A reverse source tracing analysis method combining map river channel identification and dimensionality reduction calculation

Abstract

The present invention relates to a reverse tracing analysis method combining map river identification and dimensionality reduction calculation, which belongs to the field of pollution detection technology, including: setting a river monitoring sensor, reporting alarm information when pollutants are found; obtaining monitoring point information, river hydrological information and tracing range parameter information; identifying the river network; according to the drawn center line, the map is horizontally cut to obtain the pixel map after cutting, and the pixel map is formed into a file management set; tracing grid concentration; suspected pollution discharge score; display on the map GIS according to the pollution source enterprise score for user research and judgment. The present invention takes a single monitoring point as the source point, generalizes the surface from point to point, uses the map to perform AI grid cutting and sorting of the river, and brings in the reverse tracing algorithm model to reverse the pollution grid concentration, avoiding the low accuracy caused by the calculation singularity of the existing pollution source tracing algorithm and diffusion algorithm, and the tracing error caused by the concealment of pollution discharge.

Description

Technical Field

[0001] The present invention relates to the technical field of pollution detection, and in particular to a reverse source tracing analysis method combining map river channel identification and dimensionality reduction calculation. Background Art

[0002] A common form of water pollution is river pollution. Due to human activities, substances or energy are directly or indirectly introduced into the river environment, causing harmful effects such as damage to river biological resources, damage to river water and river environmental quality, and harm to human health. The degree of pollution changes over time, and it spreads quickly and has a great impact. Although rivers have strong dilution and self-purification capabilities, which are conducive to the diffusion and degradation of pollutants, many large industrial areas and cities in the world are built on riverside areas, and a large amount of wastewater is discharged into the river, causing most rivers to be polluted to varying degrees.

[0003] Controlling river pollution is a complex systematic project with high costs. In addition to using physical, chemical and biological methods to control river pollution, it is also necessary to reduce pollution sources at the source.

[0004] There are currently a variety of pollution source diffusion model products available on the market. Such models have and take into account multi-dimensional parameters, which makes the overall diffusion calculation complex. At the same time, if it is for a reverse tracing scenario, the scenario calculation requires more computing resources.

[0005] Nowadays, with regard to the source tracing of river pollution, pollution emissions are more concealed and unorganized, resulting in normal emissions from pollution source enterprises, no solution and no evidence to trace the source, and a long search chain. Using a simple pollution source method for diffusion calculation, its simulation accuracy and experience speed can no longer achieve the purpose of tracing and governance. It is necessary to carry out model dimensionality reduction calculations in the field of tracing to solve the problem of reverse tracing of river pollutant emissions. Summary of the invention

[0006] In order to solve the above-mentioned problems in the prior art, with regard to how to accurately monitor pollution sources and conduct reverse diffusion tracing, the present invention provides a reverse tracing analysis method that combines map river channel identification and dimensionality reduction calculation.

[0007] A reverse source tracing analysis method combining map river channel identification and dimensionality reduction calculation includes the following steps:

[0008] Step S1, setting up river monitoring sensors, and reporting alarm information when pollutants are found;

[0009] Step S2, obtaining monitoring point information, river hydrological information and source tracing range parameter information;

[0010] Step S3, identifying the river network, obtaining the separated river layer, and then drawing the center line according to the river edge;

[0011] Step S4, according to the drawn center line, the map is horizontally divided to obtain a pixel map after division, and the pixel map is formed into a file management set;

[0012] Step S5, calculating the grid density through the reverse tracing algorithm model, including:

[0013] Step a: Assemble the river channel cross-cutting node information sequence set;

[0014] Step b: Based on the convection model, combined with the scaling factor of each cross-cutting node, a single node simulation is derived;

[0015] Step S6, based on the river channel cross-section node information sequence set and the enterprise river channel outlet information, a list of pollution source enterprises is obtained, and the list of pollution source enterprises is scored for suspected pollution discharge through factor matching;

[0016] Step S7, arranging the scores of the pollution source enterprises from high to low, and displaying them on the GIS map in turn for users to judge.

[0017] Furthermore, the monitoring point information in step S2 includes the geographical information of the alarm point, a set of pollution factors and pollutant concentrations; the river hydrological information includes the current water flow, flow velocity and water level information of the monitoring point; and the tracing range parameter information is the associated river range with the alarm node as the origin and the entire radius of the monitoring point to be traced.

[0018] Furthermore, the step S3 specifically includes the following sub-steps:

[0019] Step S301: using TiandiTu’s self-trained DeepLabv3+ river channel cutting model to identify the river channel and separate the river channel layers;

[0020] Step S302: Draw a center line according to the river edge.

[0021] Furthermore, the drawing of the center line according to the river edge is specifically as follows: graying the river surface, and drawing the center line using the line connecting the midpoints of the linear distances of the second line closest to the positioning point of the left line within the pixel range as the center line.

[0022] Furthermore, the positioning points are configured according to the traceability accuracy requirements with an accuracy of 50 meters.

[0023] Furthermore, the assembling of the river channel cross-cutting node information sequence set in step S5 is specifically as follows: according to the river channel flow direction grid cross-cutting node, the river channel area and factor in the node are calculated, and the river channel cross-cutting node information sequence set is organized with the river channel cross-cutting node as the association;

[0024] Calculate the area of ​​the river channel within the node: According to the river pixels in the river channel, perform pixel point statistics and area conversion to obtain the area;

[0025] Calculate node internal factors: Factors include the average flow velocity of river water, pollutant diffusion coefficient, pollutant sedimentation gradient and average river depth. When calculating node internal factors, basic scaling calculation is performed based on the current node area and the monitoring point area. The scaling factor x = 1-m1 / m0, where m1 is the current node area and m0 is the node area where the monitoring point is located. When calculating the concentration coefficient, a pollutant sedimentation gradient coefficient is added. The formula is: , v is the settling velocity of the particles, g is the acceleration due to gravity, ρ p is the density of the particle, ρ f is the density of the fluid, d is the diameter of the particle, and η is the dynamic viscosity of the fluid.

[0026] Furthermore, the single-node simulation derivation of the convection model includes forward derivation and reverse derivation. In the forward derivation and reverse derivation, the reasoning result of each cross-cutting node is calculated based on the result of the previous node and combined with the correction factor of the node, and finally the simulated pollutant emission concentration value of all nodes is obtained; the simulation derivation formula is:

[0027] ,

[0028] Among them, C is the concentration of pollutants, t is time, u is the average flow velocity of the river, D is the diffusion coefficient of pollutants, and x is the spatial coordinate along the river direction.

[0029] Furthermore, the suspected pollution discharge scoring of step S6 is specifically as follows: three rounds of matching are performed on the pollution source enterprises in the list, and weights are assigned to the matching results of each round, and the final scoring scores are obtained by combining them; the three rounds of matching specifically include:

[0030] First round of matching: Obtain enterprise filing information, match the pollutant factor catalog in the enterprise's planned emissions with the pollutants, and obtain the first round of matching scores;

[0031] Second round of matching: Match the pollutant records of the enterprise’s outlets with the pollutants to obtain the second round of matching scores;

[0032] The third round of matching: Combined with the electricity consumption information of the enterprise's production lines, the pollution control alarms for production and discharge are counted to obtain the third round of matching scores.

[0033] Furthermore, the horizontal segmentation of the map in step S4 is specifically as follows: the river map is segmented by a horizontal line that forms an angle of 90 degrees with the center line.

[0034] The beneficial effects of the present invention are as follows: the present invention takes a single monitoring point as the source point, generalizes the whole from point to point, uses the map to perform AI grid cutting and sorting of the river channel, and brings in the reverse tracing algorithm model to reverse the pollution grid concentration. At the same time, combined with the organized / unorganized emissions of provincial and ministerial enterprises, and the actual electricity production conditions of the production lines, multi-dimensional pollution discharge score screening and presentation are performed to avoid the low accuracy caused by the single calculation of the existing pollution source tracing algorithms and diffusion algorithms, the tracing errors caused by the concealment of pollution discharge, and the problem of no clues in the verification of pollution discharge in other places; at the same time, the algorithm is tailored for the tracing field, the calculation amount is greatly reduced, and computing resources are greatly saved. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a flow chart of the method of the present invention;

[0036] Figure 2 Compile data processing flow chart for river channel identification grid;

[0037] Figure 3 Draw a schematic diagram for the center line;

[0038] Figure 4 Schematic diagram for the derivation of the convection model;

[0039] Figure 5 This is the structure diagram of the DeepLabv3+ river cutting model;

[0040] Figure 6 Create flow charts for river model data sets;

[0041] Figure 7 This is the training flow chart of the river model. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] In this embodiment, refer to Figure 1 and Figure 2 , a reverse source tracing analysis method combining map river channel identification and dimensionality reduction calculation, comprising the following steps:

[0044] Step S101, setting a river monitoring sensor, and reporting an alarm message when pollutants are found.

[0045] Step S102, obtain monitoring point information, river hydrological information and traceability range parameter information; the monitoring point information includes the geographical information of the alarm point, the set of pollution factors and the concentration of pollutants; the river hydrological information includes the current water flow, flow velocity and water level information of the monitoring point, in addition, historical hydrological information can be obtained from government units for reference; the traceability range parameter information is the associated river range with the alarm node as the origin and the entire radius of the monitoring point to be traced.

[0046] Step S103, river network identification;

[0047] 1) River network identification requires the identification of rivers and separation of river layers based on the DeepLabv3+ river cutting model (Google DeepLab image segmentation) trained by Tiandi Map.

[0048] 2) Draw the center line according to the river edge; Figure 3 As shown in the figure, for the river edge, the center line is drawn: the river surface is grayed out, the positioning point of the left line (a line is composed of points), and the pixel distance of the point of the second line closest to the pixel range is the straight-line distance. The positioning point is configured according to the traceability accuracy requirements, and the default accuracy is 50 meters.

[0049] Step S104, river channel grid arrangement;

[0050] According to the drawn center line, the map is divided horizontally (that is, the river map is divided by a horizontal line that forms a 90-degree angle with the center line), and the divided pixel map is obtained, and the pixel map is formed into a file management collection.

[0051] Step S105, tracing grid concentration;

[0052] Reverse tracing algorithm model calculation:

[0053] 1) Assemble the information sequence set of river cross-section nodes: Grid cross-section nodes are identified according to the river flow direction, the river area and factor in the node are calculated, and the river cross-section node information sequence set is organized based on the river cross-section nodes;

[0054] a. Calculate the area of ​​the river channel within the node: According to the river pixels (gray part) in the river channel, perform pixel point statistics and area conversion to obtain the area;

[0055] b. Calculate the internal factors of the node: mainly including the following factors: average flow velocity of river water, pollutant diffusion coefficient, pollutant sedimentation gradient, average river depth; these factors need to be scaled based on the current node area and the monitoring point area, x = 1-m1 / m0, x is the scaling factor, m1 is the current node area, m0 is the node area where the monitoring point is located; it should be noted that when calculating the concentration coefficient, a pollutant sedimentation gradient coefficient is added, the formula is: , v is the settling velocity of the particles (m / s); g is the acceleration of gravity, the default value is 9.81) m / s²; ρ p is the density of the particles (kg / m³), with the sensor parameters as the initial input; ρ f is the density of the fluid, where the pollutant density in the pollutant information table is used as a parameter; d is the diameter of the particle (m); η is the dynamic viscosity of the fluid;

[0056] 2) If Figure 4 As shown, according to the convection model, combined with the scaling factor of each cross-cutting node, a single-node simulation is derived;

[0057] The single-node simulation derivation of the convection model includes forward derivation and reverse derivation. Regardless of forward derivation or reverse derivation, the reasoning result of each cross-cutting node is calculated based on the result of the previous node and combined with the correction factor of this node, and finally the simulated emission concentration value of pollutants of all nodes is obtained;

[0058] In this embodiment, taking forward derivation as an example, the simulation derivation formula of the convection model is:

[0059] ,

[0060] Among them, C is the concentration of pollutants; t is time; u is the average flow velocity of the river; D is the diffusion coefficient of the pollutant, which describes the rate at which the pollutant diffuses due to the concentration gradient; x is the spatial coordinate along the direction of the river.

[0061] Step S106, based on the river channel cross-section node information sequence set and the enterprise river channel outlet information, a list of pollution source enterprises is obtained, and the list of pollution source enterprises is scored for suspected pollution discharge through factor matching;

[0062] The suspected pollution discharge score in step S106 is specifically as follows: three rounds of matching are performed on the pollution source enterprises in the list, and weights are assigned to the matching results of each round to obtain the final score; three rounds of matching are performed, and the branch proportions are calculated. The matching method is as follows:

[0063] First round of matching: Obtain enterprise filing information, match the pollutant factor catalog in the enterprise's planned emissions with the pollutants, and obtain the first round of matching scores;

[0064] Second round of matching: Match the pollutant records of the enterprise’s outlets with the pollutants to obtain the second round of matching scores;

[0065] The third round of matching: Combined with the electricity consumption information of the enterprise's production lines, the pollution control alarms for production and discharge are counted to obtain the third round of matching scores.

[0066] After three rounds of matching, the matching scoring rules are combined with weights to get the final score. The larger the enterprise score, the greater the suspicion.

[0067] Step S107, sort the pollution source enterprise scores from high to low, and display them on the map GIS in sequence for users to judge.

[0068] In this embodiment, if Figure 5 As shown, the Deepalbv3+ model architecture includes:

[0069] Encoder module, using DeepLabv3 as the Encoder module, which gradually reduces feature maps and extracts higher semantic information. In the Encoder structure, it uses Xception as the deep convolutional neural network (DCNN) and uses the ASPP (Atrous Spatial Pyramid Pooling) module, which can extract image features at different scales and improve the segmentation accuracy of the model. In addition, depthwise separable convolution is also used in ASPP and Encoder modules to make the network faster;

[0070] Decoder module,After the Encoder module, DeepLabv3+ adds a simple and effective,Decoder module to gradually restore the spatial information and obtain,clearer segmentation.

[0071] In this embodiment, the training of the river model is to collect and mark materials to form a training set, and then perform multiple rounds of training through tensorflow to finally obtain a recognizable excellent model. The specific operations are as follows:

[0072] 1. Material collection and labeling, the process is as follows Figure 6 ,

[0073] (1) The crawler script collects map blocks within 200 meters of the world map and obtains multiple map blocks.

[0074] (2) Use the labelme tool to batch label and obtain images and annotation json.

[0075] (3) Finally, we get a set of labeled sets (900 sets as a training set).

[0076] 2. Training, the process is as follows Figure 7 ,

[0077] (1) Use a training set (small set).

[0078] (2) Use Tensorflow to set the learning rate, number of iterations, batch size, loss function, snapshot frequency and other information, and use the server's 24GB video memory capacity to perform iterative training.

[0079] (3) During this period, continue to pay attention to the accuracy of iterative feedback and conduct training interventions at any time.

[0080] (4) After this iteration is completed, the model is evaluated and based on the evaluation results, if the training results are not good, retraining is required.

[0081] In the description of the embodiments of the present invention, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "center", "top", "bottom", "top", "bottom", "inside", "outside", "inner side", "outer side" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. Among them, "inside" refers to an internal or enclosed area or space. "Periphery" refers to the area surrounding a specific component or a specific area.

[0082] In the description of the embodiments of the present invention, the terms "first", "second", "third", and "fourth" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", and "fourth" may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0083] In the description of the embodiments of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "install", "connect", "connect", and "assemble" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0084] In the description of the embodiments of the present invention, specific features, structures, materials or characteristics may be combined in a suitable manner in any one or more embodiments or examples.

[0085] In the description of the embodiments of the present invention, it should be understood that "-" and "~" represent the range between two values, and the range includes the endpoints. For example: "AB" represents a range greater than or equal to A and less than or equal to B. "A~B" represents a range greater than or equal to A and less than or equal to B.

[0086] In the description of the embodiments of the present invention, the term "and / or" herein is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " herein generally indicates that the associated objects before and after are in an "or" relationship.

[0087] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A reverse source tracing analysis method combining map river channel identification and dimensionality reduction calculation, characterized in that: The following steps are involved: Step S1, setting up river monitoring sensors, and reporting alarm information when pollutants are found; Step S2, obtaining monitoring point information, river hydrological information and source tracing range parameter information; Step S3, identifying the river network, obtaining the separated river layer, and then drawing the center line according to the river edge; Step S4, according to the drawn center line, the map is horizontally divided to obtain a pixel map after division, and the pixel map is formed into a file management set; Step S5, calculating the grid density through the reverse tracing algorithm model, including: Step a: Assemble a set of river channel cross-cutting node information sequence; according to the river channel flow direction grid cross-cutting nodes, calculate the river channel area and factor in the node, and organize the river channel cross-cutting node information sequence set with the river channel cross-cutting nodes as the association; Calculate the area of ​​the river channel within the node: According to the river pixels in the river channel, perform pixel point statistics and area conversion to obtain the area; Calculate node internal factors: Factors include: average flow velocity of river water, pollutant diffusion coefficient, pollutant sedimentation gradient and average river depth; when calculating node internal factors, basic scaling calculation is performed based on the current node area and the monitoring point area, and the scaling factor x = 1-m1 / m0, where m1 is the current node area and m0 is the node area where the monitoring point is located; when calculating the concentration coefficient, a pollutant sedimentation gradient coefficient is added, and the formula is: , v is the settling velocity of the particles, g is the acceleration due to gravity, ρ p is the density of the particle, ρ f is the density of the fluid, d is the diameter of the particle, and η is the dynamic viscosity of the fluid; Step b: Based on the convection model, combined with the scaling factor of each cross-cutting node, a single node simulation is derived; The single-node simulation derivation of the convection model includes forward derivation and reverse derivation. In the forward derivation and reverse derivation, the reasoning result of each cross-cutting node is calculated based on the result of the previous node and combined with the correction factor of the current node, and finally the simulated emission concentration value of pollutants of all nodes is obtained; The simulation derivation formula is: ， Where C is the concentration of pollutants; t is time; u is the average flow velocity of the river; D is the diffusion coefficient of the pollutants; x is the spatial coordinate along the river direction; Step S6, based on the river channel cross-section node information sequence set and the enterprise river channel outlet information, a list of pollution source enterprises is obtained, and the list of pollution source enterprises is scored for suspected pollution discharge through factor matching; The suspected pollution discharge scoring of step S6 is specifically as follows: three rounds of matching are performed on the pollution source enterprises in the list, and weights are assigned to the matching results of each round, and the final scoring scores are obtained by combining them; the matching method is specifically as follows: First round of matching: Obtain enterprise filing information, match the pollutant factor catalog in the enterprise's planned emissions with the pollutants, and obtain the first round of matching scores; Second round of matching: Match the pollutant records of the enterprise’s outlets with the pollutants to obtain the second round of matching scores; The third round of matching: Combined with the electricity consumption information of the enterprise's production line, the pollution control alarms of production and discharge are counted to obtain the third round of matching scores; Step S7, arranging the scores of the pollution source enterprises from high to low, and displaying them on the GIS map in turn for users to judge.

2. A reverse tracing analysis method combining map river channel identification and dimensionality reduction calculation according to claim 1, characterized in that: The monitoring point information in step S2 includes the geographical information of the alarm point, the set of pollution factors and the concentration of pollutants; the river hydrological information includes the current water flow, flow velocity and water level information of the monitoring point; the tracing range parameter information is the associated river range with the alarm node as the origin and the entire radius of the monitoring point to be traced.

3. The reverse tracing analysis method combining map river channel identification and dimensionality reduction calculation according to claim 1 is characterized in that: The step S3 specifically includes the following sub-steps: Step S301: using TiandiTu’s self-trained DeepLabv3+ river channel cutting model to identify the river channel and separate the river channel layers; Step S302: Draw a center line according to the river edge.

4. A reverse tracing analysis method combining map river channel identification and dimensionality reduction calculation according to claim 3, characterized in that: The drawing of the center line according to the river edge is specifically as follows: graying the river surface, and drawing the center line using the line connecting the midpoints of the linear distances of the second line closest to the positioning point of the left line within the pixel range as the center line.

5. A reverse tracing analysis method combining map river channel identification and dimensionality reduction calculation according to claim 4, characterized in that: The positioning points are configured according to the traceability accuracy requirements with an accuracy of 50 meters.

6. The reverse tracing analysis method combining map river channel identification and dimensionality reduction calculation according to claim 1 is characterized in that: The step S4 of horizontally dividing the map is specifically as follows: dividing the river map by a horizontal line that forms an angle of 90 degrees with the center line.

Images