CN117058549B

CN117058549B - Multi-industry secondary pollution dynamic source analysis system and analysis method

Info

Publication number: CN117058549B
Application number: CN202311050410.2A
Authority: CN
Inventors: 李雷; 王国梁; 徐盛荣; 崔泽虎; 李海智; 曹阳; 徐聪; 吴剑斌; 牛晓博; 晚军艳
Original assignee: Heilongjiang Ecological Environment Monitoring Center; 3Clear Technology Co Ltd
Current assignee: Heilongjiang Ecological Environment Monitoring Center; 3Clear Technology Co Ltd
Priority date: 2023-08-21
Filing date: 2023-08-21
Publication date: 2024-02-20
Anticipated expiration: 2043-08-21
Also published as: CN117058549A

Abstract

The invention discloses a multi-industry secondary pollution dynamic source analysis system and an analysis method, which are used for judging the change condition of pollution types according to a plurality of three-dimensional images to obtain the pollution direction. Detecting the change condition of the pollution type and the pollution degree value of the position of the pollution detector along with the time change to obtain a pollution source time type vector. And judging the pollution source category through the category neural network to obtain a pollution source geographic category vector. And obtaining the main pollution source position based on the plurality of same-category pollution values, the pollution source geographic category vector, the pollution source time category vector and the pollution direction. The pollution sources of the main discharged pollutants are accurately judged by adopting different types of pollution sources, different time of pollutant discharge, different pollutant discharge, category change of the pollutants at different time, and the movement direction of the pollutants and the pollution source distance corresponding to the pollutant content.

Description

Multi-industry secondary pollution dynamic source analysis system and analysis method

Technical Field

The invention relates to the technical field of computers, in particular to a multi-industry secondary pollution dynamic source analysis system and an analysis method.

Background

Secondary pollutant is a new pollutant which is discharged into the environment and is changed under the action of physical and chemical factors or organisms or is formed by reacting with other substances in the environment and has different physical and chemical properties from the primary pollutant, and is also called secondary pollutant. Sulfate aerosols, such as the oxidation of primary contaminant sulfur dioxide in the environment; nitrogen oxide, hydrocarbon and the like in automobile exhaust gas undergo photochemical reaction under the irradiation of sunlight, and generated ozone, peroxyacetyl nitrate (PAN), formaldehyde, ketone and the like.

The form of the pollution source exists in two types, namely a fixed source and a flowing source. By stationary source is meant a source of contamination that is stationary in position and location. Mainly refers to a large amount of pollutants discharged by industrial and mining enterprises in production. Metallurgy of metal,Industrial enterprises such as steel and building materials are all fixed sources with serious pollution to the atmospheric environment. The mobile pollution source is a pollution source formed by harmful gas discharged to the atmosphere when the vehicle is running. The method is divided into industrial pollution sources, agricultural pollution sources, transportation pollution sources, living pollution sources and the like according to the functions of human social activities. Industrial pollution source: the pollution source is formed by soot, dust, harmful compounds and the like discharged by industrial and mining enterprises such as thermal power generation, steel, chemical industry, silicate and the like in the production process. The pollution sources are different in the types and the amounts of the discharged pollutants due to the different production properties and flow processes of different industrial and mining enterprises, but have the common characteristics of concentrated emission sources, high concentration and high local pollution intensity. The large amount of pollutants discharged by the fixed source is the main culprit of urban atmospheric pollution. Agricultural pollution sources: mainly volatilizes and diffuses harmful substances generated in the processes of improper pesticide, chemical fertilizer, organic manure and the like, and NOX and CH are applied at the later stage ₄ The volatile pesticide components are dissipated from the soil into the atmosphere and the like. Traffic and transportation pollution sources: the tail gas discharged into the atmosphere when the transportation tools such as automobiles, airplanes, trains, ships and the like run. The pollution source belongs to a flowing pollution source, and the main pollutants are smoke dust, hydrocarbon, NOX, metal dust and the like, which are one of the main reasons for the deterioration of the urban atmospheric environment. The living pollution source is activities such as daily meal burning, heating and bathing, etc., and is used for burning fossil fuel to discharge smoke dust and SO to the atmosphere ₂ Contaminants such as NOX. The pollution sources belong to fixed sources, have the characteristics of wide distribution, large discharge capacity, low pollution height and the like, and are pollution sources which cannot be ignored in urban atmosphere pollution. However, with the advancement of urban electrification, urban living pollution sources are fundamentally restrained.

Detection of the type of contaminant is typically employed to match the type of emission of the corresponding source of the contaminant to find the corresponding source of the contaminant. This method is not accurate enough to judge the source of pollution. And because the categories of the secondary pollutants discharged by the industrial pollution source, the agricultural pollution source, the transportation pollution source and the living pollution are different, but overlap exists, the judgment can not be simply carried out according to the different categories.

Disclosure of Invention

The invention aims to provide a multi-industry secondary pollution dynamic source analysis system and an analysis method, which are used for solving the problems in the prior art.

In a first aspect, an embodiment of the present invention provides a method for analyzing a dynamic source of secondary pollution in multiple industries, including:

obtaining a plurality of three-dimensional images; the three-dimensional images are images obtained by detection of pollution detectors at different time points; the values in the three-dimensional image represent the pollution type and pollution degree values of the pollution detector in detecting a plurality of areas;

judging the change condition of the pollution type according to the plurality of three-dimensional images to obtain a pollution direction;

the three-dimensional images are overlapped and then input into a pollution change network, the change condition of the pollution type and the pollution degree value of the position of the pollution detector along with the time change is detected, and a pollution source time type vector is obtained; the value of the pollution source time class vector represents the probability that pollution sources at different times discharge pollutants of different pollution classes;

obtaining a geographic image; the geographic image represents the geographic condition of the area of the image detected by the pollution detector and the area around the image detected by the pollution detector; the geographic image is an area containing a pollution source; the geographic image includes a factory location, a traffic route, and a residential home;

Judging the pollution source category through a category neural network based on the three-dimensional images and the geographic images to obtain a pollution source geographic category vector; the value in the pollution source geographic category vector represents the probability of different pollutant judgment categories according to the emission of pollution sources at different positions;

adding pollution degree values with the same pollution types in the plurality of three-dimensional images to obtain a plurality of pollution values with the same type; the same-category pollution value represents a pollution degree value corresponding to the pollution category of the area where the three-dimensional image is located;

and obtaining the pollution source position based on the plurality of same-category pollution values, the pollution source geographic category vector, the pollution source time category vector and the pollution direction.

Optionally, the determining the change condition of the pollution category according to the plurality of three-dimensional images to obtain the pollution direction includes:

obtaining a first three-dimensional image; the first three-dimensional image is one image in a plurality of three-dimensional images;

obtaining a first position; the first position is one position in the first three-dimensional image;

obtaining a first category; the first category is a pollution category corresponding to the first position;

obtaining a second three-dimensional image; the second three-dimensional image is an image of the next time point of the first three-dimensional image in the plurality of three-dimensional images;

Marking the positions of the same category as the first category in the second three-dimensional image to obtain a plurality of second pollution positions;

adding the first position and the plurality of second pollution positions to the surrounding 8 values respectively to obtain a first class area and a plurality of second pollution areas; the first class region includes 9 values; the second contaminated area comprises 9 values;

based on the plurality of second pollution areas and the first category areas, expanding detection to the periphery to obtain a first pollution area and a first original pollution area; the first original polluted area represents an area with the same size as the first polluted area in the first three-dimensional image;

finding out a plurality of corresponding polluted areas and a plurality of original polluted areas by traversing the positions in the first three-dimensional image in sequence;

obtaining an optimal pollution area and an optimal original pollution area according to the plurality of pollution areas; the optimal pollution area is an area with the area larger than that of other pollution areas; the optimal original pollution area is an original pollution area corresponding to the optimal pollution area;

and acquiring a plurality of optimal pollution areas corresponding to the three-dimensional images and the original pollution areas corresponding to the three-dimensional images for multiple times, and judging to obtain the pollution direction.

Optionally, the detecting is extended to the periphery based on the plurality of second pollution areas and the first category areas, so as to obtain a first pollution area and a first original pollution area, which includes:

judging whether the pollution types of the positions corresponding to the second pollution areas and the first pollution areas are the same or not to obtain a plurality of matching areas; the matching area is an area with the same pollution type as the position corresponding to the first type area in the second pollution area; the sizes of the plurality of matching areas are different;

respectively adding a plurality of matching areas into values adjacent to the boundary for a plurality of times to obtain a plurality of expansion matching areas;

expanding the first category region into a region with the same size as the expansion matching region for multiple times to obtain multiple expansion category regions; one extension class area corresponds to one extension matching area;

judging whether the pollution types of the plurality of expansion type areas and the corresponding positions of the plurality of expansion matching areas are the same or not to obtain a plurality of first matching areas; a first matching region corresponds to an extended category region corresponds to an extended matching region; the first matching area is an area with the same pollution category as the corresponding position of the extended category area in the extended matching area pair;

Expanding a plurality of first matching areas through multiple iterations, expanding the expanded category areas into the same size of the first matching areas, judging whether the pollution categories of the corresponding positions are the same, and if the pollution categories of the corresponding positions are the same, obtaining a new matching area until the pollution categories of the corresponding positions are different, and obtaining a plurality of final matching areas and a plurality of final expanded category areas; one final extension class region corresponds to one final matching region; the final matching region represents a region which cannot be matched; the final expansion category region represents an expansion category region which corresponds to the final matching region and has the same size as the final matching region;

and taking the area of which the final matching area is larger than other final matching areas as a first pollution area, and taking the area of the first pollution area corresponding to the final expansion category area as a first original pollution area.

Optionally, the step of inputting the superimposed plurality of three-dimensional images into a pollution change network, detecting a change condition of a pollution type and a pollution level value at a position where the pollution detector is located along with time change, and obtaining a pollution source time type vector pollution level value includes:

obtaining detection time; the detection time is a time difference corresponding to a plurality of three-dimensional images;

Overlapping the three-dimensional images to obtain a time three-dimensional image;

inputting the time three-dimensional image and time into a first convolution network, extracting pollution characteristics, and obtaining a pollution source time category vector; the pollution source time category vector is a one-dimensional vector.

Optionally, the step of inputting the time three-dimensional image and the time into the first convolution network, extracting the pollution characteristic, and obtaining the pollution source time category vector includes:

extracting features of a plurality of convolution kernels in a first convolution network and the time three-dimensional image to obtain a three-dimensional feature map;

the number of channels of a plurality of convolution kernels for extracting the features in the first convolution network is the same as the number of corresponding three-dimensional images corresponding to the time three-dimensional images;

multiplying the three-dimensional feature map with a convolution kernel of the last layer of the first convolution network to obtain a pollution source time category vector;

the number of convolution kernels of the last layer of the first convolution network is the same as the number of the corresponding three-dimensional images of the time three-dimensional image; the convolution kernel size of the last layer of the first convolution network is the same as the size of the three-dimensional feature map.

Optionally, the determining, based on the plurality of three-dimensional images and the geographic image, the pollution source category through the category neural network, to obtain a pollution source geographic category vector, includes:

According to the geographic image, calculating the distance from a pollution source on the geographic image to the position of a pollution detector on the geographic image to obtain a distance proportion vector; the values in the distance scale vector represent ratios of distances of the plurality of pollution sources to the pollution detector at the location of the geographic image;

inputting the three-dimensional image into a first class convolutional network, extracting features and obtaining a first class feature vector; the values in the first class feature vector represent features of the contaminant;

inputting the distance proportion vector into a proportion neural network, extracting characteristics and obtaining a proportion output vector; the values in the scale output vector represent features of a distance scale vector;

inputting the first category characteristic vector and the proportion output vector into a fusion neural network, carrying out characteristic fusion, and judging the category of the pollution source to obtain a geographic category vector of the pollution source;

the first category convolutional network, the proportional neural network, and the fused neural network constitute a category neural network.

Optionally, the obtaining the pollution source position based on the plurality of same-category pollution values, the pollution source geographic category vector, the pollution source time category vector and the pollution direction includes:

inputting the category corresponding to the value larger than the geographic category threshold value in the pollution source geographic category vector into a pollution source geographic category set;

Inputting the category corresponding to the value larger than the time category threshold value in the pollution source time category vector into a pollution source time category set;

solving intersection of the pollution source geographic category set and the pollution source time category set to obtain a pollution source category set;

obtaining a pollution source direction set according to the pollution direction and the geographic image; the value in the pollution source direction set is a pollution source intersected with a straight line corresponding to the pollution direction in the geographic image;

forming a pollution source set by using the pollution sources which are in the same category in the pollution source direction set and the pollution source category set;

obtaining a pollution concentration distance proportion; the pollution concentration distance proportion represents the ratio of the pollution concentration stored in the database to the corresponding distance;

dividing the pollution content by the pollution concentration distance proportion to obtain a pollution distance;

and taking the position of the pollution source corresponding to the pollution distance in the pollution source set as the pollution source position.

Optionally, the determining, by acquiring a plurality of optimal pollution areas and original pollution areas corresponding to the plurality of three-dimensional images, to obtain a pollution direction includes:

constructing coordinate axes according to the three-dimensional images; calculating a slope between a first position corresponding to the original pollution area and a first position corresponding to the optimal pollution area to obtain a first pollution slope;

Calculating the slopes of a plurality of original pollution areas and the corresponding optimal pollution areas for a plurality of times to obtain a plurality of pollution slopes;

and clustering the plurality of pollution slopes, and taking a clustering center as a pollution direction.

Optionally, the class neural network training method includes:

obtaining a training set; the training set comprises a plurality of training data and a corresponding plurality of labeling data; the training data comprises a training three-dimensional image and a training distance proportion vector; the labeling data represents the labeling pollution source category; the labeling pollution source category sets the value of the pollution source for discharging pollution as 1 and sets the pollution source without discharging pollution as 0;

inputting the training data into a category convolution network, extracting the characteristics of pollution categories and pollution degree values in the three-dimensional image, and obtaining a training pollution source category vector;

calculating the loss of the training pollution source category vector and the labeling pollution source category to obtain a loss value;

obtaining the current training iteration times of the class neural network and the preset maximum iteration times of the class neural network training;

and stopping training when the loss value is smaller than or equal to a threshold value or the training iteration number reaches the maximum iteration number, and obtaining the trained class neural network.

In a second aspect, an embodiment of the present invention provides a multi-industry secondary pollution dynamic source analysis system, including:

the acquisition module is used for: obtaining a plurality of three-dimensional images; the three-dimensional images are images obtained by detection of pollution detectors at different time points; the values in the three-dimensional image represent the pollution type and pollution degree values of the pollution detector in detecting a plurality of areas; obtaining a geographic image; the geographic image represents the geographic condition of the area of the image detected by the pollution detector and the area around the image detected by the pollution detector; the geographic image is an area containing a pollution source; the geographic image includes a factory location, a traffic route, and a residential home;

pollution direction module: judging the change condition of the pollution type according to the plurality of three-dimensional images to obtain a pollution direction;

pollution source time category module: the three-dimensional images are overlapped and then input into a pollution change network, the change condition of the pollution type and the pollution degree value of the position of the pollution detector along with the time change is detected, and a pollution source time type vector is obtained; the value of the pollution source time class vector represents the probability that pollution sources at different times discharge pollutants of different pollution classes;

Pollution source geographic classification module: judging the pollution source category through a category neural network based on the three-dimensional images and the geographic images to obtain a pollution source geographic category vector; the value of the pollution source geographic category vector represents the probability that pollution sources at different positions discharge pollutants of different pollution categories;

and a pollution content module: adding pollution degree values with the same pollution types in the plurality of three-dimensional images to obtain a plurality of pollution values with the same type; the same-category pollution value represents a pollution degree value corresponding to the pollution category of the area where the three-dimensional image is located;

the main pollution source position judging module: and obtaining the main pollution source position based on the plurality of same-category pollution values, the pollution source geographic category vector, the pollution source time category vector and the pollution direction.

Compared with the prior art, the embodiment of the invention achieves the following beneficial effects:

the invention provides a multi-industry secondary pollution dynamic source analysis method and a system, wherein the method comprises the following steps: a plurality of three-dimensional images are obtained. The three-dimensional images are images detected by pollution detectors at different time points. The values in the three-dimensional image represent the contamination class and contamination level values of the contamination detector at detecting the plurality of areas. And judging the change condition of the pollution type according to the plurality of three-dimensional images to obtain the pollution direction. And (3) superposing the three-dimensional images, inputting the superposed three-dimensional images into a pollution change network, detecting the change condition of the pollution type and the pollution degree value of the position of the pollution detector along with the time change, and obtaining a pollution source time type vector. The value of the pollution source time class vector represents the probability that different pollution sources at different times emit pollutants of different pollution classes. A geographic image is obtained. The geographic image represents the geographic condition of the area of the image detected by the contamination detector and the area surrounding the image detected by the contamination detector. The geographic image is an area containing a source of contamination. The geographic image includes a factory location, a traffic route, and a residential home. And judging the pollution source category through a category neural network based on the plurality of three-dimensional images and the geographic images to obtain a pollution source geographic category vector. The value of the pollution source geographical category vector represents the probability that pollution sources at different locations will emit pollutants of different pollution categories. And adding the pollution degree values with the same pollution types in the plurality of three-dimensional images to obtain a plurality of pollution values with the same type. And the same-category pollution value represents a pollution degree value corresponding to the pollution category of the area where the three-dimensional image is located. And obtaining the main pollution source position based on the plurality of same-category pollution values, the pollution source geographic category vector, the pollution source time category vector and the pollution direction.

Since the types of pollutants emitted by industrial pollution sources, agricultural pollution sources, transportation pollution sources and living pollution sources are different, the judgment on the type of pollutants and the concentration condition can be adopted. The pollution sources of the main discharged pollutants are accurately judged by adopting different types of pollution sources, different time of pollutant discharge, different pollutant discharge, category change of the pollutants at different time, and the movement direction of the pollutants and the pollution source distance corresponding to the pollutant content.

Meanwhile, aiming at the characteristic that the pollutants are easy to drift away, the accuracy of direction prediction is easy to influence, the technical scheme of the invention adopts iteration to carry out comparison size matching to obtain the moving direction of the pollutants, so that the technical problem of the accuracy of direction prediction, which is influenced by the easy drift away of the pollutants, can be solved, and the accuracy of direction prediction can be improved. And simultaneously, three-dimensional images with a plurality of superimposed times are input into a first convolution network to obtain a pollution source time category vector. The time information of the three-dimensional images with the superimposed time is enhanced, based on the time information, the three-dimensional images with the superimposed time are input into the first convolution network, and the first convolution network is used for rapidly judging, so that corresponding pollution sources with different emission time can be judged, and the accuracy of judging the pollution sources is improved. The pollution geographic category vector adopts the method that the pollution characteristics of the three-dimensional image are extracted because the concentration and the category of pollutants discharged by pollution sources of different categories at different positions are different, the ratio of the distances from the pollution detector to the position of the geographic image is extracted, and the two characteristics are fused, so that the distance ratio vector can establish a relation to the pollution characteristics of the three-dimensional image, and the probability of judging the pollution category differently when the pollution sources are discharged at different positions is calculated.

Drawings

Fig. 1 is a flowchart of a multi-industry secondary pollution dynamic source analysis method provided by an embodiment of the invention.

Fig. 2 is a schematic block diagram of an electronic device according to an embodiment of the present invention.

The marks in the figure: a bus 500; a receiver 501; a processor 502; a transmitter 503; a memory 504; bus interface 505.

Detailed Description

The present invention will be described in detail with reference to the accompanying drawings.

Example 1

As shown in fig. 1, the embodiment of the invention provides a multi-industry secondary pollution dynamic source analysis method, which comprises the following steps:

s101: a plurality of three-dimensional images are obtained. The three-dimensional images are images obtained by detection of pollution detectors at different time points; the values in the three-dimensional image represent the contamination class and contamination level values of the contamination detector at detecting the plurality of areas. The values in the three-dimensional image include contamination class and contamination level values of the contamination detector at detecting the plurality of regions.

Wherein the first and second dimensions of the three-dimensional image represent the length and width of a square area that can be detected by the contamination detector. The third dimension of the three-dimensional image contains 11 values, 5 of which represent the pollution categories, which are classified as: carbon monoxide, NOX, CH ₄ Hydrocarbons, SO ₂ . The other 6 values represent pollution degree values, the air pollution index (AQI) is 0-50, the air quality level is one level, and the air quality condition is excellent; the air pollution index is 51-100, the air quality level is two-level, and the air quality condition is good; the air pollution index is 101-150, the air quality level is three-level, and the air quality condition belongs to light pollution; the air pollution index is 151-200, the air quality level is four, and the air quality condition belongs to moderate pollution; the air pollution index is 201-300, the air quality level is five, and the air quality condition belongs to severe pollution; the air pollution index is more than 300, the air quality level is six, and the air quality condition belongs to serious pollution. The air pollution index is a form of simplifying the monitored air concentration into a single conceptual index value.

Alternatively, the pollution detector may be a camera, or may be a dedicated instrument for monitoring air quality and pollution source exhaust gas, such as a TSP sampler, a PM10 sampler, a PM2.5 sampler, or a photochemical smog detection system. After the air pollution information is acquired by the instruments, the three-dimensional image is generated: after an empty image covering the position of the detection area (or the empty image corresponding to the detection area of the world coordinates one by one) is constructed, air pollution information (pollution information comprises pollution type and pollution concentration) is detected by an air quality and pollution source waste gas monitoring special instrument, the pollution concentration is assigned to a pixel point corresponding to the detection area in the empty image, namely, the pixel value of the pixel point corresponding to the detection area in the empty image is assigned to the pollution concentration, meanwhile, the type of the pixel point is marked as the detected pollution type, namely, the value in the three-dimensional image can be set as a key value pair comprising the pollution type and the pollution degree value. If the contamination detector is a camera, the three-dimensional image is directly captured.

In this embodiment, the pollution detector adopts a continuous automatic measurement method, and optical and chemical measurement methods are generally used.

S102: and judging the change condition of the pollution type according to the plurality of three-dimensional images to obtain the pollution direction.

S103: and (3) superposing the three-dimensional images, inputting the superposed three-dimensional images into a pollution change network, detecting the change condition of the pollution type and the pollution degree value of the position of the pollution detector along with the change of time, and obtaining a pollution source type vector. The value of the pollution source time class vector represents the probability that different pollution sources at different times emit pollutants of different pollution classes.

S104: a geographic image is obtained. The geographic image represents the geographic condition of the area of the image detected by the contamination detector and the area surrounding the image detected by the contamination detector. The geographic image is an area containing a source of contamination. The geographic image includes a factory location, a traffic route, and a residential home.

Wherein the geographic image is a three-dimensional image. The first dimension of the geographic image represents the long (long side) of a square area formed by the area of the image detected by the contamination detector and the area around the image detected by the contamination detector. The second dimension of the geographic image represents the width of a square area formed by the area of the image detected by the contamination detector and the area around the image detected by the contamination detector. The length in the third dimension of the geographic image is 3, representing the plant location, the traffic route, and the residential home, respectively.

In this embodiment, if the area of the image detected by the pollution detector is a square position of 100 meters x100 meters, the area around the image obtained by the pollution detector is a geographic image with a boundary of 2000 meters x2000 meters and a three-dimensional image as a center, wherein the width and the height of the square position of 100 meters x100 meters are extended by 1900 meters.

If a certain position in the geographic image contains a factory, the value of the corresponding position in the first layer is set to be 1, and the other positions are set to be 0. The value of the corresponding position of the traffic route in the second layer is set to 1, and the other positions are set to 0. If a resident house is present in a certain position, the value of the corresponding position in the third floor is set to 1, and the other positions are set to 0.

S105: and judging the pollution source category through a category neural network based on the plurality of three-dimensional images and the geographic images to obtain a pollution source geographic category vector. The value of the pollution source geographical category vector represents the probability that pollution sources at different locations will emit pollutants of different pollution categories.

S106: and adding the pollution degree values with the same pollution types in the plurality of three-dimensional images to obtain a plurality of pollution values with the same type. And the same-category pollution value represents a pollution degree value corresponding to the pollution category of the area where the three-dimensional image is located.

S107: based on the plurality of co-classified pollution values, the pollution source geographical classification vector, the pollution source time classification vector and the pollution direction, a main pollution source position is obtained (the detailed manner of obtaining the main pollution source position is shown in the follow-up operation).

a first three-dimensional image is obtained. The first three-dimensional image is one image of a plurality of three-dimensional images.

A first position is obtained. The first location is one of the locations in the first three-dimensional image.

A first category is obtained. The first category is a contamination category corresponding to the first location.

A second three-dimensional image is obtained. The second three-dimensional image is an image of a next time point of the first three-dimensional image in the plurality of three-dimensional images.

And marking the positions of the same category as the first category in the second three-dimensional image to obtain a plurality of second pollution positions.

Adding the first position and the plurality of second pollution positions to the surrounding 8 values respectively to obtain a first class area and a plurality of second pollution areas; the first class region includes 9 values; the second contaminated area comprises 9 values.

And finding the position which is the same as the first three-dimensional image category in the second three-dimensional image, and finding the point which is possibly matched. And respectively expanding the points matched with the first three-dimensional image and the second three-dimensional image, namely by taking the first position and the plurality of second pollution positions as centers, respectively adding 9 surrounding values, and obtaining a plurality of 3x3 squares.

Based on the plurality of second pollution areas and the first category areas, expanding detection to the periphery to obtain a first pollution area and a first original pollution area; the first contaminated region represents a region of the first three-dimensional image having the same size as the first contaminated region.

Wherein the first contaminated area represents an expanded first category area. The first contaminated area represents an extended first category area and the extended second contaminated area.

And judging whether the values in the area are the same or not by continuously expanding the area and increasing surrounding values until different values appear in the area, so as to obtain a first polluted area.

Wherein, as in the present embodiment, the same position as the first three-dimensional image category is found in the second three-dimensional imageAnd->Wherein for better explanation the value of the same position is set to 1 and the value of the different position is set to 0. Expanding the two matching areas, then find +_in the second three-dimensional image >The corresponding expanded region isIn the second three-dimensional image +.>The corresponding expanded region is +>ExpandedAnd->Are not identical.

The first pollution area is the pollution area corresponding to the position of the first category in the found second three-dimensional image.

And sequentially traversing the positions in the first three-dimensional image to find out a plurality of corresponding polluted areas and a plurality of original polluted areas.

Wherein, because different positions in the first three-dimensional image are used as starting points, the found polluted areas are different, thereby obtaining a plurality of polluted areas.

And obtaining an optimal pollution area and an optimal original pollution area according to the plurality of pollution areas. The optimal pollution area is an area larger than other pollution areas. The optimal original pollution area is an original pollution area corresponding to the optimal pollution area.

And finding a polluted area with highest matching degree in the first three-dimensional image and the second three-dimensional image.

The pollution moving direction is obtained by controlling the opening detection under the windless condition.

In the process of changing the emission pollution of a pollution source in real conditions, the concentration of different types of different areas can change, which can lead to the displacement of the position of one pollution type of the current three-dimensional image and the position of the same pollution type of the previous three-dimensional image. By the method, two images are input to find the areas which are similarly arranged, and the position of the largest area is found according to the up-down left-right association of one value, so that the moving change direction is obtained, and the accuracy of confirming the moving change direction by the method is high. That is, the technical scheme provided by the invention can confirm the direction through the change of the position corresponding to the pollution type, and can improve the accuracy of direction prediction.

and judging whether the pollution types of the positions corresponding to the second pollution areas and the first pollution areas are the same or not, and obtaining a plurality of matching areas. The matching area is an area with the same pollution type as the position corresponding to the first pollution type area in the second pollution area. The plurality of matching regions are different in size.

In this embodiment, because the contamination category is represented by a numerical value, the categories of the areas corresponding to the first and second contamination positions are subtracted, and if all of them are 0, it indicates that the matching is successful.

And respectively adding the multiple matching areas into values adjacent to the boundary for multiple times to obtain multiple expansion matching areas.

Wherein the value adjacent to the boundary is a value of 1 or 0, which is the absolute value of the subtraction of the abscissa or the ordinate of the matching region.

Expanding the first category region into a region with the same size as the expansion matching region for multiple times to obtain multiple expansion category regions; one extension class region corresponds to one extension matching region.

Wherein, by first judging the similar square part of 3X3, such as the second three-dimensional image is foundAndthe two parts are identical, wherein for a better explanation the value of the same location of the contamination category is set to 1 and the value of the different location is set to 0. Expanding the two matching areas, setting the expanded area to 2 for better explanation, thenThe corresponding expanded region is +>The corresponding expanded region is

Wherein similar locations represent locations of the same contamination category.

Judging whether the pollution types of the plurality of expansion type areas and the corresponding positions of the plurality of expansion matching areas are the same or not to obtain a plurality of first matching areas; a first matching region corresponds to an extended category region corresponds to an extended matching region; the first matching area is an area, in the pair of the extended matching areas, of which the pollution category is the same as that of the corresponding position of the extended category area.

Expanding a plurality of first matching areas through multiple iterations, expanding the expanded category areas into the same size of the first matching areas, judging whether the pollution categories of the corresponding positions are the same, and if the pollution categories of the corresponding positions are the same, obtaining a new matching area until the pollution categories of the corresponding positions are different, and obtaining a plurality of final matching areas and a plurality of final expanded category areas; one final extension class region corresponds to one final matching region; the final matching region represents a region which cannot be matched; and the final expansion category region represents an expansion category region which corresponds to the final matching region and has the same size as the final matching region.

The final matching area and the corresponding final expansion category area not only comprise the area but also comprise the position of the area.

the detection time is obtained. The detection time is the phase difference time corresponding to the three-dimensional images.

The time difference is that the phase difference between two three-dimensional images is longer than the phase difference between other two three-dimensional images.

Here, as in the present example, detection was performed every 10 minutes at 24 hours, and 24×60/10=144 three-dimensional images were obtained. The phase difference time corresponding to the plurality of three-dimensional images is 24.

And (3) overlapping the three-dimensional images, inputting the overlapped three-dimensional images into a first convolution network in time, extracting pollution characteristics, and obtaining a pollution source time category vector. The pollution source time category vector is a one-dimensional vector.

Wherein the first convolutional network is itself a convolutional neural network (Convolutional Neural Networks, CNN) in an embodiment.

Wherein the value of the pollution source time class vector represents the probability that pollution sources at different times emit pollutants of different pollution classes.

Wherein the values in the pollution source time class vector represent the overall pollution level values of the areas at different times.

the number of channels of a plurality of convolution kernels for extracting features in the first convolution network is the same as the number of corresponding three-dimensional images corresponding to the time three-dimensional image.

In this embodiment, the number of corresponding three-dimensional images corresponding to the time three-dimensional image is 144, and then the number of channels of the convolution kernel is 144. And (5) extracting features by multiplying convolution kernels with different numbers to obtain a three-dimensional feature map.

If the three-dimensional feature map in this embodiment is 7x7x1024, the number of convolution kernels of the last layer of the first convolution network is 144 as the number of convolution kernels corresponding to the plurality of three-dimensional images corresponding to the time three-dimensional image, and then the convolution kernels of the last layer of the first convolution network are convolution kernels with the channel number of 144 and the size of 7x7, and the obtained feature vector with the pollution source time category vector of 1x 144.

and calculating the distance from the pollution source on the geographic image to the position of the pollution detector on the geographic image according to the geographic image to obtain a distance proportion vector. The values in the distance scale vector represent ratios of distances of the plurality of contamination sources to the location of the contamination detector at the geographic image.

And inputting the three-dimensional image into a first class convolution network, extracting the characteristics, and obtaining a first class characteristic vector. The values in the first class feature vector represent the features of the contaminant.

And inputting the distance proportion vector into a proportion neural network, extracting features, and obtaining a proportion output vector. The values in the scale output vector represent features of the distance scale vector.

And inputting the first category characteristic vector and the proportional output vector into a fusion neural network, carrying out characteristic fusion, and judging the category of the pollution source to obtain a geographic category vector of the pollution source.

Wherein the value of the pollution source geographical category vector represents the probability that pollution sources at different locations emit pollutants of different pollution categories.

By the method, the pollution source is judged mainly by combining the pollution types under different conditions.

Optionally, the obtaining the main pollution source position based on the plurality of same-category pollution values, the pollution source geographic category vector, the pollution source time category vector and the pollution direction includes:

and inputting the category corresponding to the value larger than the geographic category threshold value in the pollution source geographic category vector into the pollution source geographic category set.

In this embodiment, the geographic category threshold is 0.85.

And inputting the category corresponding to the value larger than the time category threshold value in the pollution source time category vector into the pollution source time category set.

In this embodiment, the time class threshold is 0.9.

And solving an intersection of the pollution source geographic category set and the pollution source time category set to obtain a pollution source category set.

And obtaining a pollution source direction set according to the pollution direction and the geographic image. The value in the pollution source direction set is the pollution source intersected with the straight line corresponding to the pollution direction in the geographic image.

And writing the pollution source direction set according to all pollution sources which find the pollution direction in the geographic image.

obtaining a pollution concentration distance proportion; the pollution concentration distance ratio represents a ratio of the pollution concentration stored in the database to the corresponding distance.

The contamination distance is obtained by dividing the contamination content by the contamination concentration distance ratio.

And taking the position of the pollution source corresponding to the pollution distance in the pollution source set as a main pollution source position.

By the method, the same-category pollution value, the pollution area condition, the category greater than the threshold value in the pollution source category vector and the pollution movement condition are judged. And dynamically judging the positions of pollution sources which cause pollution at different time points on the geographic image according to the positions, the directions and the categories.

And constructing a coordinate axis according to the three-dimensional image.

The coordinate axis is a two-dimensional coordinate axis, the lower left corner point of the three-dimensional image is used as a zero point of the coordinate axis, the width of the three-dimensional image is used as an abscissa of the coordinate axis, and the length of the three-dimensional image is used as an ordinate of the coordinate axis. The lower left to lower right direction is taken as the abscissa and the lower left to upper left direction is taken as the ordinate.

And calculating the slope of the first position corresponding to the original pollution area and the first position corresponding to the optimal pollution area to obtain a first pollution slope.

The first positions corresponding to the original pollution areas and the first positions corresponding to the optimal pollution areas respectively represent the center points of the areas;

and calculating the slopes of the plurality of original pollution areas and the corresponding optimal pollution areas for a plurality of times to obtain a plurality of pollution slopes.

In this embodiment, the K-means algorithm is used for clustering.

Optionally, the class neural network training method includes:

obtaining the current training iteration times of the class neural network and the preset maximum iteration times of the class neural network training.

The preset maximum iteration number of the class neural network training is 12000.

By the method, the concentration and time change of the pollution in the three-dimensional image: the superimposed convolution judges which type of pollution source and wind direction position is possible: obtaining values and geographic positions of different azimuth positions: the map makes a judgment as to which type of pollution source is possible, such as a factory map and a traffic route, and a plurality of industries represent pollution discharge and pollution discharge modes of different industries.

Example 2

Based on the method for analyzing the secondary pollution dynamic sources in multiple industries, the embodiment of the invention also provides a system for analyzing the secondary pollution dynamic sources in multiple industries, which comprises an acquisition module, a pollution direction module, a pollution source time type module, a pollution source geographic type module, a pollution content module and a main pollution source position judging module.

The acquisition module is used for acquiring a plurality of three-dimensional images. The three-dimensional images are images detected by pollution detectors at different time points. The values in the three-dimensional image represent the contamination class and contamination level values of the contamination detector at detecting the plurality of areas. A geographic image is obtained. The geographic image represents the geographic condition of the area of the image detected by the pollution detector and the area around the image detected by the pollution detector; the geographic image is an area containing a source of contamination. The geographic image includes a factory location, a traffic route, and a residential home.

The pollution direction module is used for judging the change condition of the pollution type according to the plurality of three-dimensional images to obtain the pollution direction.

The pollution source time category module is used for inputting the superimposed three-dimensional images into a pollution change network, detecting the change condition of the pollution category and the pollution degree value of the position of the pollution detector along with the time change, and obtaining a pollution source time category vector. The value of the pollution source time class vector represents the probability that different pollution sources at different times emit pollutants of different pollution classes.

The pollution source geographic category module is used for judging the category of the pollution source based on the three-dimensional images and the geographic images through the category neural network to obtain a pollution source geographic category vector. The value of the pollution source geographical category vector represents the probability that pollution sources at different locations will emit pollutants of different pollution categories.

The pollution content module is used for adding pollution degree values with the same pollution types in the plurality of three-dimensional images to obtain a plurality of pollution values with the same type. And the same-category pollution value represents a pollution degree value corresponding to the pollution category of the area where the three-dimensional image is located.

The main pollution source position judging module is used for obtaining a main pollution source position based on a plurality of same-category pollution values, a pollution source geographic category vector, a pollution source time category vector and a pollution direction.

The specific manner in which the various modules perform the operations in the systems of the above embodiments have been described in detail herein with respect to the embodiments of the method, and will not be described in detail herein.

The embodiment of the present invention further provides an electronic device, as shown in fig. 2, including a memory 504, a processor 502, and a computer program stored in the memory 504 and capable of running on the processor 502, where the steps of any one of the methods for multi-industry secondary pollution dynamic source analysis described above are implemented when the processor 502 executes the program.

Where in FIG. 2 a bus architecture (represented by bus 500), bus 500 may include any number of interconnected buses and bridges, with bus 500 linking together various circuits, including one or more processors, represented by processor 502, and memory, represented by memory 504. Bus 500 may also link together various other circuits such as peripheral devices, voltage regulators, power management circuits, etc., as are well known in the art and, therefore, will not be described further herein. Bus interface 505 provides an interface between bus 500 and receiver 501 and transmitter 503. The receiver 501 and the transmitter 503 may be the same element, i.e. a transceiver, providing a means for communicating with various other apparatus over a transmission medium. The processor 502 is responsible for managing the bus 500 and general processing, while the memory 504 may be used to store data used by the processor 502 in performing operations.

The embodiment of the invention also provides a computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, and the program is executed by a processor to realize the steps of any one of the method for analyzing the dynamic sources of the secondary pollution in multiple industries and the related data.

The algorithms and displays presented herein are not inherently related to any particular computer, virtual system, or other apparatus. Various general-purpose systems may also be used with the teachings herein. The required structure for a construction of such a system is apparent from the description above. In addition, the present invention is not directed to any particular programming language. It will be appreciated that the teachings of the present invention described herein may be implemented in a variety of programming languages, and the above description of specific languages is provided for disclosure of enablement and best mode of the present invention.

In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.

Similarly, it should be appreciated that in the foregoing description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects.

Various component embodiments of the invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that some or all of the functions of some or all of the components in an apparatus according to embodiments of the present invention may be implemented in practice using a microprocessor or Digital Signal Processor (DSP). The present invention can also be implemented as an apparatus or device program (e.g., a computer program and a computer program product) for performing a portion or all of the methods described herein. Such a program embodying the present invention may be stored on a computer readable medium, or may have the form of one or more signals. Such signals may be downloaded from an internet website, provided on a carrier signal, or provided in any other form.

Furthermore, those skilled in the art will appreciate that while some embodiments herein include some features but not others included in other embodiments, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments.

Claims

1. A multi-industry secondary pollution dynamic source analysis method is characterized by comprising the following steps:

Judging the pollution source category through a category neural network based on the three-dimensional images and the geographic images to obtain a pollution source geographic category vector; the value of the pollution source geographic category vector represents the probability that pollution sources at different positions discharge pollutants of different pollution categories;

obtaining a pollution source position based on a plurality of same-category pollution values, a pollution source geographic category vector, a pollution source time category vector and a pollution direction;

judging the change condition of the pollution category according to the plurality of three-dimensional images to obtain the pollution direction, wherein the method comprises the following steps:

2. The multi-industry secondary pollution dynamic source analyzing method according to claim 1, wherein the expanding the detection to the periphery based on the plurality of second pollution areas and the first category areas to obtain a first pollution area and a first original pollution area comprises:

3. The method for analyzing the dynamic source of the secondary pollution in the multiple industries according to claim 1, wherein the steps of inputting the superimposed three-dimensional images into a pollution change network, detecting the change condition of the pollution type and the pollution level value of the position of the pollution detector along with the time change, and obtaining the pollution source time type vector pollution level value include:

4. The method for analyzing the dynamic sources of the secondary pollution in the multiple industries according to claim 3, wherein the steps of inputting the time three-dimensional image and the time into the first convolution network, extracting the pollution characteristics and obtaining the pollution source time category vector comprise the following steps:

5. The method for analyzing the dynamic source of the secondary pollution in the multiple industries according to claim 1, wherein the step of judging the type of the pollution source through the type neural network based on the multiple three-dimensional images and the geographic images to obtain the geographic type vector of the pollution source comprises the following steps:

6. The method for analyzing the dynamic source of the secondary pollution in multiple industries according to claim 1, wherein the obtaining the pollution source position based on the plurality of same-category pollution values, the pollution source geographic category vector, the pollution source time category vector and the pollution direction comprises:

7. The method for analyzing the dynamic source of the secondary pollution in the multiple industries according to claim 1, wherein the step of obtaining the pollution direction by obtaining the corresponding optimal pollution areas and the corresponding original pollution areas of the three-dimensional images for multiple times comprises the following steps:

constructing coordinate axes according to the three-dimensional images;

calculating a slope between a first position corresponding to the original pollution area and a first position corresponding to the optimal pollution area to obtain a first pollution slope;

8. The multi-industry secondary pollution dynamic source analysis method according to claim 1, wherein the training method of the class neural network comprises the following steps:

9. A multi-industry secondary pollution dynamic source analysis system, comprising:

a pollution source position judging module: obtaining a pollution source position based on a plurality of same-category pollution values, a pollution source geographic category vector, a pollution source time category vector and a pollution direction;