CN117054353B - Atmospheric pollution source area positioning analysis method and system - Google Patents
Atmospheric pollution source area positioning analysis method and system Download PDFInfo
- Publication number
- CN117054353B CN117054353B CN202311039879.6A CN202311039879A CN117054353B CN 117054353 B CN117054353 B CN 117054353B CN 202311039879 A CN202311039879 A CN 202311039879A CN 117054353 B CN117054353 B CN 117054353B
- Authority
- CN
- China
- Prior art keywords
- pollution
- image
- area
- spectrogram
- spectrum
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000004458 analytical method Methods 0.000 title claims description 7
- 238000001514 detection method Methods 0.000 claims abstract description 166
- 238000012544 monitoring process Methods 0.000 claims abstract description 104
- 238000001228 spectrum Methods 0.000 claims abstract description 95
- 230000004927 fusion Effects 0.000 claims abstract description 60
- 238000013528 artificial neural network Methods 0.000 claims abstract description 48
- 230000011218 segmentation Effects 0.000 claims abstract description 45
- 238000010586 diagram Methods 0.000 claims abstract description 44
- 238000000034 method Methods 0.000 claims abstract description 40
- 230000003595 spectral effect Effects 0.000 claims abstract description 40
- 238000012549 training Methods 0.000 claims description 39
- 238000011109 contamination Methods 0.000 claims description 31
- 238000002372 labelling Methods 0.000 claims description 28
- 239000013598 vector Substances 0.000 claims description 27
- 238000012935 Averaging Methods 0.000 claims description 5
- 239000000356 contaminant Substances 0.000 claims description 4
- 238000003491 array Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- RAHZWNYVWXNFOC-UHFFFAOYSA-N Sulphur dioxide Chemical compound O=S=O RAHZWNYVWXNFOC-UHFFFAOYSA-N 0.000 description 14
- 238000013527 convolutional neural network Methods 0.000 description 9
- MGWGWNFMUOTEHG-UHFFFAOYSA-N 4-(3,5-dimethylphenyl)-1,3-thiazol-2-amine Chemical compound CC1=CC(C)=CC(C=2N=C(N)SC=2)=C1 MGWGWNFMUOTEHG-UHFFFAOYSA-N 0.000 description 7
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 description 7
- 229910002091 carbon monoxide Inorganic materials 0.000 description 7
- JCXJVPUVTGWSNB-UHFFFAOYSA-N nitrogen dioxide Inorganic materials O=[N]=O JCXJVPUVTGWSNB-UHFFFAOYSA-N 0.000 description 7
- 238000003915 air pollution Methods 0.000 description 4
- 238000004590 computer program Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 239000007789 gas Substances 0.000 description 4
- 230000006399 behavior Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000009792 diffusion process Methods 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 239000000779 smoke Substances 0.000 description 2
- CBENFWSGALASAD-UHFFFAOYSA-N Ozone Chemical compound [O-][O+]=O CBENFWSGALASAD-UHFFFAOYSA-N 0.000 description 1
- 239000000443 aerosol Substances 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000000701 chemical imaging Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/01—Arrangements or apparatus for facilitating the optical investigation
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
- G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
- G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A50/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE in human health protection, e.g. against extreme weather
- Y02A50/20—Air quality improvement or preservation, e.g. vehicle emission control or emission reduction by using catalytic converters
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Remote Sensing (AREA)
- Chemical & Material Sciences (AREA)
- Pathology (AREA)
- Immunology (AREA)
- Biochemistry (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Analytical Chemistry (AREA)
- Biomedical Technology (AREA)
- Astronomy & Astrophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Biodiversity & Conservation Biology (AREA)
- Artificial Intelligence (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a system for positioning and analyzing an atmospheric pollution source region. A first spectral diagram is obtained. And extracting spectral features based on the secondary detection spectrogram and the secondary detection pollution monitoring image through a spectrum fusion network, and combining the image features to obtain a pollution area. And establishing a relation between the pollution concentration of the pollution area and the color of the secondary detection pollution monitoring image, and dividing the pollution area by dividing the neural network to obtain a divided pollution area image. And obtaining a plurality of pollution concentrations based on the plurality of pollution blocks corresponding to the segmented pollution area image. And (5) monitoring twice to prevent missed detection. The spectrum fusion network outputs a three-dimensional map. The detection point information on the ground is used for carrying out initial segmentation through closed point lines according to the clustering detection point information, and then the closed point lines and the closed point lines are combined through a segmentation neural network, so that more accurate pollution blocks and pollution center points are obtained.
Description
Technical Field
The invention relates to the technical field of computers, in particular to an atmospheric pollution source area positioning analysis method and system.
Background
Atmospheric pollution can be detected by satellite spectrum because the exhausted gas is detected by different bands of light with different information. However, the spectrum needs to emit light from all positions to obtain detailed region positions, or detection is inaccurate or a lot of material resources are consumed. The detection can also be performed by on-line detection, but a large amount of instruments and equipment are required, so that the cost is high.
Meanwhile, when the pollution is detected in the area, the specific pollution area containing irregular boundaries cannot be accurately detected due to the diffusion characteristic of the gas. And the pollution area cannot be divided according to the concentration and the pollution center point cannot be detected, so that the atmospheric pollution source cannot be accurately positioned.
Disclosure of Invention
The invention aims to provide a method and a system for positioning and analyzing an atmospheric pollution source area, 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 positioning and analyzing an atmospheric pollution source area, including:
obtaining a first spectrogram; the first spectrogram includes features of information for a plurality of regions;
according to the first spectrogram, performing secondary detection through a local detection neural network, and detecting a local area to obtain a secondary detection spectrogram and a secondary detection pollution monitoring image;
Extracting spectral features based on the secondary detection spectrogram and the secondary detection pollution monitoring image through a spectrum fusion network, and combining the image features to obtain a pollution area;
establishing a relation between the pollution concentration of the pollution area and the color of the secondary detection pollution monitoring image, and dividing the pollution area through a division neural network to obtain a divided pollution area image; the split pollution area image comprises a plurality of pollution blocks with different pollution degrees and a pollution center point;
obtaining a plurality of pollution concentrations based on a plurality of pollution blocks corresponding to the split pollution area image; one pollution concentration corresponds to one pollution block;
and replacing the values in the corresponding split pollution area images with the plurality of pollution concentrations, marking the pollution center point, and obtaining a pollution source area positioning image.
Optionally, the extracting spectral features and combining image features to obtain a polluted region based on the secondary detection spectrogram and the secondary detection pollution monitoring image through a spectrum fusion network includes:
inputting the secondary detection pollution monitoring image into an image convolution network, extracting image characteristics and obtaining a pollution monitoring characteristic diagram; the size of the pollution monitoring characteristic diagram is the same as that of the spectrum characteristic diagram;
The size of the convolution kernel of the last layer of the image convolution network is the same as that of the spectrum characteristic diagram;
superposing and fusing the secondary detection spectrogram and the pollution monitoring feature image to obtain a fused feature image;
inputting the fusion feature map into a spectrum fusion network, extracting features, detecting boundaries and obtaining a pollution area;
the spectrum fusion network includes a first convolutional network, a second convolutional network, and a third convolutional network.
Optionally, the establishing a relationship between the pollution concentration of the pollution area and the color of the secondary detection pollution monitoring image, and dividing the pollution area by dividing the neural network to obtain a divided pollution area image includes:
obtaining a plurality of detection point information; the detection point information is the concentration of the pollution detected in the pollution area;
clustering the detection point information to obtain a plurality of clustering sets; the values in the cluster set represent values of the same degree of contamination;
connecting values in a plurality of clustering sets at positions corresponding to the secondary detection pollution monitoring images for a plurality of times to obtain a rough region image;
inputting the rough area image and the secondary pollution detection monitoring image into a segmentation neural network, and predicting concentration distribution conditions to obtain a segmentation pollution area image.
Optionally, the training method of the spectrum fusion network includes:
obtaining a training set; the training set comprises a plurality of training feature graphs and a plurality of corresponding labeling data; the training feature map is a fusion feature map obtained by fusing a spectrum feature map and a pollution monitoring feature map; the annotation data comprises an annotation graph and an annotation vector; the labeling graph is a three-dimensional graph; the labeling graph is a binary graph in which a pollution area is labeled as 1 according to the pollution type and other areas are set as 0; the labeling vector represents whether the pollution center point exists and the existence position;
inputting the training feature map into a first convolution network, extracting fusion features and obtaining a segmentation feature map;
inputting the segmentation feature map into a second convolution network, extracting segmentation features, and carrying out multidimensional judgment to obtain a second convolution map; the second convolution map is three-dimensional; the values of the second convolution map represent whether a contaminant class of a region is contaminated;
inserting a numerical value into the second convolution map through an interpolation method, and converting the second convolution feature map into a feature map of a pollution area, wherein the size of the feature map is the same as that of the spectrum map;
inputting the segmentation feature map into a third convolution network, extracting segmentation features, superposing the features into one dimension, and classifying to obtain a discrimination vector; the value in the discrimination vector represents whether the pollution center point exists or not and the position of the pollution center point;
Obtaining losses from the characteristic map and the labeling map of the polluted area to obtain a first loss value;
solving the loss of the discrimination vector and the labeling vector to obtain a second loss value;
adding the first loss value and the second loss value to obtain a loss value;
obtaining the current training iteration times of a spectrum fusion network and the preset maximum iteration times of the spectrum fusion 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 spectrum fusion network.
Optionally, the inputting the rough area image and the secondary detection pollution monitoring image into a segmented neural network, predicting concentration distribution conditions, and obtaining segmented pollution area images includes:
superposing the rough area image and the secondary pollution detection monitoring image to obtain a three-dimensional array;
inputting the three-dimensional array into a segmentation neural network, extracting features, and obtaining a segmentation pollution area image; the image of the partitioned polluted area is an image for dividing the polluted image into a plurality of polluted blocks;
the segmented neural network is obtained by training a plurality of training three-dimensional arrays and labeling a segmented polluted area graph; the labeling segmentation contaminated region map is an image that divides the contaminated region into a plurality of contaminated blocks.
Optionally, the obtaining a plurality of pollution concentrations based on the plurality of pollution blocks corresponding to the segmented pollution area image includes:
extracting detection point information corresponding to the pollution blocks of the split pollution area image to obtain a plurality of detection concentrations;
averaging the plurality of detection concentrations to obtain a pollution concentration;
and extracting detection point information corresponding to a plurality of pollution blocks of the split pollution area image for a plurality of times, and then averaging to obtain a plurality of pollution concentrations.
Optionally, the performing secondary detection through the local detection neural network according to the first spectrogram, detecting a local area, and obtaining a secondary detection spectrogram and a secondary detection pollution monitoring image includes:
inputting the first spectrogram into a local detection neural network, and extracting spectral features to obtain a first spectral feature map;
comparing the values of the elements in the first spectrogram with spectrum thresholds respectively, and judging whether the values are in a set pollution range or not to obtain a first comparison chart; the values in the first ratio graph represent whether there is contamination and a type of contamination in different areas;
performing OR operation on the values in the first spectrum characteristic diagram and the first ratio diagram to obtain a first pollution diagram;
If pollution exists in the first pollution map, dividing the pollution area into a plurality of areas with the same size, and performing spectrum detection to obtain a secondary detection spectrum image and a secondary detection pollution monitoring image.
In a second aspect, an embodiment of the present invention provides an atmospheric pollution source area positioning analysis system, including:
the acquisition module is used for: obtaining a first spectrogram; the first spectrogram includes features of information for a plurality of regions; the information of the region is obtained by emitting light rays to regions with the same size in a plurality of atmosphere layers for a plurality of times through a spectrometer;
and a secondary detection module: according to the first spectrogram, performing secondary detection through a local detection neural network, and detecting a local area to obtain a secondary detection spectrogram and a secondary detection pollution monitoring image;
and a spectrum fusion module: extracting spectral features based on the secondary detection spectrogram and the secondary detection pollution monitoring image through a spectrum fusion network, and combining the image features to obtain a pollution area;
region segmentation module: establishing a relation between the pollution concentration of the pollution area and the color of the secondary detection pollution monitoring image, and dividing the pollution area through a division neural network to obtain a divided pollution area image; the split pollution area image comprises a plurality of pollution blocks with different pollution degrees and a pollution center point;
The concentration calculation module: obtaining a plurality of pollution concentrations based on a plurality of pollution blocks corresponding to the split pollution area image; one pollution concentration corresponds to one pollution block;
and a marking module: and replacing the values in the corresponding split pollution area images with the plurality of pollution concentrations, marking the pollution center point, and obtaining a pollution source area positioning image.
Optionally, the extracting spectral features and combining image features to obtain a polluted region based on the secondary detection spectrogram and the secondary detection pollution monitoring image through a spectrum fusion network includes:
inputting the secondary detection pollution monitoring image into an image convolution network, extracting image characteristics and obtaining a pollution monitoring characteristic diagram; the size of the pollution monitoring characteristic diagram is the same as that of the spectrum characteristic diagram;
the size of the convolution kernel of the last layer of the image convolution network is the same as that of the spectrum characteristic diagram;
superposing and fusing the spectrum characteristic diagram and the pollution monitoring characteristic diagram to obtain a fused characteristic diagram;
inputting the fusion feature map into a spectrum fusion network, extracting features, detecting boundaries and obtaining a pollution area;
the spectrum fusion network includes a first convolutional network, a second convolutional network, and a third convolutional network.
Optionally, the establishing a relationship between the pollution concentration of the pollution area and the color of the secondary detection pollution monitoring image, and dividing the pollution area by dividing the neural network to obtain a divided pollution area image includes:
obtaining a plurality of detection point information; the detection point information is the concentration of the pollution detected in the pollution area;
clustering the detection point information to obtain a plurality of clustering sets; the values in the cluster set represent values of the same degree of contamination;
connecting values in a plurality of clustering sets at positions corresponding to the secondary detection pollution monitoring images for a plurality of times to obtain a rough region image;
inputting the rough area image and the secondary pollution detection monitoring image into a segmentation neural network, and predicting concentration distribution conditions to obtain a segmentation pollution area image.
Compared with the prior art, the embodiment of the invention achieves the following beneficial effects:
the embodiment of the invention also provides a method and a system for positioning and analyzing the air pollution source area, wherein the method comprises the following steps: a first spectral diagram is obtained. The first spectrogram includes features of information for a plurality of regions. And according to the first spectrogram, performing secondary detection through a local detection neural network, and detecting a local area to obtain a secondary detection spectrogram and a secondary detection pollution monitoring image. And extracting spectral features based on the secondary detection spectrogram and the secondary detection pollution monitoring image through a spectrum fusion network, and combining the image features to obtain a pollution area. And establishing a relation between the pollution concentration of the pollution area and the color of the secondary detection pollution monitoring image, and dividing the pollution area by dividing the neural network to obtain a divided pollution area image. The segmented contamination region image includes a plurality of contamination patches of varying contamination levels and a contamination center point. And obtaining a plurality of pollution concentrations based on the plurality of pollution blocks corresponding to the segmented pollution area image. One pollution concentration corresponds to one pollution block. And replacing the values in the corresponding split pollution area images with the plurality of pollution concentrations, marking the pollution center point, and obtaining a pollution source area positioning image.
And (5) monitoring twice to prevent missed detection. And firstly, using spectrum and image fusion to judge whether pollution exists in the area. The spectrum is used to assist the image rather than simply fusing, and the detection of colorless contaminant gases by the image is corrected. The spectrum fusion network outputs a three-dimensional graph for detecting pollution areas of different pollution types. The contaminated area can be detected more accurately.
The polluted area obtained by combining the detection point information with the atmospheric pollution is initially segmented by closing the dotted line according to the clustering detection point information, and then the polluted area is segmented by a segmentation neural network, and the two are combined to obtain more accurate polluted blocks and a polluted central point. And the detection concentration is obtained through the detection point information in the pollution block, so that the air pollution area can be conveniently identified and used later according to the concentration information when the air pollution area is positioned. Thus, the air pollution areas are accurately positioned, and the concentration of each positioning area is recorded.
In summary, the invention performs spectrum secondary detection according to the spectrum and the satellite photo partition, and simultaneously performs fusion of a feature, increases feature information, and finds out a pollution area with obvious boundary division. And detecting a region with pollution concentration according to the dispersed detection points by the pollution region, predicting the region blocks according to the distribution condition, dividing the region blocks into a plurality of pollution blocks with similar concentration, and predicting to obtain a central detection point.
Drawings
FIG. 1 is a flow chart of a method for positioning and analyzing an atmospheric pollution source area according to an embodiment of the present 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.
Description of the embodiments
The present invention will be described in detail with reference to the accompanying drawings.
Example 1
As shown in fig. 1, an embodiment of the present invention provides a method for analyzing location of an atmospheric pollution source, where the method includes:
s101: a first spectral diagram is obtained. The first spectrogram includes features of information for a plurality of regions.
The information of the region is obtained by emitting light to regions with the same size in a plurality of atmosphere layers through a spectrometer.
The first spectrogram is a graph marked by establishing a coordinate axis by taking the lower left corner as an origin. The values on the coordinate axes correspond to the distances of the actual positions.
The area to be detected is divided into a rectangular area, and the rectangular area is uniformly divided according to fixed length and height to obtain a plurality of identical areas. The spectrometer emits light to a plurality of identical regions to obtain a plurality of spectral features, and a first spectrogram is formed.
Wherein the first spectrogram is used for detecting pollution conditions in a large-range area and finding a small-area part of pollution.
S102: and according to the first spectrogram, performing secondary detection through a local detection neural network, and detecting a local area to obtain a secondary detection spectrogram and a secondary detection pollution monitoring image.
And detecting the first spectrum for the first time to obtain a local area with pollution, wherein the local area is a divided rectangular area. And performing secondary detection on the rectangular area, and finding out a spectrum image formed by dividing the rectangular area into corresponding spectrum values of smaller rectangular areas and an image shot by the image pickup equipment of the rectangular area to obtain a secondary detection spectrum image and a secondary detection pollution monitoring image.
S103: and extracting spectral features based on the secondary detection spectrogram and the secondary detection pollution monitoring image through a spectrum fusion network, and combining the image features to obtain a pollution area.
The content of important trace gases and aerosols such as sulfur dioxide, ozone, nitrogen dioxide, methane, carbon monoxide and the like in the atmosphere can be analyzed through a multispectral imaging spectrometer. In this example, the spectral information of sulfur dioxide, nitrogen dioxide and carbon monoxide is detected.
Wherein, the spectrum information is combined with the image to detect irregular pollution areas.
S104: and establishing a relation between the pollution concentration of the pollution area and the color of the secondary detection pollution monitoring image, and dividing the pollution area by dividing the neural network to obtain a divided pollution area image.
Wherein the contaminated block represents a different region divided into contaminated regions according to a degree of contamination.
Wherein the contamination center point represents a point where the contamination level is greater than other contamination locations.
Wherein, according to different pollution degrees, find the gradient change of color. The approximate region position is found from the color gradient change.
And establishing a relation between the monitored concentration and the color of the image to obtain the image, and predicting the accurate position. To address this behavior, a center point of contamination that may be a site of contamination is found due to the diffusion behavior of smoke.
S105: obtaining a plurality of pollution concentrations based on a plurality of pollution blocks corresponding to the split pollution area image; one pollution concentration corresponds to one pollution block.
Wherein the concentration is obtained according to the approximate region position. Dividing it into regions of different concentrations.
S106: and replacing the values in the corresponding split pollution area images with the plurality of pollution concentrations, marking the pollution center point, and obtaining a pollution source area positioning image.
The method comprises the steps of dividing a pollution area image into different pollution blocks according to pollution levels, calculating pollution concentration, and replacing the value of the position of the pollution block with the corresponding pollution concentration.
Optionally, the extracting spectral features and combining image features to obtain a polluted region based on the secondary detection spectrogram and the secondary detection pollution monitoring image through a spectrum fusion network includes:
inputting the secondary detection pollution monitoring image into an image convolution network, extracting image characteristics and obtaining a pollution monitoring characteristic diagram; the size of the pollution monitoring feature is the same as the size of the spectral feature.
The size of the convolution kernel of the last layer of the image convolution network is the same as that of the spectrum characteristic diagram;
and superposing and fusing the secondary detection spectrogram and the pollution monitoring feature image to obtain a fused feature image.
In this embodiment, the spectrum feature map size is 28x28x1, the pollution monitoring feature map size is 28x28x255, and the secondary detection spectrum map and the pollution monitoring feature map are overlapped and fused to obtain a fused feature map size of 28x28x256.
And inputting the fusion feature map into a spectrum fusion network, extracting features, detecting boundaries and obtaining a pollution area.
The spectrum fusion network includes a first convolutional network, a second convolutional network, and a third convolutional network.
By the method, a region is detected by using a spectrometer, and the part of the region with pollution can be obtained according to the spectrum. The contaminated area is obtained through the spectrum and the image, and the contaminated area can also be directly obtained through input. The image can be characterized, and then the spectral characteristics of the position monitored by the spectrometer are combined to carry out characteristic judgment. Emphasis is placed on using spectra to assist the image rather than simple fusion.
Optionally, the establishing a relationship between the pollution concentration of the pollution area and the color of the secondary detection pollution monitoring image, and dividing the pollution area by dividing the neural network to obtain a divided pollution area image includes:
obtaining a plurality of detection point information; the detection point information is the concentration of the pollution detected in the pollution area.
And clustering the detection point information to obtain a plurality of clustering sets. The values in the cluster set represent values of the same degree of contamination.
And connecting the values in the plurality of clustering sets at the positions corresponding to the secondary detection pollution monitoring images for a plurality of times to obtain a rough region image.
Wherein, a two-dimensional matrix with the same size as the secondary detection pollution monitoring image and value of 0 is obtained; and changing the value in the clustering set corresponding to the position in the boundary region, namely the value in the position corresponding to the secondary detection pollution monitoring image from 0 to the value in the clustering set, so as to obtain the approximate region image. In this embodiment, the rough area image is 224x224x1.
Inputting the rough area image and the secondary pollution detection monitoring image into a segmentation neural network, and predicting concentration distribution conditions to obtain a segmentation pollution area image.
By the method, the pollution area is not easy to find the boundary or to divide the pollution area by distance through the boundary vertical, so that the actual relationship is established according to the concentration relationship obtained by monitoring each position.
Optionally, the training method of the spectrum fusion network includes:
the training method of the image segmentation network comprises the following steps:
obtaining a training set; the training set comprises a plurality of training feature graphs and a plurality of corresponding labeling data; the training feature map is a fusion feature map obtained by fusing a spectrum feature map and a pollution monitoring feature map; the annotation data comprises an annotation graph and an annotation vector; the labeling graph is a three-dimensional graph; the labeling graph is a binary graph for labeling a polluted area as 1 and setting the area outside the polluted area as 0 according to the labeling; the annotation vector indicates whether the center point of contamination exists and where it exists.
In this embodiment, the labeling chart sets the area where nitrogen dioxide is located in the chart corresponding to the first dimension to be 1, and sets other areas to be 0. The region where carbon monoxide corresponding to the second dimension is located is set to 1, and the other regions are set to 0 in the figure. The region where sulfur dioxide corresponding to the third dimension is located is set to 1, and the other regions are set to 0 in the figure.
And inputting the training feature map into a first convolution network, extracting fusion features, and obtaining a segmentation feature map.
Wherein the first convolutional network is a convolutional neural network (Convolutional Neural Networks, CNN).
And inputting the segmentation feature map into a second convolution network, extracting segmentation features, and carrying out multidimensional judgment to obtain a second convolution map. The second convolution map is three-dimensional. The values of the second convolution map represent whether a contaminant class of an area is contaminated.
And performing feature extraction by using a convolution check segmentation feature map in the second convolution network. In this embodiment, the final convolution kernel is 3 convolution kernels of 7x7 size, so the second convolution map obtained in this embodiment is a three-dimensional map of 7x7x 3. The first dimension indicates whether there is nitrogen dioxide pollution in 7x7, i.e. 49 zones, the second dimension indicates whether there is carbon monoxide pollution in 7x7, i.e. 49 zones, and the third dimension indicates whether there is sulfur dioxide pollution in 7x7, i.e. 49 zones.
And inserting values into the second convolution map through an interpolation method, and converting the second convolution characteristic map into a characteristic map of the pollution area, wherein the size of the characteristic map is the same as that of the spectrum map.
In this embodiment, interpolation is performed using a bilinear interpolation algorithm.
In this embodiment, the convolution map of the 7x7x3 size is converted into a contaminated area feature map of 224x224x 3. The contaminated area feature map represents an image divided into 224x224 sized areas, each area having three judgment values, respectively representing the presence or absence of nitrogen dioxide, carbon monoxide and sulfur dioxide contamination.
Inputting the segmentation feature map into a third convolution network, extracting segmentation features, and classifying the features after superimposing the features into one dimension to obtain a discrimination vector. The values in the discrimination vector indicate whether a contamination center point exists and the location of the contamination center point.
The third convolutional network is a convolutional neural network (Convolutional Neural Networks, CNN), features are extracted through the third convolutional network, and finally, only a convolutional kernel with the same size as the feature map is used for converting the segmented feature map into one-dimensional feature vectors, and the one-dimensional feature vectors are classified through a softmax function to obtain discrimination vectors with the size of 1x 3. The first element of the discrimination vector represents whether a pollution center point exists, the second element of the discrimination vector represents the abscissa of the pollution center point, and the third element of the discrimination vector represents the ordinate of the pollution center point.
And obtaining the loss from the characteristic map and the labeling map of the polluted area to obtain a first loss value.
In this embodiment, the contaminated area feature map and the labeling map are respectively elongated to be one-dimensional vectors, and the loss value is calculated by using a cross entropy loss function.
And solving the loss of the discrimination vector and the labeling vector to obtain a second loss value.
And solving the loss of the discrimination vector and the labeling vector through a cross entropy loss function.
And adding the first loss value and the second loss value to obtain a loss value.
Obtaining the current training iteration times of the spectrum fusion network and the preset maximum iteration times of the spectrum fusion network training.
The preset maximum iteration number of the spectrum fusion network training is 12000.
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 spectrum fusion network.
In this embodiment, the threshold is 0.9.
By the above method, spectrum acquisition is performed for each minute portion. Because the actual image may also appear as smoke, but sometimes color monitoring cannot be performed, but error monitoring is provided, so that the judgment is more accurate by fusing the spectrum information. In the training stage, the network tries to divide all the image blocks, and even if targets do not exist in the image blocks, prediction is divided, so that judgment is performed, and whether the judgment is the same as the judgment of the label passing loss is performed, so that a more accurate network is obtained.
Optionally, the inputting the rough area image and the secondary detection pollution monitoring image into a segmented neural network, predicting concentration distribution conditions, and obtaining segmented pollution area images includes:
and superposing the rough area image and the secondary pollution detection monitoring image to obtain a three-dimensional array. The three-dimensional array includes a general area image and a secondary detection contamination monitoring image.
Inputting the three-dimensional array into a segmentation neural network, extracting features, and obtaining a segmentation pollution area image. The segmented contaminated region image is an image that divides the contaminated image into a plurality of contaminated blocks.
The segmented neural network is obtained by training a plurality of training three-dimensional arrays and labeling a segmented polluted area graph; the labeling segmentation contaminated region map is an image that divides the contaminated region into a plurality of contaminated blocks.
According to the method, the value of the corresponding position in the secondary detection pollution monitoring image is extracted according to the position in the approximate area image, and a new image is obtained. The image of the separated pollution area is an image which corresponds to the pollution area and can reflect pollution conditions. The importance of the spectral and image monitoring can be obtained at this time. After the rough region is obtained, the region is divided.
Optionally, the obtaining a plurality of pollution concentrations based on the plurality of pollution blocks corresponding to the segmented pollution area image includes:
and extracting detection point information corresponding to the pollution blocks of the split pollution area image to obtain a plurality of detection concentrations.
However, since the detection point information is not divided on average when dividing the polluted block, there is a possibility that one polluted block exists in the detection point information in two clusters, and the detection concentration is to be obtained again.
The plurality of detected concentrations are averaged to obtain a pollution concentration.
And extracting detection point information corresponding to a plurality of pollution blocks of the split pollution area image for a plurality of times, and then averaging to obtain a plurality of pollution concentrations.
Optionally, the performing secondary detection through the local detection neural network according to the first spectrogram, detecting a local area, and obtaining a secondary detection spectrogram and a secondary detection pollution monitoring image includes:
and inputting the first spectrogram into a local detection neural network, extracting spectral characteristics, and obtaining a first spectral characteristic graph.
The local detection neural network takes a spectrogram as input and marked pollution points as input trained images.
The first spectrogram is three-dimensional, the first dimension represents the abscissa position of the first spectrogram, the first dimension represents the ordinate position of the first spectrogram, and the third dimension represents a plurality of detection values of the position determined by the abscissa and the ordinate of the first spectrogram. The peak value, half-width and variation trend detected by different wavelengths are shown in the embodiment.
In this embodiment, the local detection neural network is a convolutional neural network (Convolutional Neural Networks, CNN).
The first dimension of the first spectral feature map represents the abscissa position of the corresponding first spectral map, and the first dimension represents the ordinate position of the corresponding first spectral map. The third dimension indicates whether or not there is contamination and the type of contamination, as in this embodiment, the number of third dimensions is 4, one indicates whether or not there is contamination, one indicates whether or not there is excess nitrogen dioxide, one indicates whether or not there is excess carbon monoxide, and one indicates whether or not there is excess sulfur dioxide.
And comparing the values of the elements in the first spectrogram with spectrum thresholds respectively, and judging whether the values are in a set pollution range or not to obtain a first comparison chart. The values in the first ratio graph indicate whether there is contamination and type of contamination in different areas.
The spectral threshold in this embodiment is 0.7.
The peak value, the half-width and the variation trend detected by using the light with different wavelengths in the first spectrogram are compared with the peak value, the half-width and the variation trend of different pollution types under different wavelengths, so that a plurality of comparison results are obtained, wherein the comparison results have 4 values, one value indicates whether pollution exists, one value indicates whether nitrogen dioxide is too high, one value indicates whether carbon monoxide is too high, and one value indicates whether sulfur dioxide is too high. The comparison results of the different regions constitute a first ratio map.
And performing OR operation on the values in the first spectrum characteristic diagram and the first ratio diagram to obtain a first pollution diagram.
Wherein the pollution map indicates whether or not each area has pollution and the type of pollution.
If pollution exists in the first pollution map, dividing the pollution area into a plurality of areas with the same size, and performing spectrum detection to obtain a secondary detection spectrum image and a secondary detection pollution monitoring image.
The secondary detection spectrogram and the secondary detection pollution monitoring image are shot well, and whether the secondary detection spectrogram and the secondary detection pollution monitoring image are acquired is judged according to whether pollution exists or not.
The secondary detection spectral image is obtained by cutting a first pollution area into a fixed position and emitting light to the fixed position; the secondary detection pollution monitoring image is an image of a first pollution area shot by a camera. A spectral signature is obtained. The spectrometer characteristic diagram is obtained by emitting light to different areas.
By the method, the first spectral feature map and the secondary detection spectral image are detected twice to prevent missed detection. The spot is located at a position to obtain a size which is a common size of the divided area and contains the spot, and a contaminated area is obtained.
Example 2
Based on the above-mentioned method for positioning and analyzing the atmospheric pollution source region, the embodiment of the invention also provides a system for positioning and analyzing the atmospheric pollution source region, which comprises an acquisition module, a secondary detection module, a spectrum fusion module, a region segmentation module and a marking module.
The acquisition module is used for acquiring a first spectrogram; the first spectrogram includes features of information for a plurality of regions. The information of the region is obtained by emitting light rays to regions with the same size in a plurality of atmosphere layers through a spectrometer.
The secondary detection module is used for carrying out secondary detection through the local detection neural network according to the first spectrogram, detecting the local area and obtaining a secondary detection spectrogram and a secondary detection pollution monitoring image.
The spectrum fusion module is used for extracting spectrum characteristics and combining image characteristics through a spectrum fusion network based on the secondary detection spectrogram and the secondary detection pollution monitoring image to obtain a pollution area.
The region segmentation module is used for establishing a relation between the pollution concentration of the pollution region and the color of the secondary detection pollution monitoring image, and segmenting the pollution region through a segmentation neural network to obtain a segmented pollution region image. The segmented contamination region image includes a plurality of contamination patches of varying contamination levels and a contamination center point.
The concentration calculation module is used for obtaining a plurality of pollution concentrations based on a plurality of pollution blocks corresponding to the split pollution area images; one pollution concentration corresponds to one pollution block.
The marking module is used for replacing the values in the corresponding split pollution area images with the plurality of pollution concentrations, marking the pollution center point and obtaining a pollution source area positioning image.
Optionally, the extracting spectral features and combining image features to obtain a polluted region based on the secondary detection spectrogram and the secondary detection pollution monitoring image through a spectrum fusion network includes:
inputting the secondary detection pollution monitoring image into an image convolution network, extracting image characteristics and obtaining a pollution monitoring characteristic diagram. The size of the pollution monitoring feature is the same as the size of the spectral feature.
The size of the convolution kernel of the last layer of the image convolution network is the same as that of the spectrum characteristic diagram.
And superposing and fusing the spectrum characteristic diagram and the pollution monitoring characteristic diagram to obtain a fused characteristic diagram.
And inputting the fusion feature map into a spectrum fusion network, extracting features, detecting boundaries and obtaining a pollution area.
The spectrum fusion network includes a first convolutional network, a second convolutional network, and a third convolutional network.
Optionally, the establishing a relationship between the pollution concentration of the pollution area and the color of the secondary detection pollution monitoring image, and dividing the pollution area by dividing the neural network to obtain a divided pollution area image includes:
a plurality of detection point information is obtained. The detection point information is the concentration of the pollution detected in the pollution area.
Clustering the detection point information to obtain a plurality of clustering sets; the values in the cluster set represent values of the same degree of contamination.
And connecting the values in the plurality of clustering sets at the positions corresponding to the secondary detection pollution monitoring images for a plurality of times to obtain a rough region image.
Inputting the rough area image and the secondary pollution detection monitoring image into a segmentation neural network, and predicting concentration distribution conditions to obtain a segmentation pollution area image.
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.
An 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 processor 502 implements the steps of any one of the above-described methods for positioning and analyzing an atmospheric pollution source region when executing 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.
Embodiments of the present invention also provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of any of the methods of the above-described method of atmospheric pollution source area localization analysis and the data referred to above.
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. However, the disclosed method should not be construed as reflecting the intention that: i.e., the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.
Those skilled in the art will appreciate that the modules in the apparatus of the embodiments may be adaptively changed and disposed in one or more apparatuses different from the embodiments. The modules or units or components of the embodiments may be combined into one module or unit or component and, furthermore, they may be divided into a plurality of sub-modules or sub-units or sub-components. Any combination of all features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or units of any method or apparatus so disclosed, may be used in combination, except insofar as at least some of such features and/or processes or units are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.
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. For example, in the following claims, any of the claimed embodiments can be used in any combination.
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.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, third, etc. do not denote any order. These words may be interpreted as names.
Claims (5)
1. An atmospheric pollution source area positioning analysis method is characterized by comprising the following steps:
obtaining a first spectrogram; the first spectrogram includes features of information for a plurality of regions;
according to the first spectrogram, performing secondary detection through a local detection neural network, and detecting a local area to obtain a second spectrogram and a second pollution monitoring image;
based on the second spectrogram and the second pollution monitoring image, extracting spectral characteristics through a spectral fusion network, and combining the image characteristics to obtain a pollution area;
establishing a relation between the pollution concentration of the pollution area and the color of the second pollution monitoring image, and dividing the pollution area through a dividing neural network to obtain a divided pollution area image; the split pollution area image comprises a plurality of pollution blocks with different pollution degrees and a pollution center point;
obtaining a plurality of pollution concentrations based on a plurality of pollution blocks corresponding to the split pollution area image; one pollution concentration corresponds to one pollution block;
inputting the plurality of pollution concentrations into the positions of corresponding pollution blocks in the split pollution area image, and marking a pollution center point to obtain a pollution source area positioning image;
the step of extracting spectral features and combining image features to obtain a polluted region through a spectrum fusion network based on the second spectrogram and the second pollution monitoring image comprises the following steps:
Inputting the second pollution monitoring image into an image convolution network, extracting image features and obtaining a pollution monitoring feature map; the size of the pollution monitoring characteristic diagram is the same as that of the spectrum characteristic diagram;
the size of the convolution kernel of the last layer of the image convolution network is the same as that of the spectrum characteristic diagram;
superposing and fusing the second spectrogram and the pollution monitoring feature image to obtain a fused feature image;
inputting the fusion feature map into a spectrum fusion network, extracting features, detecting boundaries and obtaining a pollution area;
the spectrum fusion network comprises a first convolution network, a second convolution network and a third convolution network;
establishing a relation between the pollution concentration of the pollution area and the color of the second pollution monitoring image, and dividing the pollution area by dividing the neural network to obtain a divided pollution area image, wherein the method comprises the following steps:
obtaining a plurality of detection point information; the detection point information is the concentration of the pollution detected in the pollution area;
clustering the detection point information to obtain a plurality of clustering sets; the values in the cluster set represent values of the same degree of contamination;
connecting values in the plurality of clustering sets at positions corresponding to the second pollution monitoring images for a plurality of times to obtain a rough area image;
Inputting the rough area image and the second pollution monitoring image into a segmentation neural network, and predicting concentration distribution conditions to obtain a segmentation pollution area image;
performing secondary detection by using a local detection neural network according to the first spectrogram, and detecting a local area to obtain a second spectrogram and a second pollution monitoring image, wherein the method comprises the following steps:
inputting the first spectrogram into a local detection neural network, and extracting spectral features to obtain a first spectral feature map;
comparing the values of the elements in the first spectrogram with threshold values respectively, and judging whether the values are in a set pollution range or not to obtain a first comparison value chart; the values in the first ratio graph represent whether there is contamination and a type of contamination in different areas;
performing OR operation on the values in the first spectrum characteristic diagram and the first ratio diagram to obtain a first pollution diagram;
if pollution exists in the first pollution map, dividing the pollution area into a plurality of areas with the same size, and performing spectrum detection to obtain a second spectrum image and a second pollution monitoring image.
2. The method for positioning and analyzing an atmospheric pollution source area according to claim 1, wherein the training method of the spectrum fusion network comprises the following steps:
Obtaining a training set; the training set comprises a plurality of training feature graphs and a plurality of corresponding labeling data; the training feature map is a fusion feature map obtained by fusing a spectrum feature map and a pollution monitoring feature map; the annotation data comprises an annotation graph and an annotation vector; the labeling graph is a three-dimensional graph; the labeling graph is a binary graph in which a pollution area is labeled as 1 according to the pollution type and other areas are set as 0; the labeling vector represents whether the pollution center point exists and the existence position;
inputting the training feature map into a first convolution network, extracting fusion features and obtaining a segmentation feature map;
inputting the segmentation feature map into a second convolution network, extracting segmentation features, and carrying out multidimensional judgment to obtain a second convolution map; the second convolution map is three-dimensional; the values of the second convolution map represent whether a contaminant class of a region is contaminated;
inserting a numerical value into the second convolution map through an interpolation method, and converting the second convolution feature map into a feature map of a pollution area, wherein the size of the feature map is the same as that of the spectrum map;
inputting the segmentation feature map into a third convolution network, extracting segmentation features, superposing the features into one dimension, and classifying to obtain a discrimination vector; the value in the discrimination vector represents whether the pollution center point exists or not and the position of the pollution center point;
Obtaining losses from the characteristic map and the labeling map of the polluted area to obtain a first loss value;
solving the loss of the discrimination vector and the labeling vector to obtain a second loss value;
adding the first loss value and the second loss value to obtain a loss value;
obtaining the current training iteration times of a spectrum fusion network and the preset maximum iteration times of the spectrum fusion 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 spectrum fusion network.
3. The method of claim 1, wherein inputting the general area image and the second pollution monitoring image into a segmented neural network, predicting a concentration distribution condition, and obtaining a segmented pollution area image comprises:
superposing the rough area image and the second pollution monitoring image to obtain a three-dimensional array;
inputting the three-dimensional array into a segmentation neural network, extracting features, and obtaining a segmentation pollution area image; the image of the partitioned polluted area is an image for dividing the polluted image into a plurality of polluted blocks;
The segmented neural network is obtained by training a plurality of training three-dimensional arrays and labeling a segmented polluted area graph; the labeling segmentation contaminated region map is an image that divides the contaminated region into a plurality of contaminated blocks.
4. The method for positioning and analyzing an atmospheric pollution source according to claim 1, wherein the obtaining a plurality of pollution concentrations based on a plurality of pollution blocks corresponding to the segmented pollution region image comprises:
extracting detection point information corresponding to the pollution blocks of the split pollution area image to obtain a plurality of detection concentrations;
averaging the plurality of detection concentrations to obtain a pollution concentration;
and extracting detection point information corresponding to a plurality of pollution blocks of the split pollution area image for a plurality of times, and then averaging to obtain a plurality of pollution concentrations.
5. An atmospheric pollution source area location analysis system, comprising:
the acquisition module is used for: obtaining a first spectrogram; the first spectrogram includes features of information for a plurality of regions; the information of the region is obtained by emitting light rays to regions with the same size in a plurality of atmosphere layers for a plurality of times through a spectrometer;
and a secondary detection module: according to the first spectrogram, performing secondary detection through a local detection neural network, and detecting a local area to obtain a second spectrogram and a second pollution monitoring image;
And a spectrum fusion module: based on the second spectrogram and the second pollution monitoring image, extracting spectral characteristics through a spectral fusion network, and combining the image characteristics to obtain a pollution area;
region segmentation module: establishing a relation between the pollution concentration of the pollution area and the color of the second pollution monitoring image, and dividing the pollution area through a dividing neural network to obtain a divided pollution area image; the split pollution area image comprises a plurality of pollution blocks with different pollution degrees and a pollution center point;
the concentration calculation module: obtaining a plurality of pollution concentrations based on a plurality of pollution blocks corresponding to the split pollution area image; one pollution concentration corresponds to one pollution block;
and a marking module: inputting the plurality of pollution concentrations into the positions of corresponding pollution blocks in the split pollution area image, and marking a pollution center point to obtain a pollution source area positioning image;
the step of extracting spectral features and combining image features to obtain a polluted region through a spectrum fusion network based on the second spectrogram and the second pollution monitoring image comprises the following steps:
inputting the second pollution monitoring image into an image convolution network, extracting image features and obtaining a pollution monitoring feature map; the size of the pollution monitoring characteristic diagram is the same as that of the spectrum characteristic diagram;
The size of the convolution kernel of the last layer of the image convolution network is the same as that of the spectrum characteristic diagram;
superposing and fusing the spectrum characteristic diagram and the pollution monitoring characteristic diagram to obtain a fused characteristic diagram;
inputting the fusion feature map into a spectrum fusion network, extracting features, detecting boundaries and obtaining a pollution area;
the spectrum fusion network comprises a first convolution network, a second convolution network and a third convolution network;
establishing a relation between the pollution concentration of the pollution area and the color of the second pollution monitoring image, and dividing the pollution area by dividing the neural network to obtain a divided pollution area image, wherein the method comprises the following steps:
obtaining a plurality of detection point information; the detection point information is the concentration of the pollution detected in the pollution area;
clustering the detection point information to obtain a plurality of clustering sets; the values in the cluster set represent values of the same degree of contamination;
connecting values in the plurality of clustering sets at positions corresponding to the second pollution monitoring images for a plurality of times to obtain a rough area image;
inputting the rough area image and the second pollution monitoring image into a segmentation neural network, and predicting concentration distribution conditions to obtain a segmentation pollution area image;
Performing secondary detection by using a local detection neural network according to the first spectrogram, and detecting a local area to obtain a second spectrogram and a second pollution monitoring image, wherein the method comprises the following steps:
inputting the first spectrogram into a local detection neural network, and extracting spectral features to obtain a first spectral feature map;
comparing the values of the elements in the first spectrogram with threshold values respectively, and judging whether the values are in a set pollution range or not to obtain a first comparison value chart; the values in the first ratio graph represent whether there is contamination and a type of contamination in different areas;
performing OR operation on the values in the first spectrum characteristic diagram and the first ratio diagram to obtain a first pollution diagram;
if pollution exists in the first pollution map, dividing the pollution area into a plurality of areas with the same size, and performing spectrum detection to obtain a second spectrum image and a second pollution monitoring image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311039879.6A CN117054353B (en) | 2023-08-17 | 2023-08-17 | Atmospheric pollution source area positioning analysis method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311039879.6A CN117054353B (en) | 2023-08-17 | 2023-08-17 | Atmospheric pollution source area positioning analysis method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117054353A CN117054353A (en) | 2023-11-14 |
| CN117054353B true CN117054353B (en) | 2024-03-19 |
Family
ID=88667293
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311039879.6A Active CN117054353B (en) | 2023-08-17 | 2023-08-17 | Atmospheric pollution source area positioning analysis method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117054353B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2586939C1 (en) * | 2015-04-24 | 2016-06-10 | Федеральное государственное бюджетное научное учреждение "Научно-исследовательский институт аэрокосмического мониторинга "АЭРОКОСМОС" (НИИ "АЭРОКОСМОС") | Method of determining index of state of atmosphere for anthropogenic pollution sources |
| CN107340364A (en) * | 2017-05-31 | 2017-11-10 | 北京市环境保护监测中心 | Polluted space analysis method and device based on magnanimity air pollution concentration data |
| KR20220122381A (en) * | 2021-02-26 | 2022-09-02 | 창원대학교 산학협력단 | Water Quality Monitoring Method and System for Using Unmanned Aerial Vehicle |
| CN115359431A (en) * | 2022-10-21 | 2022-11-18 | 山东省生态环境监测中心 | Atmospheric environment pollution source pollution degree evaluation method and system |
| CN115436573A (en) * | 2022-08-30 | 2022-12-06 | 南京云创大数据科技股份有限公司 | Intelligent monitoring method and device for atmospheric pollution source |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR102082419B1 (en) * | 2019-09-27 | 2020-02-27 | 국방과학연구소 | Apparatus and method for spectrum simulation of air polluted |
-
2023
- 2023-08-17 CN CN202311039879.6A patent/CN117054353B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2586939C1 (en) * | 2015-04-24 | 2016-06-10 | Федеральное государственное бюджетное научное учреждение "Научно-исследовательский институт аэрокосмического мониторинга "АЭРОКОСМОС" (НИИ "АЭРОКОСМОС") | Method of determining index of state of atmosphere for anthropogenic pollution sources |
| CN107340364A (en) * | 2017-05-31 | 2017-11-10 | 北京市环境保护监测中心 | Polluted space analysis method and device based on magnanimity air pollution concentration data |
| KR20220122381A (en) * | 2021-02-26 | 2022-09-02 | 창원대학교 산학협력단 | Water Quality Monitoring Method and System for Using Unmanned Aerial Vehicle |
| CN115436573A (en) * | 2022-08-30 | 2022-12-06 | 南京云创大数据科技股份有限公司 | Intelligent monitoring method and device for atmospheric pollution source |
| CN115359431A (en) * | 2022-10-21 | 2022-11-18 | 山东省生态环境监测中心 | Atmospheric environment pollution source pollution degree evaluation method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117054353A (en) | 2023-11-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111307727B (en) | Water body water color abnormity identification method and device based on time sequence remote sensing image | |
| US9251420B2 (en) | System for mapping and identification of plants using digital image processing and route generation | |
| CN111398176B (en) | Water body water color abnormity remote sensing identification method and device based on pixel scale characteristics | |
| CN107292886B (en) | Target object intrusion detection method and device based on grid division and neural network | |
| US20130279795A1 (en) | Optimization of unknown defect rejection for automatic defect classification | |
| CN111401418A (en) | Employee dressing specification detection method based on improved Faster r-cnn | |
| CN112686274B (en) | Target object detection method and device | |
| CN112365491A (en) | Method for detecting welding seam of container, electronic equipment and storage medium | |
| CN112686595A (en) | Method, device, equipment and storage medium for detecting illegal behavior of logistics operation | |
| CN115359431B (en) | Atmospheric environment pollution source pollution degree evaluation method and system | |
| CN111046950A (en) | Image processing method and device, storage medium and electronic device | |
| CN115578377B (en) | Panel defect detection method, training method, device, equipment and medium | |
| CN110766027A (en) | Image area positioning method and training method of target area positioning model | |
| CN111526099A (en) | Internet of things application flow detection method based on deep learning | |
| CN110910360A (en) | Power grid image positioning method and image positioning model training method | |
| US9177215B2 (en) | Sparse representation for dynamic sensor networks | |
| CN113239227A (en) | Image data structuring method and device, electronic equipment and computer readable medium | |
| CN116823793A (en) | Device defect detection method, device, electronic device and readable storage medium | |
| KC | Enhanced pothole detection system using YOLOX algorithm | |
| CN114581654A (en) | Mutual inductor based state monitoring method and device | |
| CN117054353B (en) | Atmospheric pollution source area positioning analysis method and system | |
| CN112989768A (en) | Method and device for correcting connection questions, electronic equipment and storage medium | |
| CN112858181A (en) | Black and odorous water body monitoring method and device and electronic equipment | |
| CN111931721A (en) | Method and device for detecting color and number of annual inspection label and electronic equipment | |
| CN115631197B (en) | Image processing method, device, medium, equipment and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |