A method and system for intelligent management and control of environmental data

By collecting and processing image difference data in dust detection technology and updating the model with the actual measured values ​​of dust sensors, the problem of inaccurate acquisition of dust degree in the existing technology is solved, and the accuracy of dust degree at each location in the area is achieved. Accurate control of measurement and dust suppression equipment.

CN118692029BActive Publication Date: 2024-12-03SHAANXI FENGHUA TIMES ENVIRONMENTAL ENG CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202411170623.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-26
Publication Date
2024-12-03
Estimated Expiration
2044-08-26

AI Technical Summary

Technical Problem

The existing dust detection technology is difficult to accurately obtain the dust level at various locations in the area, which makes it difficult to accurately control the dust suppression equipment.

Method used

By adopting the intelligent control method of environmental protection data, the absolute value of the real-time image and standard image difference value in the control area is collected, and the trained control model is input to output the dust prediction diagram, and the actual measured value and preset loss of the dust sensor are combined for back propagation. Update the management model to improve the accuracy of the prediction graph.

Benefits of technology

It realizes accurate acquisition of dust levels at various locations in the control area, improves the accuracy of dust prediction maps, and ensures accurate control of dust suppression equipment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118692029B_ABST
    Figure CN118692029B_ABST
Patent Text Reader

Abstract

The present application relates to the technical field of dust detection, and in particular to a method and system for intelligent management and control of environmental data, the method comprising: collecting real-time images within the control area, calculating the absolute value of the difference between the real-time image and the standard image at each position point, and obtaining a difference map; inputting the difference map into the trained control model to output a dust prediction map, the dust prediction map including the dust prediction values ​​at each position point within the control area; deploying dust sensors at key positions in the control area, and in response to the absolute value of the difference between the dust prediction value at the key position point and the dust measured value collected by the dust sensor being greater than a preset error, back-propagating the control model using a preset loss. Through the technical solution of the present application, the degree of dust at each position point within the control area can be accurately obtained.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of dust detection, and in particular to a method and system for intelligent management and control of environmental protection data. Background Art

[0002] Dust control and dust control have always been an important part of environmental governance. In the production activities of many industries such as construction, electricity, coal chemical industry, etc., dust may be generated both indoors and outdoors, causing environmental pollution and affecting the health of workers.

[0003] At present, the patent application document with application publication number CN112528968A discloses a dust detection method and system for urban governance, which includes: obtaining a video data stream and a sensor data stream of a fixed scene, and preprocessing the video data stream and the sensor data stream; obtaining a dust map, comparing the dust map with the picture in the video data stream, and identifying the dust map; obtaining multiple sensor data streams, comparing them with the preset dust threshold, and the data stream above the dust threshold is the dust data stream; using the dust map and the dust data stream as samples, inputting them into a pre-trained classifier to detect whether there is dust, and transmitting the detection results to the server, and visually displaying them on the host computer. The pre-trained classifier uses an SVM classifier.

[0004] The above method inputs the video data stream in a fixed scene and the sensor data stream collected by the sensor as input information to a pre-trained classifier to determine whether there is dust in the fixed scene. However, in the process of dust control, dust suppression equipment such as fog cannons are required to spray water mist towards the area where dust exists to achieve the purpose of dust suppression. In order to accurately control the dust suppression equipment, only judging whether there is dust in a fixed scene cannot meet the requirements. Therefore, how to accurately obtain the dust level at each location in an area is a problem that needs to be solved urgently. Summary of the invention

[0005] In order to solve the technical problem of inaccurate surface defect detection results, the present application provides an environmental data intelligent management and control method and system, which can accurately obtain the dust level at each location in an area.

[0006] In a first aspect, the present application provides an intelligent management and control method for environmental data, the management and control method comprising: collecting real-time images within a management and control area, calculating the absolute value of the difference between the real-time image and the standard image at each position point, and obtaining a difference map; inputting the difference map into a trained management and control model to output a dust prediction map, the dust prediction map comprising dust prediction values ​​for each position point within the management and control area; deploying dust sensors at key positions within the management and control area, and in response to the absolute value of the difference between the dust prediction value at the key position point and the actual dust value collected by the dust sensor being greater than a preset error, back-propagating the management and control model using a preset loss, the preset loss Satisfies the relationship: ; The key position points in the dust prediction map The dust forecast value is Key location points The measured value of dust emission is is the number of key location points, is the position point in the dust prediction map The local Moran index, The location point in the historical dust forecast map The average local Moran index, is the total number of locations within the control area.

[0007] The accuracy of the dust prediction map is judged by the actual dust measured values ​​collected by dust sensors at key locations. When the accuracy of the dust prediction map is low, the preset loss is calculated based on the correlation of dust levels at different locations in the control area in the spatial dimension and the actual dust measured values ​​at key locations, and the control model is updated again to improve the accuracy of the dust prediction map.

[0008] In one embodiment, the training method of the control model includes: collecting difference map samples and sample labels, the sample labels including dust label values ​​of key position points; inputting the difference map samples into the control model of this iteration to obtain a first output result, and taking the absolute value of the difference between the dust prediction value of the key position point and the dust label value in the first output result as the full supervision loss; inputting the difference map samples into the control model of the previous iteration to obtain a second output result, and calculating the absolute value of the difference between the second output result and the dust prediction value of the non-key position point in the first output result as the unsupervised loss; performing weighted summation of the full supervision loss and the unsupervised loss to obtain a semi-supervised loss to update the control model and obtain a control model of the next iteration; performing multiple iterations on the control model until the semi-supervised loss is less than a preset loss or the number of iterations is greater than a preset number, and then stopping the training.

[0009] The actual dust values ​​collected by dust sensors at key locations are used as dust label values, and the management and control model is trained using an unsupervised training method to constrain the management and control model to learn the mapping relationship between the difference map and the dust prediction map, ensuring that the management and control model can output accurate dust prediction maps.

[0010] In one embodiment, the semi-supervised loss satisfies the relationship:

[0011] ,in is the difference map sample The fully supervised loss is is the difference map sample The unsupervised loss is is the adjustment coefficient, is the difference map sample The semi-supervised loss.

[0012] In one embodiment, the adjustment coefficient is positively correlated with the number of historical iterations.

[0013] Since the dust label value is the actual value collected, the fully supervised loss can provide accurate and effective supervision information; with the increase of the number of historical iterations, under the constraint of the fully supervised loss, the more accurate the mapping relationship between the difference map and the dust prediction map learned by the control model, that is, the more accurate the second output result in the unsupervised loss, the more effective supervision information the unsupervised loss can provide. At this time, the adjustment coefficient is increased so that the control model can fully learn the mapping relationship between the difference map and the dust prediction map.

[0014] In one embodiment, the adjustment coefficient is negatively correlated with the full supervision loss.

[0015] Since the dust label value is the actual value collected, the fully supervised loss can provide accurate and effective supervision information; the fully supervised loss can intuitively reflect the accuracy of the second output result. The smaller the fully supervised loss, the more accurate the second output result, and the more effective supervision information the unsupervised loss can provide. At this time, the adjustment coefficient is increased so that the control model can fully learn the mapping relationship between the difference map and the dust prediction map.

[0016] In one embodiment, the location point The average local Moran index Satisfies the relationship:

[0017] ,in, For the Location points in the historical dust forecast map The local Moran index, is the number of historical dust prediction maps.

[0018] In one embodiment, the location point The average local Moran index Satisfies the relationship:

[0019] ,in, For the Location points in the historical dust forecast map The local Moran index, is the number of historical dust forecast maps, For the The prediction error of the historical dust forecast map, It is the sum of the prediction errors of all historical dust prediction maps; wherein the prediction error is the maximum absolute value of the difference between the dust prediction value and the dust measured value at the key location point.

[0020] Different historical dust forecast maps have different prediction errors. In order to average the local Moran index Can accurately reflect the location point The correlation with the dust degree at other locations is used to assign a smaller weight coefficient to the historical dust forecast map with a larger prediction error, and the local Moran index of all historical dust forecast maps is weighted summed.

[0021] In the second aspect of the present application, there is also provided an environmental data intelligent management and control system, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an environmental data intelligent management and control method according to the first aspect of the present application is implemented.

[0022] The technical solution of this application has the following beneficial technical effects:

[0023] Collect real-time images within the control area, calculate the absolute value of the difference between the real-time image and the standard image at each position point, and obtain a difference map. The standard image is the image information within the control area when there is no dust. The difference map can reflect the difference between the standard pixel value of each position point in the control area at the current moment and the standard image; input the difference map into the trained control model to output a dust prediction map, which includes the dust prediction value of each position point in the control area; the dust degree of each position point in the control area can be accurately obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become easy to understand. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0025] Figure 1 is a flow chart of a method for intelligent management and control of environmental data according to an embodiment of the present application;

[0026] Figure 2 is a flow chart of a training method for a control model according to an embodiment of the present application;

[0027] Figure 3 is a flow chart of an environmental data intelligent management and control method according to another embodiment of the present application;

[0028] Figure 4 It is a structural block diagram of an environmental data intelligent management and control system according to an embodiment of the present application. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0030] It should be understood that when the terms "first", "second", etc. are used in the claims, specification and drawings of the present application, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the specification and claims of the present application indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections.

[0031] According to the first aspect of the present application, the present application provides a method for intelligent management and control of environmental data. Figure 1 1 is a flow chart of an environmental data intelligent management and control method according to an embodiment of the present application. Figure 1 As shown, the environmental data intelligent management and control method 100 includes steps S101 to S102, which are described in detail below.

[0032] S101, collecting a real-time image within a controlled area, calculating the absolute value of the difference between each position point of the real-time image and a standard image, and obtaining a difference map.

[0033] In one embodiment, the control area can be an indoor area such as a production workshop or factory, or an outdoor area such as a road or industrial park, and this application does not impose any restrictions. A real-time image is collected within the control area, and the real-time image includes the real-time pixel value of each position point within the control area; the standard image includes the standard pixel value of each position point within the control area when there is no dust, and the absolute value of the difference between the real-time pixel value and the standard pixel value of each position point is calculated to obtain a difference map, which can reflect the difference between the standard pixel value of each position point in the control area at the current moment, and provide a data basis for obtaining the dust prediction value of each position point.

[0034] S102, inputting the difference map into a trained control model to output a dust prediction map, wherein the dust prediction map includes dust prediction values ​​for each location point in the control area.

[0035] In one embodiment, the difference map is used as the input of the control model to output a dust prediction map. The control model includes an image encoder and an image decoder. The image encoder performs multiple downsampling on the difference map to obtain a feature map; the image decoder performs multiple upsampling on the feature map to output a dust prediction map; the dust prediction map and the difference map have the same size, and the dust prediction map includes dust prediction values ​​for each location point in the control area.

[0036] The number of times of downsampling performed in the image encoder is the same as the number of times of downsampling performed in the image decoder, thereby ensuring that the dust prediction map and the difference map have the same size. The implementer can adopt the network structure of existing networks such as autoencoder network and U-Net network.

[0037] In one embodiment, in order to enable the control model to output an accurate dust prediction map, the control model needs to be trained to constrain the control model to learn the mapping relationship between the difference map and the dust prediction map. Figure 2 , which is a flow chart of a training method for a control model according to an embodiment of the present application, wherein the training method 200 for a control model includes steps S201 to S205.

[0038] S201, collecting difference map samples and sample labels, where the sample labels include dust label values ​​at key locations.

[0039] Among them, multiple dust sensors are deployed in the control area, and the location points where the dust sensors are deployed are key location points, and the location points other than the key location points are recorded as non-key points; it can be understood that the union of key location points and non-key location points can cover the entire control area.

[0040] Difference map samples are collected within the control area, and the actual dust measurement values ​​of the dust sensors at each key position point when collecting the difference map samples are used as the dust label values ​​of the corresponding key position points to obtain the sample labels corresponding to the difference map samples.

[0041] S202, input the difference map sample into the control model of this iteration to obtain the first output result, and use the absolute value of the difference between the dust prediction value and the dust label value of the key position point in the first output result as the full supervision loss.

[0042] Among them, the full supervision loss satisfies the relationship: ,in is the difference map sample The key position points in the first output result The dust forecast value is is the difference map sample Key location points The measured value of dust emission is is the number of key position points in the difference map sample, is the difference map sample The corresponding fully supervised loss.

[0043] S203, input the difference map sample into the control model of the previous iteration to obtain a second output result, and calculate the absolute value of the difference between the dust prediction value of the non-critical position point in the second output result and the first output result as the unsupervised loss.

[0044] Specifically, the difference map sample is input into the control model of the previous iteration to obtain the second output result, and the dust prediction value of the non-critical position point in the second output result is used as the pseudo label of the non-critical position point. It can be understood that the absolute value of the difference between the second output result and the dust prediction value of the non-critical position point in the first output result is calculated as the unsupervised loss, that is, the absolute value of the difference between the dust prediction value of the non-critical position point in the first output result and the pseudo label is calculated as the unsupervised loss.

[0045] For non-critical position points, the difference map sample is input into the management and control model of the previous iteration to obtain a second output result, which includes the dust prediction value of each non-critical position point in the difference map sample, and the dust prediction value is used as the pseudo-label corresponding to the non-critical position point in this iteration; in other words, in this iteration, one critical position point corresponds to one dust label value in the sample label, and one non-critical position point corresponds to one pseudo-label.

[0046] The unsupervised loss satisfies the relationship: ,in is the difference map sample The non-critical position points in the first output result The dust forecast value is is the difference map sample The non-critical position points in the second output result The dust prediction value (i.e. pseudo label), is the number of non-critical position points in the difference map sample, is the difference map sample The corresponding unsupervised loss.

[0047] S204, performing weighted summation on the full supervised loss and the unsupervised loss to obtain a semi-supervised loss, so as to update the control model and obtain a control model for the next iteration.

[0048] In one embodiment, the semi-supervised loss satisfies the relationship:

[0049] ,in is the difference map sample The fully supervised loss is is the difference map sample The unsupervised loss is is the adjustment coefficient, is the difference map sample The semi-supervised loss. The adjustment coefficient is positively correlated with the number of historical iterations. Specifically, the adjustment coefficient Satisfies the relationship: ,in is the number of historical iterations.

[0050] In another embodiment, the adjustment coefficient is negatively correlated with the full supervision loss. Specifically, the adjustment coefficient Satisfies the relationship: ,in is the difference map sample The fully supervised loss.

[0051] The semi-supervised loss is used to back-propagate the control model, and the model parameters of the image encoder and image decoder in the control model are adjusted to update the control model. The updated control model is the control model for the next iteration.

[0052] S205, performing multiple iterations on the control model until the semi-supervised loss is less than a preset loss or the number of iterations is greater than a preset number, and then stopping the training.

[0053] Among them, the preset loss value is 0.01, and the preset number of times is 5000.

[0054] In this way, the dust sensors deployed in the control area collect the actual dust values ​​at key locations and use them as the dust label values ​​of the key locations to train the control model. This allows the control model to learn the mapping relationship between the pixel values ​​in the difference map and the actual dust values, thereby improving the accuracy of the dust prediction map.

[0055] It should be noted that, during the training process of the control model, a difference map sample corresponds to a sample label, the key position points in the sample label correspond to the dust label value, and the non-key position points have no dust label value. That is to say, some position points in the difference map sample have label information, while some position points have no label information. The training process of the control model is a semi-supervised training process. In addition to the above-mentioned control model training method 200, the control model can also be trained by a consistency-based method.

[0056] In another embodiment, the difference map between the real-time image and the standard image at any time is calculated, and the difference map is input into the trained control model to obtain the dust prediction map at the corresponding time, wherein the dust prediction map includes the dust prediction values ​​of each position point in the control area. In order to ensure that the control model can continuously output accurate dust prediction maps, it is necessary to use dust sensors deployed at key locations to monitor the accuracy of the dust prediction map in real time. Figure 3 , is a flow chart of an environmental data intelligent management and control method according to another embodiment of the present application. The environmental data intelligent management and control method 300 includes steps S301 to S303, which are described in detail below.

[0057] Specifically, the specific process of step S301 is the same as step S101, and the specific process of step S302 is the same as step S102, which will not be repeated here. After obtaining the dust prediction map in step S302, the control method further includes:

[0058] S303, deploying dust sensors at key locations in the control area, and in response to the absolute value of the difference between the dust prediction value at the key location and the dust measured value collected by the dust sensor being greater than a preset error, back-propagating the control model using a preset loss.

[0059] Among them, the preset loss Satisfies the relationship: ; The key position points in the dust prediction map The dust forecast value is Key location points The measured value of dust emission is is the number of key location points, is the position point in the dust prediction map The local Moran index, The location point in the historical dust forecast map The average local Moran index, is the total number of locations within the control area.

[0060] Among them, the absolute value of the difference between the dust prediction value at the key position point and the dust measurement value collected by the dust sensor can represent the error size of the dust prediction map. The larger the absolute value of the difference, the less accurate the dust prediction map. When the absolute value of the difference is greater than the preset error, the preset loss is calculated, and the control model is back-propagated to update the control model again to ensure the accuracy of the dust prediction map. The value of the preset error is 2.

[0061] It should be noted that the number of key position points in the control area is one or more. When the number of key position points in the control area is multiple, the absolute value of the difference between the dust prediction value and the dust measured value at each key position point is calculated, and the maximum value of the absolute value of the difference is compared with the preset error. If the maximum value of the absolute value of the difference is greater than the preset error, the preset loss is calculated, the control model is back-propagated, and the control model is updated again.

[0062] In one embodiment, there is a certain correlation between the dust levels at different locations in the control area, that is, there is a correlation between the dust level at one location and the dust level at other locations. The local Moran index can be used to characterize the correlation between the dust level at one location and other locations.

[0063] Specifically, the position point in the dust prediction map The local Moran index Satisfies the relationship:

[0064] ,in, , , ,in, and The location point in the dust prediction map and location points The dust deviation value, and are the position points in the dust prediction map and location points The dust forecast value is is the average value of the dust prediction value in the dust prediction map, The location point in the dust prediction map The dust deviation value, It represents the sum of squares of dust deviation values ​​at all locations in the dust prediction map. The number of all location points in the dust prediction map, For location point and location points The spatial weight of It is the average value of the sum of squares of dust deviation values ​​at all locations in the dust prediction map.

[0065] Among them, in response to the position point and location points Mutually adjacent points, spatial weight The value is 1, in response to the position point and location points Not adjacent to each other, spatial weight The value of is 0.

[0066] In this way, the local Moran's index of each position point in the dust prediction map is obtained. The local Moran's index of a position point is used to reflect the correlation of the dust degree between the position point and its adjacent position points.

[0067] In one embodiment, the correlation between the dust level at one location and the dust level at other locations tends to be stable, and the dust level at the location in the historical dust prediction map is calculated. The average local Moran index can reflect the position point in the stable state. The correlation between the dust level at the location and the dust level at other locations; therefore, the average local Moran index is used as the The standard value of the local Moran index is included in the preset loss In the calculation, the correlation between the dust levels of each location point is constrained in the spatial dimension, so that the control model can output an accurate dust prediction map.

[0068] Specifically, the location point The average local Moran index Satisfies the relationship:

[0069] ,in, For the Location points in the historical dust forecast map The local Moran index, is the number of historical dust prediction maps.

[0070] In another embodiment, since the historical dust prediction map is the output result of the control model in historical time, each historical dust prediction map corresponds to a prediction error, and the prediction error is equal to the maximum absolute value of the difference between the dust prediction value and the dust measured value at the key location point.

[0071] Specifically, the location point The average local Moran index Satisfies the relationship:

[0072] ,in, For the Location points in the historical dust forecast map The local Moran index, is the number of historical dust forecast maps, For the The prediction error of the historical dust forecast map, It is the sum of the prediction errors of all historical dust prediction maps; wherein the prediction error is the maximum absolute value of the difference between the dust prediction value and the dust measured value at the key location point.

[0073] In this way, the accuracy of the dust prediction map is judged based on the dust sensors deployed at key locations. When the accuracy of the dust prediction map is low, the actual dust values ​​collected by the dust sensors at key locations and the local Moran's index of each location point in the dust prediction map are collected to construct a preset loss. The preset loss aims to minimize the difference between the dust prediction value and the actual dust value at the key location and to stabilize the correlation between the dust levels of each location point. The control model is updated to improve the accuracy of the dust prediction map.

[0074] The technical principles and implementation details of the environmental protection data intelligent management and control method of the present application are introduced above through specific embodiments. Real-time images within the control area are collected, and the absolute value of the difference between the real-time image and the standard image at each position point is calculated to obtain a difference map. The standard image is the image information within the control area when there is no dust. The difference map can reflect the difference between the standard pixel value of each position point in the control area at the current moment and the standard image; the difference map is input into the trained control model to output a dust prediction map, which includes the dust prediction value of each position point in the control area; the dust degree of each position point in the control area can be accurately obtained.

[0075] According to the second aspect of the present application, the present application also provides an intelligent management and control system for environmental data. Figure 4 is a structural block diagram of the environmental protection data intelligent management and control system according to an embodiment of the present application. Figure 4 As shown, the system 50 includes a processor and a memory, and the memory stores computer program instructions. When the computer program instructions are executed by the processor, an environmental data intelligent management and control method according to the first aspect of the present application is implemented. The system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, whose settings and functions are known in the art, so they are not described here.

[0076] In the present application, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in this application may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0077] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0078] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the patent application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent application shall be subject to the attached claims.

Claims

1. A method for intelligent management and control of environmental data, characterized in that: The control method includes: Collecting a real-time image within the control area, calculating the absolute value of the difference between the real-time image and the standard image at each position point, and obtaining a difference map; Inputting the difference map into the trained control model to output a dust prediction map, wherein the dust prediction map includes dust prediction values ​​for each location point within the control area; Dust sensors are deployed at key locations in the control area. In response to the absolute value of the difference between the dust prediction value at the key location and the dust measurement value collected by the dust sensor being greater than a preset error, the control model is back-propagated using a preset loss. Satisfies the relationship: ; The key position points in the dust prediction map The dust forecast value is Key location points The measured value of dust emission is is the number of key location points, is the position point in the dust prediction map The local Moran index, The location point in the historical dust forecast map The average local Moran index of is the total number of location points within the control area; The difference map can reflect the difference between the standard pixel value of each location point in the control area at the current moment and the standard image.

2. The method for intelligent management and control of environmental data according to claim 1, characterized in that: The training method of the control model includes: Collect difference map samples and sample labels, where the sample labels include dust label values ​​at key locations; Input the difference map sample into the control model of this iteration to obtain the first output result, and use the absolute value of the difference between the dust prediction value of the key position point in the first output result and the dust label value as the full supervision loss; input the difference map sample into the control model of the previous iteration to obtain the second output result, and calculate the absolute value of the difference between the dust prediction value of the non-key position point in the second output result and the first output result as the unsupervised loss; perform weighted summation of the full supervision loss and the unsupervised loss to obtain the semi-supervised loss, so as to update the control model and obtain the control model of the next iteration; The control model is iterated multiple times until the semi-supervised loss is less than a preset loss or the number of iterations is greater than a preset number, and then the training is stopped.

3. The method for intelligent management and control of environmental data according to claim 2, characterized in that: The semi-supervised loss satisfies the relationship: ,in is the difference map sample The fully supervised loss is is the difference map sample The unsupervised loss is is the adjustment coefficient, is the difference map sample The semi-supervised loss.

4. The method for intelligent management and control of environmental data according to claim 3, characterized in that: The adjustment coefficient is positively correlated with the number of historical iterations. Satisfies the relationship: ,in is the number of historical iterations.

5. The method for intelligent management and control of environmental data according to claim 3, characterized in that: The adjustment coefficient is negatively correlated with the full supervision loss. Satisfies the relationship: ,in is the difference map sample Fully supervised loss.

6. The method for intelligent management and control of environmental data according to claim 1, characterized in that: Location Point The average local Moran index Satisfies the relationship: ,in, For the Location points in the historical dust forecast map The local Moran index, is the number of historical dust prediction maps.

7. The method for intelligent management and control of environmental data according to claim 1, characterized in that: Location Point The average local Moran index Satisfies the relationship: ,in, For the Location points in the historical dust forecast map The local Moran index, is the number of historical dust forecast maps, For the The prediction error of the historical dust forecast map, It is the sum of the prediction errors of all historical dust prediction maps; wherein the prediction error is the maximum absolute value of the difference between the dust prediction value and the dust measured value at the key location point.

8. An intelligent environmental data management and control system, characterized in that: It comprises a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an environmental protection data intelligent management and control method according to any one of claims 1 to 7 is implemented.

Citation Information

Patent Citations

  • Flying dust detection method and system applied to urban governance

    CN112528968A

  • Model training method, system and device based on generative adversarial network and medium

    CN117095216A

  • Intelligent planting method for agricultural operation robot

    CN117882546A