series analysis method is used to perform linear regression analysis on the historical sequence of static pressure deviation values ​​in the most recent consecutive sampling periods to calculate the slope and mean of the change, thereby generating trend analysis results describing the direction and magnitude of the change.

[0177] Next, the absolute value of the static pressure deviation and the trend analysis results are input into a predefined control rule lookup table. This predefined control rule lookup table is indexed by the deviation magnitude and trend, and outputs the corresponding control parameters. The lookup operation determines the expected adjustment direction and specific expected adjustment range for increasing or decreasing the power amplification factor. Then, through arithmetic operations, the power amplification factor after the previous adjustment of the drive circuit is added to or subtracted from the expected adjustment range determined this time, according to the direction, to calculate the target power amplification factor for this time.

[0178] Then, a digital pulse width modulation generator linearly maps the target power amplification factor to a corresponding target pulse duty cycle based on a pre-stored proportional conversion coefficient, and generates a pulse width modulation signal with a fixed frequency and corresponding duty cycle accordingly. Finally, the pulse width modulation signal is sent to the control terminal of the drive circuit through the digital output port. The drive circuit adjusts the conduction time of its power switch according to the duty cycle of this signal, thereby amplifying the input power according to the target power amplification factor. At this time, the actual gain of the circuit is the adjusted power amplification factor.

[0179] For example, following the specific implementation of the previous step, firstly, based on the issued air volume command, the fresh air system of office area A begins to adjust and simultaneously starts to stabilize the pipeline pressure; secondly, the real-time static pressure value measured by the pressure sensor is read and compared with the preset target value to obtain the static pressure deviation value; then, the change sequence of the static pressure deviation value over a recent period is analyzed to determine its trend; then, based on the current deviation magnitude and the identified trend, the preset control strategy is queried to determine the direction and magnitude of the adjustment to the drive circuit gain; then, based on the previous gain value, this adjustment is applied to calculate a new target gain value; subsequently, this target gain value is converted into a pulse width modulation signal with a corresponding duty cycle; finally, this pulse width modulation signal is sent to the motor driver, and the drive circuit operates with the new gain accordingly, thereby adjusting the fan power and causing the pipeline static pressure to tend towards the target value.

[0180] Step 107: Control the operation of the fresh air system fan according to the initial air volume command and the adjusted power amplification factor.

[0181] Optionally, step 107 may specifically include:

[0182] Step 1071: Obtain the initial air volume command and parse the target air volume value corresponding to each fresh air outlet in the target area from the initial air volume command.

[0183] Step 1072: Calculate the target power value required to drive the fan of the fresh air system based on the adjusted power amplification factor.

[0184] Step 1073: Based on the target air volume value corresponding to the fresh air outlet and the fan characteristic curve of the fresh air system, determine the control signal parameters required to drive the opening degree of the corresponding air valve of the fresh air outlet.

[0185] Step 1074: Combine the target power value with the control signal parameters to generate a combined control command for driving the fan motor and the corresponding air valve.

[0186] Step 1075: Output the synthetic control command to the motor driver and damper actuator of the fresh air system to control the fan to run according to the target power value and synchronously adjust the dampers of each fresh air outlet to the target opening degree.

[0187] In this step, fan operation refers to the physical process in which the fan motor rotates at a specific speed and power under the control of the drive circuit to deliver air, thereby achieving fresh air supply to the target area.

[0188] The target air volume value refers to the specific air volume that is expected to be achieved for each fresh air outlet in the generated initial air volume command, and is used to precisely control the air supply capacity of each outlet.

[0189] The target power value refers to the total air volume required to drive the fan, calculated based on the adjusted power amplification factor, and is the power value that needs to be supplied to the fan motor to control the main drive energy of the fan.

[0190] A fan characteristic curve is a mathematical model or data chart that describes the relationship between the air volume, static pressure and required input power of a specific fan model at different speeds. It is used to determine appropriate control parameters based on the target air volume and is obtained through performance data provided by the fan manufacturer or experimental calibration.

[0191] Control signal parameters refer to the specific signal values ​​required to control the action of the damper actuator, which can accurately correspond to a specific damper opening degree. They are used to convert the target air volume value into a drive command for the damper. They are calculated by looking up the correspondence between the damper opening degree and the signal value and combining it with the fan characteristic curve.

[0192] A composite control command is a complex data structure or signal that contains the target power value information required to drive the fan motor and the control signal parameter information required to control each air valve. It is used to control the fan and air valves simultaneously and in a coordinated manner. It is obtained by encapsulating the target power value and multiple control signal parameters according to a predetermined protocol.

[0193] A motor driver is a power electronic device that receives control commands and converts them into three-phase AC or DC power suitable for driving the fan motor. It is used to precisely control the speed and power of the fan and is achieved through power electronic conversion technology.

[0194] A damper actuator is an electromechanical device installed at each fresh air outlet, which receives electrical signals and drives the damper blades to rotate to adjust the opening. It is used to precisely control the air volume of each outlet and is achieved through a stepper motor or servo motor combined with a gear mechanism.

[0195] In this step, firstly, by using data packet parsing or memory reading technology, the identifier of each fresh air outlet and its corresponding target air volume value are extracted from the received initial air volume command data structure; secondly, by using multiplication, the adjusted power amplification factor is multiplied by a preset system reference power value to calculate the target power value required to drive the fan of the entire fresh air system.

[0196] Next, for the target air volume value of each air outlet, a lookup table and interpolation algorithm is used for processing. With the target air volume value as input, the pre-stored fan characteristic curve database is queried to find the valve opening required to achieve the target air volume value under the expected operating conditions. Then, according to the calibration mapping table between the valve opening and the control signal, the control signal parameters required to drive the valve to the target opening, such as a specific pulse width or voltage value, are determined.

[0197] Then, through data encapsulation and protocol framing technology, the target power value and the control signal parameters of all air outlets are combined and packaged according to a predefined communication protocol format to generate a structured synthetic control command. Finally, this synthetic control command is simultaneously sent to the motor driver and each damper actuator via a fieldbus or digital-analog output interface. The motor driver adjusts its power output according to the target power value in the command to control the fan operation; each damper actuator drives the valve plate to move according to the corresponding control signal parameters in the command, adjusting the damper to the target opening, thereby achieving coordinated operation of the fan and dampers.

[0198] For example, following the specific implementation of the previous step, the initial air volume command and the adjusted power amplification factor are first obtained, and the final control begins. Next, the target air volume value for each outlet is parsed from the initial air volume command; for example, outlet F1 is 430 cubic meters per hour, and F2 is 50 cubic meters per hour. Then, the adjusted power amplification factor is multiplied by the system base power to calculate the target power value required to drive the fan. Then, for each target air volume value, the pre-stored fan characteristic curve is consulted to determine the required valve opening to achieve the target air volume value. Based on the calibration relationship between the opening and the control signal, the corresponding control signal parameters are obtained. These parameters and the target power value are then encapsulated into a synthetic control command according to a predetermined communication protocol. Finally, this command is simultaneously sent to the motor driver and each valve actuator via the bus, driving the fan to operate at the target power and adjusting each valve to the target opening, thereby collaboratively achieving precise air delivery.

[0199] Figure 2 This application provides a structural schematic diagram of a fresh air purification and energy-saving system based on indoor environmental monitoring, as shown below. Figure 2 As shown, the system includes:

[0200] The acquisition module 21 is used to acquire the video stream captured by the image acquisition device, the ambient sound signal captured by the sound acquisition array, and the static pressure feedback signal in the air supply duct within the target area.

[0201] Analysis module 22 is used to perform time delay analysis on the environmental sound signals acquired by the sound acquisition array, locate the spatial position of the sound source in the target area, and generate the sound intensity distribution of the target area in combination with the amplitude of the sound signal;

[0202] The recognition module 23 is used to perform moving target segmentation and morphological recognition on the video stream and generate a personnel distribution map reflecting the real-time location of personnel within the target area;

[0203] The extrapolation module 24 is used to correlate the sound intensity distribution with the personnel distribution map in a time sequence, extrapolate the aggregation trend of each partition of the target area within a preset time period, and form a predictive personnel density distribution map;

[0204] Output module 25 is used to input the predictive population density distribution map into the trained network model to output the initial air volume command corresponding to each fresh air outlet in the target area.

[0205] The adjustment module 26 is used to dynamically adjust the power amplification factor of the drive circuit according to the static pressure feedback signal, so as to obtain the adjusted power amplification factor.

[0206] The control module 27 is used to control the operation of the fan of the fresh air system according to the initial air volume command and the adjusted power amplification factor.

[0207] Figure 2 The aforementioned fresh air purification and energy-saving system based on indoor environmental monitoring can perform... Figure 1 The implementation principle and technical effects of the fresh air purification and energy-saving method based on indoor environmental monitoring described in the above embodiments will not be repeated here. The specific operation methods of each module and unit in the fresh air purification and energy-saving system based on indoor environmental monitoring in the above embodiments have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0208] In one possible design, Figure 2 The illustrated embodiment of a fresh air purification and energy-saving system based on indoor environmental monitoring can be implemented as a computing device, such as... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0209] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0210] The processing component 32 is used for the above Figure 1 The embodiment describes a fresh air purification and energy-saving method based on indoor environmental monitoring.

[0211] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0212] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A fresh air purification and energy-saving method based on indoor environmental monitoring, characterized in that, include: Acquire video streams captured by image acquisition devices within the target area, as well as ambient sound signals captured by sound acquisition arrays and static pressure feedback signals within air supply ducts within the target area; The ambient sound signal acquired by the sound acquisition array is subjected to time delay analysis to locate the spatial position of the sound source within the target area, and the sound intensity distribution of the target area is generated by combining the amplitude of the sound signal. The video stream is segmented into moving targets and its shape is recognized to generate a personnel distribution map reflecting the real-time location of people within the target area; By correlating the sound intensity distribution with the population distribution map in a time sequence, the aggregation trend of each zone in the target area within a preset time period is deduced, and a predictive population density distribution map is formed. The predictive population density distribution map is input into the trained network model to output the initial air volume command corresponding to each fresh air outlet in the target area; The power amplification factor of the drive circuit is dynamically adjusted based on the static pressure feedback signal to obtain the adjusted power amplification factor. The operation of the fresh air system's fan is controlled based on the initial air volume command and the adjusted power amplification factor.

2. The method according to claim 1, characterized in that, Perform time delay analysis on the ambient sound signals acquired by the sound acquisition array to locate the spatial position of the sound source within the target area, and generate the sound intensity distribution of the target area by combining the amplitude of the sound signals, including: Receives multiple audio signals collected by audio acquisition units located at different physical locations within the target area; For the multiple sound signals, based on the time difference between the arrival of the sound signals at any two sound acquisition units and the speed of sound propagation in the air, the distance difference from the potential sound source point to any two sound acquisition units is determined. Based on the multiple distance differences, a system of equations is constructed regarding the spatial location of potential sound source points, and the system of equations is solved to obtain the spatial coordinates of at least one sound source within the target area. For each sound source obtained by the solution, select a sound acquisition unit that is closest to the sound source, and obtain the signal amplitude corresponding to the sound source in the sound signal acquired by the sound acquisition unit. Based on the spatial coordinates of each sound source and the corresponding signal amplitude, the energy contribution of the sound source to each preset grid point in the target area is calculated; The energy contributions of all sound sources to the same grid point are superimposed to obtain the sound intensity distribution of the target area.

3. The method according to claim 1, characterized in that, Perform moving target segmentation and morphological recognition on the video stream to generate a personnel distribution map reflecting the real-time location of personnel within the target area, including: Extract any two consecutive video frames from the video stream, compare the difference in brightness values ​​of corresponding pixels in the two video frames, and mark pixels whose brightness value changes exceed a preset change threshold as changed pixels. Spatial connectivity analysis is performed on the changed pixels to group multiple spatially adjacent changed pixels into an independent motion region. For each independent motion region, the shape contour features representing the independent motion region are extracted from the current video frame. The shape contour features of the independent motion region are compared with the pre-stored human body contour shape feature set to determine whether the shape contour of the independent motion region conforms to human features. Independent motion regions whose shape contours conform to human features are determined as valid human body regions. Calculate the geometric center coordinates of the effective personnel area in the current video frame, and use the geometric center coordinates as the real-time position of the personnel within the effective personnel area; Based on the real-time locations of all valid personnel within the target area, a personnel distribution map reflecting the real-time locations of personnel within the target area is generated.

4. The method according to claim 1, characterized in that, By sequentially associating the sound intensity distribution with the population distribution map, the aggregation trend of each zone in the target area within a preset time period is deduced, and a predictive population density distribution map is formed, including: Obtain sound intensity distribution and personnel distribution maps corresponding to the current moment and multiple consecutive historical moments to form a historical sequence; The historical sequence was analyzed to identify the temporal and spatial correspondence between high-value areas of sound intensity distribution and densely populated areas on the population distribution map, and to establish a correlation mapping between the sound activity distribution pattern of the high-value areas and the population location distribution pattern of the densely populated areas. Based on the aforementioned association mapping, sound feature regions related to personnel gathering activities are extracted; Based on the historical sequence, calculate the number of people in each partition within the target area at the current moment, and combine this with the sound intensity value of the sound feature area at the current moment to infer the intensity level of gathering activity in each partition within the target area at the current moment. The intensity level of gathering activity in each partition at the current moment, the number of people in each partition at the current moment, and the rate of change of the number of people in each partition in the historical sequence are used as a comprehensive feature vector to describe the current gathering state. The comprehensive feature vector is input into a pre-trained state inference model. Based on the changes in the number of people in subsequent periods corresponding to similar comprehensive feature vectors in the historical sequence, the predicted number of people in each partition of the target area after a preset period is inferred. Based on the predicted number of people in each zone, a predictive population density distribution map is generated.

5. The method according to claim 1, characterized in that, The predictive population density distribution map is input into the trained network model to output initial airflow commands corresponding to each fresh air outlet in the target area, including: The predicted number of people in each partition represented by the predictive population density distribution map is mapped onto a grid map corresponding to the physical space of the target area to generate a gridded density data map. The gridded density data map is processed by channelization, with the predicted number of people at each grid point as an independent numerical channel and the relative position coordinates of the grid point in its respective partition as an independent spatial channel, so as to generate a multi-channel feature map containing numerical and spatial channels. The multi-channel feature map is input into the first processing branch of the trained network model to extract multi-scale neighborhood information from the spatial channels in the multi-channel feature map and generate a first intermediate feature map. The gridded density data map is input into the second processing branch of the trained network model to extract global statistical features from the gridded density data map and generate a second intermediate feature map that reflects the overall population distribution in the target area. The first intermediate feature map and the second intermediate feature map are concatenated along the feature dimension, and the concatenated combined feature map is subjected to nonlinear transformation and information compression to obtain a weight distribution map containing the wind volume demand weights of each location in the target area. According to the weight distribution diagram, the preset total system air volume requirement is allocated according to the weight of each location, and the theoretical air volume value corresponding to each fresh air outlet is calculated. Based on the theoretical air volume value corresponding to each fresh air outlet, and combined with the duct resistance characteristics corresponding to each outlet, the theoretical air volume value is adjusted to generate an initial air volume command.

6. The method according to claim 1, characterized in that, The power amplification factor of the drive circuit is dynamically adjusted based on the static pressure feedback signal to obtain the adjusted power amplification factor, including: The real-time static pressure value is compared with the preset static pressure target value to calculate the static pressure deviation value; Perform trend analysis on the static pressure deviation value to determine the direction and magnitude of change of the static pressure deviation value in multiple consecutive sampling periods, and generate trend analysis results; Based on the absolute value of the static pressure deviation and the trend analysis results, the expected adjustment range and direction for adjusting the power amplification factor of the drive circuit are determined. Based on the power amplification factor of the drive circuit after the previous adjustment, as well as the expected adjustment range and direction, calculate the target power amplification factor for this adjustment; Generate a corresponding pulse width modulation signal based on the target power amplification factor; The pulse width modulation signal is output to the driving circuit, which controls the driving circuit to amplify the input power according to the target power amplification factor to obtain the adjusted power amplification factor.

7. The method according to claim 1, characterized in that, Based on the initial air volume command and the adjusted power amplification factor, control the operation of the fresh air system's fan, including: Obtain the initial air volume command, and parse the target air volume value corresponding to each fresh air outlet in the target area from the initial air volume command; Based on the adjusted power amplification factor, calculate the target power value required to drive the fan of the fresh air system. Based on the target air volume value corresponding to the fresh air outlet and the fan characteristic curve of the fresh air system, determine the control signal parameters required to drive the opening degree of the air valve corresponding to the fresh air outlet; By combining the target power value with the control signal parameters, a combined control command is generated to drive the fan motor and the corresponding air valve. The synthesized control command is output to the motor driver and damper actuator of the fresh air system to control the fan to operate according to the target power value and synchronously adjust the dampers of each fresh air outlet to the target opening degree.

8. A fresh air purification and energy-saving system based on indoor environmental monitoring, characterized in that, include: The acquisition module is used to acquire video streams captured by the image acquisition device, ambient sound signals captured by the sound acquisition array, and static pressure feedback signals in the air supply duct within the target area. The analysis module is used to perform time delay analysis on the environmental sound signals acquired by the sound acquisition array, locate the spatial position of the sound source in the target area, and generate the sound intensity distribution of the target area by combining the amplitude of the sound signal. The recognition module is used to perform moving target segmentation and morphological recognition on the video stream and generate a personnel distribution map reflecting the real-time location of personnel within the target area; The extrapolation module is used to correlate the sound intensity distribution with the personnel distribution map in a time sequence, extrapolate the aggregation trend of each partition of the target area within a preset time period, and form a predictive personnel density distribution map; The output module is used to input the predictive population density distribution map into the trained network model to output the initial air volume command corresponding to each fresh air outlet in the target area. The adjustment module is used to dynamically adjust the power amplification factor of the drive circuit according to the static pressure feedback signal, so as to obtain the adjusted power amplification factor. The control module is used to control the operation of the fresh air system's fan according to the initial air volume command and the adjusted power amplification factor.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a fresh air purification and energy-saving method based on indoor environmental monitoring as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The device contains a computer program that, when executed by a computer, implements a fresh air purification and energy-saving method based on indoor environmental monitoring as described in any one of claims 1 to 7.

Images