Intelligent sewage identification method and system for smart city

Abstract

The application provides a sewage intelligent identification method and system for smart city. By using deep learning technology, more complex and deep features can be extracted from sewage image data, which can more accurately identify and predict the sewage relationship between urban areas than traditional identification methods. Moreover, a large amount of sewage image data can be processed in real time to generate urban area sewage relationship estimation data, which has high application value for environmental protection tasks that require real-time monitoring and processing. The generated urban area sewage relationship estimation data can help environmental protection departments understand the sewage discharge characteristics and mutual relationship between various urban areas, so as to develop more effective sewage treatment strategies, optimize resource allocation, and improve sewage treatment efficiency. In addition, cross-domain sample sewage image data and local sample sewage image data are used for training, so that the model can adapt to different regions and different types of sewage identification and prediction tasks.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart city, in particular to a sewage intelligent identification method and system for smart city. BACKGROUND

[0002] With the acceleration of industrialization and urbanization, water pollution problems are becoming increasingly serious. Especially in large urban areas, due to the large population, frequent industrial activities, and large sewage discharge, sewage treatment has become an important environmental problem. Traditional sewage treatment methods usually rely on manual sampling, laboratory analysis, and other methods to identify sewage types and pollution levels, and then determine specific treatment strategies based on the results. This method is time-consuming, inefficient, and highly dependent on human resources.

[0003] In recent years, the development of image processing and deep learning technology has provided a new solution for sewage treatment. Through the analysis of sewage image data, the type and pollution level of sewage can be quickly and accurately identified, which helps environmental protection departments to develop more effective treatment strategies. However, current deep learning models mostly focus on single type of sewage identification, rarely considering the relationship between sewage in urban areas. In addition, how to effectively use deep learning technology to extract useful information from a large amount of sewage image data and generate urban area sewage relationship estimation data is still a problem to be solved.

[0004] Therefore, developing a method that can accurately identify and predict the relationship between sewage in urban areas to improve the decision-making efficiency of environmental protection departments has become a pressing problem to be solved. SUMMARY

[0005] Therefore, the purpose of the present application is to provide a sewage intelligent identification method and system for smart city.

[0006] According to the first aspect of the present application, a sewage intelligent identification method for smart city is provided, which is applied to a sewage intelligent identification system for smart city, and the method comprises:

[0007] Obtaining a sample sewage image data set, the sample sewage image data set comprising cross-domain sample sewage image data and local sample sewage image data;

[0008] Based on all the cross-domain sample sewage image data, updating the parameters of the preset cross-domain sewage feature extraction network to generate a target cross-domain sewage feature extraction network, and determining the corresponding cross-domain sewage feature vector of the target cross-domain sewage feature extraction network;

[0009] updating parameters of the preset sewage feature extraction network based on all the sewage image data in the current domain, to generate a target sewage feature extraction network in the current domain, and determining a sewage feature vector corresponding to the target sewage feature extraction network in the current domain;

[0010] optimizing target network function information in the preset sewage correlation estimation network based on all the cross-domain sewage feature vectors and all the current-domain sewage feature vectors, to generate a target sewage correlation estimation network;

[0011] obtaining a target sewage image data sequence; the target sewage image data sequence includes at least one target sewage image data cluster of a city area, and each target sewage image data cluster of a city area includes cross-domain sewage image data of the city area and current-domain sewage image data of the city area;

[0012] loading the target sewage image data sequence into the target sewage correlation estimation network to generate city area sewage correlation estimation data.

[0013] In a possible implementation of the first aspect, after the loading of the target sewage image data sequence into the target sewage correlation estimation network to generate city area sewage correlation estimation data, the method further includes:

[0014] determining at least one first target city area corresponding to the city area sewage correlation estimation data;

[0015] For each first target city area, analyzing sewage feature correlation data of the first target city area based on a target sewage image data cluster of the first target city area, and each sewage feature correlation data of the first target city area includes a sewage feature path corresponding to the first target city area;

[0016] obtaining sewage source flow direction data of each first target city area in a set time domain interval;

[0017] For each first target city area, generating a sewage discharge scheduling strategy of the first target city area based on the sewage feature correlation data of the first target city area and the sewage source flow direction data of the first target city area in the set time domain interval, and performing a sewage discharge scheduling action corresponding to the sewage discharge scheduling strategy of the first target city area on the first target city area based on the sewage discharge scheduling strategy of the first target city area.

[0018] In a possible implementation of the first aspect, after the loading of the target sewage image data sequence into the target sewage correlation estimation network to generate city area sewage correlation estimation data, the method further includes:

[0019] generate a sewage cross-region situation value of each of the urban areas based on the sewage relationship estimation data of the urban areas, and extract one or more second target urban areas from all of the urban areas based on the sewage cross-region situation value of each of the urban areas, wherein the sewage cross-region situation value of the second target urban areas is not greater than a threshold set value;

[0020] generate a sewage flow direction vector of each of the second target urban areas based on the target sewage image data cluster of each of the second target urban areas;

[0021] generate sewage discharge guidance information based on the sewage flow direction vector of all of the second target urban areas, and send the sewage discharge guidance information to all of the remaining urban areas except all of the second target urban areas among all of the urban areas.

[0022] In a possible implementation of the first aspect, the generating, for each of the first target urban areas, of the sewage discharge scheduling strategy of the first target urban area based on the sewage characteristic relationship data of the first target urban area and the sewage source flow direction data of the first target urban area in the set time domain interval comprises:

[0023] the generating, for each of the target urban areas, of the sewage simulation constraint model of the target urban area based on the sewage characteristic relationship data of the target urban area and the sewage source flow direction data of the target urban area in the set time domain interval;

[0024] the generating, for each of the target urban areas, of the sewage discharge scheduling strategy of the target urban area based on the sewage simulation constraint model of the target urban area.

[0025] In a possible implementation of the first aspect, after the optimizing of the target network function function information in the preset sewage correlation estimation network based on all of the cross-region sewage characteristic vectors and all of the local sewage characteristic vectors to generate a target sewage relationship estimation network, the method further comprises:

[0026] determining a function coefficient convergence factor of each network function function information in the target sewage relationship estimation network and a function coefficient convergence value of each network function function information;

[0027] calculating a network convergence value of the target sewage relationship estimation network according to the function coefficient convergence factor of each network function function information in the target sewage relationship estimation network and the function coefficient convergence value of each network function function information;

[0028] analyzing whether the network convergence value of the target sewage relationship estimation network is less than a set network convergence value;

[0029] If the network convergence value of the target sewage relationship estimation network is less than the set network convergence value, extracting a to-be-updated function coefficient from all the network function function information in the target sewage relationship estimation network, and determining a convergence factor optimization weight of each to-be-updated function coefficient;

[0030] For each to-be-updated function coefficient, based on the convergence factor optimization weight of the to-be-updated function coefficient, optimizing the function coefficient convergence factor of the to-be-updated function coefficient to optimize the to-be-updated function coefficient, and re-triggering the operation of calculating the network convergence value of the target sewage relationship estimation network according to the function coefficient convergence factor of each network function function information in the target sewage relationship estimation network and the function coefficient convergence value of each network function function information, and analyzing whether the network convergence value of the target sewage relationship estimation network is less than a set network convergence value.

[0031] In a possible implementation of the first aspect, after the target sewage image data sequence is obtained, the method further includes:

[0032] Performing sewage phenomenon label prediction on all the target sewage image data included in the obtained target sewage image data sequence to generate a sewage phenomenon label of each target sewage image data, the sewage phenomenon label including a cross-domain sewage phenomenon label or an in-domain sewage phenomenon label;

[0033] According to the sewage phenomenon label of each target sewage image data, for each target sewage image data, assigning a network learning attribute to the target sewage image data so that the target sewage image data has a corresponding network learning attribute;

[0034] Performing feature cleaning on all the target sewage image data to optimize all the target sewage image data and trigger the operation of loading the target sewage image data sequence into the target sewage relationship estimation network to generate urban area sewage relationship estimation data.

[0035] In a possible implementation of the first aspect, the optimization of the target network function function information in the preset sewage relationship estimation network based on all the cross-domain sewage feature vectors and all the in-domain sewage feature vectors to generate a target sewage relationship estimation network includes:

[0036] optimizing target network function information in a preset sewage correlation estimation network based on all the cross-domain sewage feature vectors and all the in-domain sewage feature vectors, to generate a basic sewage relationship estimation network;

[0037] extracting at least one target sample sewage image data from the sample sewage image data set;

[0038] For each target sample sewage image data, loading the target sample sewage image data into the basic sewage relationship estimation network, generating a basic learning result corresponding to the target sample sewage image data, calculating an error parameter between the target sample sewage image data and the basic learning result corresponding to the target sample sewage image data, and generating a learning cost corresponding to the target sample sewage image data;

[0039] According to the learning cost corresponding to each target sample sewage image data, generating a network learning cost of the basic sewage relationship estimation network;

[0040] analyzing whether the network learning cost converges, and when it is analyzed that the network learning cost converges, determining the basic sewage relationship estimation network as a target sewage relationship estimation network;

[0041] When it is analyzed that the network learning cost does not converge, re-triggering the extraction of at least one target sample sewage image data from the sample sewage image data set, for each target sample sewage image data, loading the target sample sewage image data into the basic sewage relationship estimation network, generating a basic learning result corresponding to the target sample sewage image data, calculating an error parameter between the target sample sewage image data and the basic learning result corresponding to the target sample sewage image data, and generating a learning cost corresponding to the target sample sewage image data; according to the learning cost corresponding to each target sample sewage image data, generating a network learning cost of the basic sewage relationship estimation network; analyzing whether the network learning cost converges.

[0042] According to a second aspect of the present application, a sewage intelligent identification system for smart city is provided, which comprises a machine readable storage medium and a processor, the machine readable storage medium stores machine executable instructions, and the processor, when executing the machine executable instructions, implements the sewage intelligent identification method for smart city as described above.

[0043] According to a third aspect of the present application, a computer readable storage medium is provided, which stores computer executable instructions, and when the computer executable instructions are executed, the sewage intelligent identification method for smart city as described above is implemented.

[0044] According to any one of the above aspects, in the present application, more complex and deep features can be extracted from sewage image data by using deep learning technology, which can more accurately identify and predict the sewage relationship between urban areas than traditional identification methods. The generated urban area sewage relationship estimation data can help environmental protection departments understand the sewage discharge characteristics and mutual relationship between various urban areas, so as to develop more effective sewage treatment strategies, optimize resource allocation, and improve sewage treatment efficiency. In addition, cross-domain sample sewage image data and in-domain sample sewage image data are used for training, so that the model has better generalization ability and can adapt to different regions and different types of sewage identification and prediction tasks. Since deep learning technology is used, this method can process a large amount of sewage image data in real time to generate urban area sewage relationship estimation data, which has high application value for environmental protection tasks that need real-time monitoring and processing. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0046] Figure 1 A flowchart of the sewage intelligent identification method for smart city provided by the embodiments of the present application is shown.

[0047] Figure 2 The component structure schematic diagram of the sewage intelligent identification system for smart city provided by the embodiments of the present application for realizing the sewage intelligent identification method for smart city described above is shown. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be described clearly and completely below according to the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of illustration and description, and do not limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn to scale. The flowchart shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can not be implemented in sequence, and the steps without logical context relationship can be reversed in sequence or implemented simultaneously. In addition, one or more other operations can be added to the flowchart or one or more operations can be deleted from the flowchart under the guidance of the content of the present application.

[0049] In addition, the described embodiments are only some embodiments of the present application, not all embodiments. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. According to the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor correspond to the scope of protection of the present application.

[0050] Figure 1 The flowchart of the sewage intelligent identification method and system for smart city provided by the embodiments of the present application is shown. It should be understood that in other embodiments, the order of some steps of the sewage intelligent identification method for smart city of the present embodiment can be shared according to actual needs, or some steps can be omitted or maintained. The detailed steps of the sewage intelligent identification method for smart city include:

[0051] Step S110, obtaining a sample sewage image data set, wherein the sample sewage image data set includes cross-domain sample sewage image data and local sample sewage image data.

[0052] For example, a large amount of sewage image data can be collected from various places. Among them, the cross-domain sample data may include sewage samples of different regions, and the local sample data may focus on sewage samples of a specific urban area, such as A1 area of A city.

[0053] Step S120, based on all the cross-domain sample sewage image data, updating the parameters of the preset cross-domain sewage feature extraction network to generate a target cross-domain sewage feature extraction network, and determining the corresponding cross-domain sewage feature vector of the target cross-domain sewage feature extraction network.

[0054] For example, a neural network model is trained and optimized using sewage image data from multiple geographic regions, so as to effectively identify and extract features of cross-regional sewage.

[0055] Step S130, based on all the local sample sewage image data, updating the parameters of the preset local sewage feature extraction network to generate a target local sewage feature extraction network, and determining the corresponding local sewage feature vector of the target local sewage feature extraction network.

[0056] For example, similarly, another neural network model is trained and optimized using sewage image data only for a certain region, so as to accurately identify and extract features of sewage in the certain region.

[0057] Step S140, based on all the cross-domain sewage feature vectors and all the local-domain sewage feature vectors, the target network function information in the pre-set sewage correlation estimation network is optimized to generate the target sewage relationship estimation network.

[0058] For example, there is a sewage correlation estimation network for predicting and estimating the sewage discharge feature relationship between different regions. For example, the sewage feature vectors extracted from the cross-domain and local-domain can be used to train and optimize this model, so that it can more accurately predict the sewage discharge features of different regions and their mutual influence.

[0059] Step S150, obtaining a target sewage image data sequence. The target sewage image data sequence includes at least one city area target sewage image data cluster, and each city area target sewage image data cluster includes cross-domain sewage image data of the city area and local-domain sewage image data of the city area.

[0060] Step S160, loading the target sewage image data sequence into the target sewage relationship estimation network to generate city area sewage relationship estimation data.

[0061] For example, when a new city, such as A, is to be monitored for sewage, sewage image data of A can be collected, including sewage image data of various urban areas. Finally, the collected sewage image data of A is input into the optimized sewage correlation estimation model, and the model outputs predicted sewage feature relationship data, which can help understand the sewage discharge features and mutual relationships between various urban areas in A, and thus develop more accurate sewage treatment strategies.

[0062] For example, city area sewage relationship estimation data can include the following data:

[0063] 1. Sewage flow prediction: For example, the data may show that a portion of the sewage from the commercial area eventually flows into the sewer system of the residential area, while a portion flows into the nearby river.

[0064] 2. Sewage composition influence: For example, the sewage discharged by the industrial area may contain certain harmful substances, which may enter other areas through the sewer system and affect the water quality there.

[0065] 3. Sewage discharge estimation: For example, the data may show that the sewage discharge of certain areas (such as residential areas) will significantly increase during certain time periods (such as the morning rush hour).

[0066] 4. Inter-regional sewage exchange: For example, the data may reveal the degree of sewage exchange between different regions, such as between residential areas and commercial areas, or between industrial areas and other regions.

[0067] These information can help the city management department better understand the sewage discharge situation of the city and develop more effective sewage treatment strategies. For example, if the data shows that the sewage of the industrial area has a greater impact on the water quality of the residential area, the city management department may need to consider improving the sewage treatment facilities of the industrial area or re-planning the sewer system of the city to reduce the impact of industrial area sewage on other areas.

[0068] Based on the above steps, by using deep learning technology, more complex and deep features can be extracted from sewage image data, which can more accurately identify and predict the sewage relationship between urban areas than traditional identification methods. The generated urban area sewage relationship estimation data can help environmental protection departments understand the sewage discharge characteristics and mutual relationship between various urban areas, so as to develop more effective sewage treatment strategies, optimize resource allocation, and improve sewage treatment efficiency. In addition, cross-domain example sewage image data and in-domain example sewage image data are used for training, so that the model has better generalization ability and can adapt to different regions and different types of sewage identification and prediction tasks. Since deep learning technology is used, this method can process a large amount of sewage image data in real time to generate urban area sewage relationship estimation data, which has high application value for environmental protection tasks that require real-time monitoring and processing.

[0069] In one possible implementation, after step S160, the method further includes:

[0070] Step A110, determining at least one first target urban area corresponding to the urban area sewage relationship estimation data.

[0071] Step A120, for each first target urban area, based on the target sewage image data cluster of the first target urban area, analyzing the sewage feature relationship data of the first target urban area, and the sewage feature relationship data of each first target urban area includes the sewage feature path corresponding to the first target urban area.

[0072] Step A130, obtaining the sewage source flow direction data of each first target urban area within a set time domain interval.

[0073] Step A140, for each first target urban area, based on the sewage feature relationship data of the first target urban area and the sewage source flow direction data of the first target urban area within the set time domain interval, generating the sewage discharge scheduling strategy of the first target urban area, and based on the sewage discharge scheduling strategy of the first target urban area, performing the sewage discharge scheduling action corresponding to the sewage discharge scheduling strategy of the first target urban area on the first target urban area.

[0074] For example, according to the previous prediction results, the region A1 and the region A2 of the city A are determined as the target city regions, because the sewage discharge characteristics and the mutual influence degree of the two regions are more significant. For the region A1 and the region A2, further analysis is performed on the sewage characteristic relationship data thereof. For example, it can be found that the industrial sewage of the region A1 flows to the region A2 through the sewer system and contains a certain specific harmful substance. For example, the sewage source flow data of the region A1 and the region A2 in the past year are collected, which can help better understand the sewage flow of the two regions. Based on the above data analysis, the following scheduling strategies can be proposed: one is to optimize the industrial sewage treatment facilities of the region A1 to reduce the content of the harmful substance; and the other is to modify the sewer system between the two regions to reduce the inflow of the sewage of the region A1 to the region A2. Then, the strategies are submitted to the city management department for them to perform the specific modification and optimization work.

[0075] In one possible implementation, after step S160, the method further includes:

[0076] Step B110, generating a sewage cross-region situation value of each of the city regions based on the city region sewage relationship estimation data, and extracting one or more second target city regions from all the city regions based on the sewage cross-region situation value of each of the city regions, wherein the sewage cross-region situation value of the second target city region is not greater than a threshold set value.

[0077] Step B120, generating a sewage flow direction vector of each of the second target city regions based on the target sewage image data cluster of each of the second target city regions.

[0078] Step B130, generating sewage discharge guidance information based on the sewage flow direction vectors of all the second target city regions, and sending the sewage discharge guidance information to all the remaining city regions except all the second target city regions among all the city regions.

[0079] For example, based on the predicted sewage relationship estimation data of each region A, the sewage cross-region situation value of each region can be calculated. This value may represent the degree of influence of the sewage of the region on other regions. A threshold value is set, and if the sewage cross-region situation value of a certain region is lower than the threshold value, it is selected as the second target city region. For example, if it is found that the sewage cross-region situation values of region A3 and region A4 are low, they are selected as the second target city regions. Next, the sewage image data of region A3 and region A4 is analyzed to generate the sewage flow vector of the two regions. This vector may represent where the sewage comes from (source) and where it goes (flow direction). Then, based on the obtained sewage flow vector of region A3 and region A4, it may be found that if other regions can adjust their sewage discharge direction or method, the impact on region A3 and region A4 can be effectively reduced. Therefore, they generate a sewage discharge guidance information and send it to other regions (such as region A1 and region A2) to guide them to adjust the sewage discharge strategy to reduce the impact on region A3 and region A4.

[0080] In a possible implementation, step A140 can include:

[0081] Step A141, for each of the target city regions, based on the sewage characteristic relationship data of the target city region and the sewage source flow direction data of the target city region in the set time domain interval, generating a sewage simulation constraint model of the target city region.

[0082] For example, for region A1 and region A2, two sewage simulation constraint models are established using their sewage characteristic relationship data and sewage source flow direction data in the past year. These models can simulate the sewage discharge of each region and its impact on other regions.

[0083] Step A142, for each of the target city regions, based on the sewage simulation constraint model of the target city region, generating a sewage discharge scheduling strategy of the target city region.

[0084] For example, for region A1, the sewage discharge of some factories may be limited in some time period to reduce the impact on region A2; for region A2, more sewage treatment facilities may be recommended to handle the sewage from region A1. These strategies are calculated and inferred based on the simulation constraint model, aiming to minimize the impact of sewage discharge on the environment.

[0085] In a possible implementation, after step S140, the method further includes:

[0086] Step S141, determine the function coefficient convergence factor of each network function function information in the target sewage relationship estimation network, and the function coefficient convergence value of each network function function information.

[0087] For example, after training and optimizing the target sewage relationship estimation network, the function coefficient convergence factor of each network function (i.e. each layer of the network or certain specific nodes) will be checked, which can reflect the speed of parameter update of the function in the training process; At the same time, they also determine the function coefficient convergence value of each network function, which represents the parameter value when the network training reaches a stable state.

[0088] Step S142, according to the function coefficient convergence factor of each network function function information in the target sewage relationship estimation network and the function coefficient convergence value of each network function function information, calculate the network convergence value of the target sewage relationship estimation network.

[0089] For example, according to the function coefficient convergence factor and the function coefficient convergence value mentioned above, the convergence value of the whole network is calculated, which can measure the stability of the whole network training and the generalization ability of the model.

[0090] Step S143, analyze whether the network convergence value of the target sewage relationship estimation network is less than the set network convergence value.

[0091] Step S144, if the network convergence value of the target sewage relationship estimation network is less than the set network convergence value, extract the function coefficient to be updated from all the network function function information in the target sewage relationship estimation network, and determine the convergence factor optimization weight of each function coefficient to be updated.

[0092] Step S145, for each function coefficient to be updated, based on the convergence factor optimization weight of the function coefficient to be updated, optimize the function coefficient convergence factor of the function coefficient to be updated to optimize the function coefficient to be updated, and retrigger the operation of calculating the network convergence value of the target sewage relationship estimation network according to the function coefficient convergence factor of each network function function information in the target sewage relationship estimation network and the function coefficient convergence value of each network function function information, and analyzing whether the network convergence value of the target sewage relationship estimation network is less than the set network convergence value.

[0093] For example, if the calculated network convergence value is less than the preset network convergence value, it means that the network training has reached a stable state, and the model can be considered to have good generalization ability.

[0094] After the conditions in the previous step are met, the function coefficients that need further optimization will be extracted, and their convergence factor optimization weights will be determined. These weights can be set according to factors such as the location of the function coefficients in the network, their role, etc.

[0095] Then, according to the convergence factor optimization weights obtained in the previous step, the function coefficient convergence factors of the function coefficients to be updated are optimized. The purpose of this step is to adjust these function coefficients so that the network converges faster, thereby improving the training efficiency and accuracy of the model. Finally, after the above optimization steps, the entire process is restarted until the network convergence value no longer decreases, i.e., the network training has reached a new stable state. At this point, the training and optimization of the target sewage relationship estimation network is complete.

[0096] In one possible implementation, after step S150, the method further comprises:

[0097] Step S151, sewage phenomenon label prediction is performed on all target sewage image data included in the target sewage image data sequence obtained, and sewage phenomenon labels of each target sewage image data are generated, which include cross-domain sewage phenomenon labels or domain-specific sewage phenomenon labels.

[0098] Step S152, according to the sewage phenomenon label of each target sewage image data, the network learning attribute corresponding to each target sewage image data is assigned to each target sewage image data, so that the target sewage image data has a corresponding network learning attribute.

[0099] Step S156, feature cleaning is performed on all target sewage image data to optimize all target sewage image data and trigger the operation of loading the target sewage image data sequence into the target sewage relationship estimation network to generate urban area sewage relationship estimation data.

[0100] For example, after collecting sewage image data of each district A, they use the trained sewage recognition model to label these images, which may label "industrial sewage", "domestic sewage" and other labels. These labels are divided into cross-domain sewage phenomenon labels (such as "pollutant content exceeds standard", which may occur in multiple regions) and domain-specific sewage phenomenon labels (such as "serious river pollution in region A1", which is specific to a certain region).

[0101] Then, according to the predicted sewage phenomenon labels, the research team will assign corresponding network learning attributes to each sewage image data. These attributes may include sewage type, pollution degree, etc., which will be used for subsequent network training and optimization.

[0102] Then, all the sewage image data will be cleaned of features, for example, they may exclude some poor quality or not expected data, or normalize the data, etc., to improve the effect of model training.

[0103] Finally, after the above steps, the prepared sewage image data will be loaded into the target sewage relationship estimation network, starting training and generating urban area sewage relationship estimation data. These data can help environmental protection departments understand the sewage discharge characteristics and mutual relationship between various urban areas in Shanghai, and then develop more accurate sewage treatment strategies.

[0104] In one possible implementation, step A140 can include:

[0105] Step C110, based on all the cross-domain sewage feature vectors and all the local sewage feature vectors, optimizing the target network function function information in the preset sewage correlation estimation network to generate a basic sewage relationship estimation network.

[0106] For example, according to the cross-domain sewage feature vectors (such as excessive pollutant content, etc.) and the local sewage feature vectors (such as the type of sewage in a specific area, etc.) extracted from the sewage image data of each district in the sample area, the preset sewage correlation estimation network is trained and optimized to generate a basic sewage relationship estimation network.

[0107] Step C120, extracting at least one target sample sewage image data from the sample sewage image data set.

[0108] For example, a part of a sample data set containing a large number of different types of sewage images may be selected as target sample sewage image data.

[0109] Step C130, for each target sample sewage image data, loading the target sample sewage image data into the basic sewage relationship estimation network, generating the basic learning result corresponding to the target sample sewage image data, calculating the error parameter between the target sample sewage image data and the basic learning result corresponding to the target sample sewage image data, and generating the learning cost corresponding to the target sample sewage image data.

[0110] For example, input these target sample sewage image data into the basic sewage relationship estimation network to obtain the basic learning result of each sample, and then calculate the error between the predicted result and the actual result of each sample as the learning cost of the sample.

[0111] Step C140, generating the network learning cost of the basic sewage relationship estimation network according to the learning cost corresponding to each target sample sewage image data.

[0112] For example, the learning cost of the entire network can be calculated according to the learning cost of all the samples. This value can be used to measure the prediction performance of the network model.

[0113] Step C150, analyze whether the network learning cost converges, and when it is analyzed that the network learning cost converges, determine the basic sewage relationship estimation network as the target sewage relationship estimation network.

[0114] For example, if the network learning cost has converged (i.e., no longer significantly changes), it means that the training of the network model has reached a stable state, and at this time the basic sewage relationship estimation network can be determined as the target sewage relationship estimation network.

[0115] Step C160, when it is analyzed that the network learning cost does not converge, re-trigger the execution of the extraction of at least one target sample sewage image data from the sample sewage image data set. For each of the target sample sewage image data, load the target sample sewage image data into the basic sewage relationship estimation network, generate the basic learning result corresponding to the target sample sewage image data, calculate the error parameter between the target sample sewage image data and the basic learning result corresponding to the target sample sewage image data, generate the learning cost corresponding to the target sample sewage image data. According to the learning cost corresponding to each of the target sample sewage image data, generate the network learning cost of the basic sewage relationship estimation network. Analyze whether the network learning cost converges.

[0116] If the network learning cost does not converge, the research team needs to re-extract samples from the sample sewage image data set and continue training and optimization until the network learning cost converges.

[0117] Figure 2 A sewage intelligent identification system 100 for smart city that can be used to implement various embodiments described in the present application is schematically shown.

[0118] For one embodiment, Figure 2 A sewage intelligent identification system 100 for smart city is shown, which has one or more processors 102, a control module (chipset) 104 coupled to one or more of the processor(s) 102, a memory 106 coupled to the control module 104, a non-volatile memory (NVY) / storage device 108 coupled to the control module 104, one or more input / output devices 110 coupled to the control module 104, and a network interface 112 coupled to the control module 104.

[0119] The processor 102 can include one or more single core or multicore processors, which can include any combination of general-purpose processors or dedicated processors (e.g., graphics processors, application processors, baseband processors, etc.). In an alternative implementation, the intelligent sewage identification system for smart city 100 can function as a server device such as a gateway as described in embodiments of the present application.

[0120] Figure 2 An intelligent sewage identification system for smart city 100 that can be used to implement various embodiments described in the present application is shown schematically.

[0121] For one embodiment, Figure 2 An intelligent sewage identification system for smart city 100 is shown having one or more processors 102, a control module (chipset) 104 coupled to one or more of the processor(s) 102, a memory 106 coupled to the control module 104, a non-volatile memory (NVM) / storage device 108 coupled to the control module 104, one or more input / output devices 110 coupled to the control module 104, and a network interface 112 coupled to the control module 104.

[0122] The processor 102 can include one or more single core or multicore processors, which can include any combination of general-purpose processors or dedicated processors (e.g., graphics processors, application processors, baseband processors, etc.). In an alternative implementation, the intelligent sewage identification system for smart city 100 can function as a server device such as a gateway as described in embodiments of the present application.

[0123] In an alternative implementation, the intelligent sewage identification system for smart city 100 can include one or more computer readable media (e.g., the memory 106 or the NVM / storage device 108) having instructions 114 and one or more processors 102 in communication with the one or more computer readable media configured to execute the instructions 114 to implement modules to perform the actions described in the present disclosure.

[0124] For one embodiment, the control module 104 can include any suitable interface controllers to provide any suitable interface to one or more of the processor(s) 102 and / or any suitable device or component in communication with the control module 104.

[0125] The control module 104 can include a memory controller module to provide an interface to the memory 106. The memory controller module can be a hardware module, a software module, and / or a firmware module.

[0126] Memory 106 can be used to load and store data and / or instructions 114, for example, for intelligent sewer identification system 100 for smart city. For one embodiment, memory 106 can include any suitable volatile memory, such as suitable DRAM. In an alternative implementation, memory 106 can include double data rate type four synchronous dynamic random access memory (DDR4 SDRAM).

[0127] For one embodiment, control module 104 can include one or more input / output controllers to provide an interface to NVM / storage 108 and input / output device(s) 110.

[0128] For example, NVM / storage 108 can be used to store data and / or instructions 114. NVM / storage 108 can include any suitable non-volatile memory (e.g., flash memory) and / or can include any suitable non-volatile storage device(s) (e.g., one or more hard disk drives (HDDs), one or more compact disk (CD) drives, and / or one or more digital versatile disk (DVD) drives).

[0129] NVM / storage 108 can include storage resources that are physically part of the device on which intelligent sewer identification system 100 for smart city is installed, or it can be accessed by the device remotely and / or be accessible to the device via a network, without necessarily being part of the device. For example, NVM / storage 108 can be accessed by intelligent sewer identification system 100 for smart city via input / output device(s) 110 over a network.

[0130] Input / output device(s) 110 can provide an interface for intelligent sewer identification system 100 for smart city to communicate with any other suitable device(s), which can include communication components, pinyin components, sensor components, etc. Network interface 112 can provide an interface for intelligent sewer identification system 100 for smart city to communicate over one or more networks, which can wirelessly communicate with one or more components of a wireless network according to any of one or more wireless network standards and / or protocols, such as accessing a wireless network according to a communication standard, or a combination thereof.

[0131] For one embodiment, one or more of the processor(s) 102 can be loaded with logic of one or more controllers (e.g., memory controller modules) of the control module 104. For one embodiment, one or more of the processor(s) 102 can be loaded with logic of one or more controllers of the control module 104 to form a system level load. For one embodiment, one or more of the processor(s) 102 can be integrated on the same die as logic of one or more controllers of the control module 104. For one embodiment, one or more of the processor(s) 102 can be integrated on the same die as logic of one or more controllers of the control module 104 to form a system on a chip (SoC).

[0132] In various embodiments, the intelligent sewage identification system for smart city 100 can be, but is not limited to, a server, a desktop computing device, or a mobile computing device (e.g., a laptop computer, a handheld computing device, a tablet computer, a netbook, etc.), and the like. In various embodiments, the intelligent sewage identification system for smart city 100 can have more or less components and / or different architecture. For example, in an alternative implementation, the intelligent sewage identification system for smart city 100 includes one or more cameras, a keyboard, a liquid crystal display (LCD) screen (including touch screen displays), a non- volatile memory port, multiple antennas, a graphics chip, an application specific integrated circuit (ASIC), and a speaker.

[0133] The above describes the embodiments of the present application in detail, and the principles and implementation modes of the present application are described by applying specific examples. The above description of the embodiments is only used to help understand the method of the present application and its core idea; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range will be changed, and the above description of the present application should not be understood as a limitation.

Claims

1. A sewage intelligent identification method for a smart city, characterized in that, The application discloses a sewage intelligent identification system applied to a smart city. An example sewage image data set is acquired, and the example sewage image data set includes cross-domain example sewage image data and local-domain example sewage image data. A preset cross-domain sewage feature extraction network is parameter updated based on all the cross-domain example sewage image data, a target cross-domain sewage feature extraction network is generated, and a cross-domain sewage feature vector corresponding to the target cross-domain sewage feature extraction network is determined. A preset local-domain sewage feature extraction network is parameter updated based on all the local-domain example sewage image data, a target local-domain sewage feature extraction network is generated, and a local-domain sewage feature vector corresponding to the target local-domain sewage feature extraction network is determined. A target network function function information in a preset sewage correlation estimation network is optimized based on all the cross-domain sewage feature vectors and all the local-domain sewage feature vectors, and a target sewage relationship estimation network is generated. A target sewage image data sequence is acquired, and the target sewage image data sequence includes target sewage image data clusters of at least one city area, and each target sewage image data cluster of the city area includes cross-domain sewage image data of the city area and local-domain sewage image data of the city area. The target sewage image data sequence is loaded into the target sewage relationship estimation network, and city area sewage relationship estimation data is generated. 2.The intelligent sewage identification method for smart city according to claim 1, characterized in that, After the target sewage image data sequence is loaded into the target sewage relationship estimation network to generate the city area sewage relationship estimation data, the method further includes the following steps. At least one first target city area corresponding to the city area sewage relationship estimation data is determined. For each first target city area, sewage feature relationship data of the first target city area is analyzed based on a target sewage image data cluster of the first target city area, and the sewage feature relationship data of each first target city area includes a sewage feature path corresponding to the first target city area. Sewage source flow direction data of each first target city area in a set time domain interval is acquired. For each first target city area, a sewage discharge scheduling strategy of the first target city area is generated based on sewage feature relationship data of the first target city area and sewage source flow direction data of the first target city area in the set time domain interval, and a sewage discharge scheduling action corresponding to the sewage discharge scheduling strategy of the first target city area is performed on the first target city area based on the sewage discharge scheduling strategy of the first target city area. 3.The intelligent sewage identification method for smart city according to claim 2, characterized in that, After the target sewage image data sequence is loaded into the target sewage relationship estimation network to generate the city area sewage relationship estimation data, the method further includes the following steps. Based on the city area sewage relationship estimation data, a sewage cross-domain situation value of each city area is generated, and one or more second target city areas are extracted from all the city areas based on the sewage cross-domain situation value of each city area, and the sewage cross-domain situation value of the second target city area is not greater than a threshold set value. generating, for each of the second target urban areas, a sewage flow direction vector of the second target urban area based on the target sewage image data cluster of the second target urban area; generating sewage discharge guidance information based on the sewage flow direction vectors of all the second target urban areas, and sending the sewage discharge guidance information to all the remaining urban areas except all the second target urban areas among all the urban areas. 4.The intelligent sewage identification method for smart city according to claim 2 or 3, characterized in that, The generating, for each of the first target urban areas, of the sewage discharge scheduling strategy of the first target urban area based on the sewage characteristic relationship data of the first target urban area and the sewage source flow direction data of the first target urban area in the set time domain interval comprises: The generating, for each of the target urban areas, of the sewage discharge scheduling strategy of the target urban area based on the sewage characteristic relationship data of the target urban area and the sewage source flow direction data of the target urban area in the set time domain interval comprises: The generating, for each of the target urban areas, of the sewage discharge scheduling strategy of the target urban area based on the sewage characteristic relationship data of the target urban area and the sewage source flow direction data of the target urban area in the set time domain interval comprises: 5.The intelligent sewage identification method for smart city according to claim 4, characterized in that, After the generating of the target sewage relationship estimation network based on all the cross-domain sewage characteristic vectors and all the local-domain sewage characteristic vectors and the optimization of the target network function function information in the preset sewage correlation estimation network, the method further comprises: determining a function coefficient convergence factor of each network function function information in the target sewage relationship estimation network and a function coefficient convergence value of each network function function information; calculating a network convergence value of the target sewage relationship estimation network according to the function coefficient convergence factor of each network function function information in the target sewage relationship estimation network and the function coefficient convergence value of each network function function information; analyzing whether the network convergence value of the target sewage relationship estimation network is less than a set network convergence value; if the network convergence value of the target sewage relationship estimation network is less than the set network convergence value, extracting a to-be-updated function coefficient from all the network function function information in the target sewage relationship estimation network, and determining a convergence factor optimization weight of each to-be-updated function coefficient; for each to-be-updated function coefficient, optimizing a function coefficient convergence factor of the to-be-updated function coefficient based on the convergence factor optimization weight of the to-be-updated function coefficient, so as to optimize the to-be-updated function coefficient, and re-triggering the operation of calculating the network convergence value of the target sewage relationship estimation network according to the function coefficient convergence factor of each network function function information in the target sewage relationship estimation network and the function coefficient convergence value of each network function function information and analyzing whether the network convergence value of the target sewage relationship estimation network is less than the set network convergence value. 6.The intelligent sewage identification method for smart city according to claim 1, wherein, After the obtaining of the target sewage image data sequence, the method further comprises: performing sewage phenomenon label prediction on all target sewage image data included in the target sewage image data sequence to generate a sewage phenomenon label for each target sewage image data, the sewage phenomenon label including a cross-domain sewage phenomenon label or an in-domain sewage phenomenon label; allocating a network learning attribute to each target sewage image data according to the sewage phenomenon label of each target sewage image data, so that the target sewage image data has a corresponding network learning attribute; performing feature cleaning on all target sewage image data to optimize all target sewage image data and trigger the operation of loading the target sewage image data sequence into the target sewage relationship estimation network to generate urban area sewage relationship estimation data. 7.The intelligent sewage identification method for smart city according to claim 1, wherein, optimizing target network function function information in a preset sewage correlation estimation network based on all cross-domain sewage feature vectors and all in-domain sewage feature vectors to generate a target sewage relationship estimation network, including: optimizing target network function function information in a preset sewage correlation estimation network based on all cross-domain sewage feature vectors and all in-domain sewage feature vectors to generate a target sewage relationship estimation network, including: extracting at least one target sample sewage image data from the sample sewage image data set; loading each target sample sewage image data into the basic sewage relationship estimation network to generate a basic learning result corresponding to the target sample sewage image data, calculating an error parameter between the target sample sewage image data and the basic learning result corresponding to the target sample sewage image data, and generating a learning cost corresponding to the target sample sewage image data; generating a network learning cost of the basic sewage relationship estimation network according to the learning cost corresponding to each target sample sewage image data; analyzing whether the network learning cost converges, and determining the basic sewage relationship estimation network as the target sewage relationship estimation network when it is analyzed that the network learning cost converges; when it is analyzed that the network learning cost does not converge, retriggering the operation of extracting at least one target sample sewage image data from the sample sewage image data set, loading each target sample sewage image data into the basic sewage relationship estimation network to generate a basic learning result corresponding to the target sample sewage image data, calculating an error parameter between the target sample sewage image data and the basic learning result corresponding to the target sample sewage image data, and generating a learning cost corresponding to the target sample sewage image data; generating a network learning cost of the basic sewage relationship estimation network according to the learning cost corresponding to each target sample sewage image data; and analyzing whether the network learning cost converges.

8. A sewage intelligent identification system for a smart city, characterized in that, The application discloses a sewage intelligent identification method for a smart city, and comprises a processor and a computer readable storage medium. The computer readable storage medium stores machine executable instructions, and the machine executable instructions are executed by the processor to realize the sewage intelligent identification method for the smart city.

Images