CN115792919A - Method for identifying pollution hot spot area through horizontal scanning and monitoring of aerosol laser radar - Google Patents
Method for identifying pollution hot spot area through horizontal scanning and monitoring of aerosol laser radar Download PDFInfo
- Publication number
- CN115792919A CN115792919A CN202310061211.5A CN202310061211A CN115792919A CN 115792919 A CN115792919 A CN 115792919A CN 202310061211 A CN202310061211 A CN 202310061211A CN 115792919 A CN115792919 A CN 115792919A
- Authority
- CN
- China
- Prior art keywords
- scanning
- hot spot
- pollution
- radar
- model
- 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.)
- Granted
Links
Images
Classifications
-
- 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
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to identification of a pollution hot spot region, in particular to an identification method for aerosol laser radar horizontal scanning monitoring of a pollution hot spot region, which comprises the steps of obtaining radar scanning monitoring data and drawing a scanning map with obvious color contrast; calling a radar scanning map classification model, and classifying the drawn scanning map through the radar scanning map classification model to obtain a scanning map classification result; calling a radar scanning spectrum pollution hot spot area identification model based on the scanning spectrum classification result, and performing pollution hot spot area identification on the drawn scanning spectrum through the radar scanning spectrum pollution hot spot area identification model to obtain a pollution hot spot area identification result; calculating relevant information of the pollution hot spot area according to the identification result of the pollution hot spot area, and comprehensively displaying the scanning map and the pollution hot spot area; the technical scheme provided by the invention can effectively overcome the defect that the prior art cannot accurately identify and distinguish the pollution hot spot areas with different pollution types.
Description
Technical Field
The invention relates to identification of a pollution hot spot area, in particular to an identification method for aerosol laser radar horizontal scanning monitoring of the pollution hot spot area.
Background
At present, various monitoring devices and monitoring means are continuously emerging in the field of atmospheric environment monitoring and pollution tracing. The atmospheric aerosol laser radar has the characteristics of high monitoring data frequency, strong timeliness, high monitoring height, long monitoring distance and the like, can perform vertical monitoring and scanning monitoring in atmospheric environment monitoring, and can visually display a pollution high-value area through a scanning map formed by radar scanning monitoring data so as to determine related information such as a pollution position, a pollution area, a pollution type and the like.
The existing aerosol laser radar equipment has the problem of low intelligent degree to a certain extent, the scanning atlas needs to be analyzed and judged manually, the use effect of the equipment, the early warning real-time performance and other aspects are greatly reduced, and the cost of manpower and material resources consumed by the equipment is high. How to rapidly, effectively and automatically identify and extract the polluted hot spot area in the scanning map, automatically plot the outline of the polluted hot spot area, automatically calculate the polluted area of the polluted hot spot area and the like is a technical problem which is difficult to overcome in the industry for a long time.
However, the problems of poor adaptability and poor effect generally exist in the current method of analyzing the pollution related information by using a signal analysis means of a related mathematical method, and the method can only be used under the condition of ideal radar data with typical characteristics, and is difficult to accurately identify and distinguish pollution hot spot areas with different pollution types under the complex conditions of a plurality of pollution hot spot areas and the like. Because the scanning map is influenced by factors such as relevant field use conditions, external meteorological conditions, internal equipment limitations and the like, the scanning result presents larger difference and uncertainty, and the rules of data and signals are difficult to capture and analyze by adopting a conventional method, so that the methods are difficult to achieve a satisfactory effect in the actual environment monitoring process.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides the method for identifying the pollution hot spot region through horizontal scanning monitoring of the aerosol laser radar, which can effectively overcome the defect that the pollution hot spot regions with different pollution types cannot be accurately identified in the prior art.
In order to achieve the purpose, the invention is realized by the following technical scheme:
an aerosol laser radar horizontal scanning pollution hotspot area identification method comprises the following steps:
s1, radar scanning monitoring data are obtained and a scanning map with obvious color contrast is drawn;
s2, calling a radar scanning map classification model, and classifying the drawn scanning map through the radar scanning map classification model to obtain a scanning map classification result;
s3, calling a radar scanning map pollution hot spot region identification model based on the scanning map classification result, and performing pollution hot spot region identification on the drawn scanning map through the radar scanning map pollution hot spot region identification model to obtain a pollution hot spot region identification result;
and S4, calculating relevant information of the pollution hot spot area according to the identification result of the pollution hot spot area, and comprehensively displaying the scanning map and the pollution hot spot area.
Preferably, the radar scanning monitoring data is acquired in S1, and the scanning map with distinct color contrast is drawn, including:
s11, carrying out horizontal scanning monitoring on the aerosol laser radar, and obtaining atmospheric extinction coefficient profiles of all angles of the horizontal scanning monitoring through an inversion algorithm of the laser radar monitoring;
s12, after horizontal scanning monitoring is finished, calculating atmospheric extinction coefficient data of a scanning result to obtain drawing parameters of an extinction coefficient scanning map;
and S13, drawing a scanning map with obvious color contrast based on the corresponding relation between the atmospheric extinction coefficient value and the color by combining drawing parameters of the extinction coefficient scanning map according to the spatial resolution and the scanning angle information of the atmospheric extinction coefficient profile.
Preferably, after the horizontal scanning monitoring in S12 is finished, calculating the atmospheric extinction coefficient data of the scanning result of this time to obtain a drawing parameter of the extinction coefficient scanning map, including:
s121, forming a set of atmospheric extinction coefficient data in a certain height range in the atmospheric extinction coefficient profiles of all angles through horizontal scanning monitoring, and sequencing;
s122, taking out a set number of atmospheric extinction coefficient data positioned in the middle of the sequence from the set, and calculating the average value of the atmospheric extinction coefficient data to obtain the median value of atmospheric extinction coefficient intensity;
and S123, dividing the atmospheric extinction coefficient intensity median into numerical value gear intervals, and correspondingly giving drawing parameters of the extinction coefficient scanning spectrum of the numerical value gear intervals by combining the system condition and the actual experience of the aerosol laser radar.
Preferably, the step S2 of calling a radar scanning spectrum classification model, and classifying the scanning spectrum obtained by drawing through the radar scanning spectrum classification model to obtain a scanning spectrum classification result, includes:
establishing a radar scanning pattern classification data set, and performing model training on a radar scanning pattern classification model by using the radar scanning pattern classification data set, wherein the method specifically comprises the following steps:
s21, collecting a radar scanning map classification training data set and a radar scanning map classification verification data set, and classifying and labeling the radar scanning map classification training data set;
s22, setting global parameters of model training;
s23, reading training images in the radar scanning atlas classification training data set, converting the training images into floating point arrays, normalizing the training images, and setting labels for the training images;
s24, carrying out image amplification on the training image, and expanding the size of the radar scanning spectrum classification training data set;
s25, calling a radar scanning spectrum classification model, and performing model training by using a radar scanning spectrum classification training data set;
s26, continuously adjusting global parameters of model training to obtain the optimal recognition rate, and storing the optimal radar scanning spectrum classification model into a model file;
s27, loading a model file, verifying by using a radar scanning map classification verification data set, and analyzing a model classification effect;
and S28, exporting the model file for subsequent calling.
Preferably, the radar scanning atlas classification model is a ResNet50 residual error network model based on a Convolutional Neural Network (CNN) algorithm in a deep learning framework Tensflow.
Preferably, in S3, the identification of the pollution hot spot region is performed on the scanning spectrum obtained by drawing through the radar scanning spectrum pollution hot spot region identification model, so as to obtain a pollution hot spot region identification result, and the identification result includes:
establishing a radar scanning spectrum pollution hot spot area identification data set, and performing model training on a radar scanning spectrum pollution hot spot area identification model by using the radar scanning spectrum pollution hot spot area identification data set, wherein the model training specifically comprises the following steps:
s31, collecting a radar scanning spectrum pollution hotspot region identification training data set and a radar scanning spectrum pollution hotspot region identification verification data set;
s32, reading a training image in the radar scanning spectrum pollution hotspot region identification training data set, labeling the pollution hotspot region of the training image, and converting the training image and the verification image into a target data format required by model training;
s33, setting global parameters of model training;
s34, calling a radar scanning map pollution hotspot area recognition model, performing model training by using a radar scanning map pollution hotspot area recognition training data set, and obtaining a final check point file after the training is completed;
s35, loading a check point file, identifying and verifying a data set by using a radar scanning map pollution hot spot region, and analyzing a model identification effect;
and S36, exporting the checkpoint file for subsequent calling.
Preferably, the radar scanning spectrum pollution hot spot region identification model is a DeepLabV3+ model using an image semantic segmentation algorithm in a deep learning framework TensorFlow.
Preferably, in S3, invoking a radar scanning pattern pollution hot spot region identification model based on the scanning pattern classification result includes:
and when the classification result of the scanning map is a normal scanning map, calling a radar scanning map pollution hot spot region identification model.
Preferably, in S4, the calculating of the relevant information of the pollution hot spot region according to the identification result of the pollution hot spot region, and the comprehensive display of the scanning spectrum and the pollution hot spot region include:
s41, analyzing the recognition result of the pollution hot spot region, and reserving the recognition result of the pollution hot spot region with certain confidence coefficient;
s42, determining a polluted hot spot area according to the mask layer two-dimensional array returned by the identification result of the polluted hot spot area;
s43, performing contour polygon calculation based on the two-dimensional mask layer array of each pollution hot spot area to obtain a contour polygon coordinate array of each pollution hot spot area;
s44, calculating the longitude and latitude of the outline polygon according to the outline polygon coordinate array of each pollution hot spot area by combining the longitude and latitude information of the radar scanning central point, and calculating the area and the gravity center position of the pollution hot spot area;
s45, carrying out comprehensive display of a scanning map and the pollution hot spot area in the platform in combination with an electronic map, carrying out contour plotting on each pollution hot spot area, and displaying the pollution type, the pollution area and the pollution gravity center position of the pollution hot spot area in detail.
Preferably, the determining the dirty hot spot region according to the mask layer two-dimensional array returned by the dirty hot spot region identification result in S42 includes:
if the identification result of the polluted hot spot region returns a group of mask layer two-dimensional arrays, the group of mask layer two-dimensional arrays corresponds to a polluted hot spot region, namely the polluted hot spot region comprises pixel positions covered by the polluted hot spot region;
if the identification result of the polluted hot spot region returns a plurality of mask layer two-dimensional arrays, calculating whether the polluted hot spot regions are overlapped, and combining the overlapped polluted hot spot regions to obtain a plurality of non-overlapped polluted hot spot regions.
Compared with the prior art, the aerosol laser radar horizontal scanning pollution hotspot area identification method provided by the invention has the following beneficial effects:
1) The collected scanning maps are classified and image segmentation is realized through machine learning, the pollution hot spot areas can be effectively identified for the scanning maps under different conditions, and the pollution types of the pollution hot spot areas can be accurately distinguished;
2) In the actual environment monitoring process, in most cases, the scanning spectrum does not ideally present typical pollution hot spot characteristics every time, due to meteorological conditions, radar performance, related faults and the like, the scanning spectrum may present situations without practical significance or does not have typical pollution hot spot characteristics, abnormal scanning spectra can be filtered out through a radar scanning spectrum classification model, normal scanning spectra are reserved, and a basis is provided for accurately identifying pollution hot spot areas subsequently;
3) For various scanning maps with different forms and complex conditions, the pollution hot spot features are different in shapes, different in sizes and changeable in directions, and the pollution hot spot region in the scanning map needs to be accurately identified and extracted.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It is obvious that the drawings in the following description are only some embodiments of the invention, and that for a person skilled in the art, other drawings can be derived from them without inventive effort.
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a schematic diagram showing the effect of scanning atlas drawing on the same scanning monitoring result by using drawing parameters of different extinction coefficients scanning atlases in the present invention;
FIG. 3 is a schematic illustration of an abnormal scan pattern of the present invention;
FIG. 4 is a schematic diagram of classification labeling of a radar scanning spectrum classification training data set according to the present invention;
FIG. 5 is a schematic view of the types of pollution in different hot spot areas according to the present invention;
FIG. 6 is a schematic diagram of a recognition training data set for identifying a pollution hot spot region of a radar scanning spectrum according to the present invention;
FIG. 7 is a schematic view of the present invention showing the scanning spectrum and the hot spot area of contamination comprehensively;
FIG. 8 is a diagram illustrating the recognition effect of the present application on hot spot areas with different pollution types;
fig. 9 is a diagram illustrating an identification effect of the present application on a plurality of pollution hot spot regions of different pollution types in the same scanning spectrum;
fig. 10 is a diagram illustrating an identification effect of a plurality of pollution hot spot regions of different pollution types in a non-continuous scanning spectrum according to the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. It is to be understood that the embodiments described are only a few embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
A method for identifying a pollution hot spot region through horizontal scanning and monitoring of an aerosol laser radar is disclosed, as shown in figure 1, (1) radar scanning and monitoring data are obtained and are drawn into a scanning map with obvious color contrast, and the method specifically comprises the following steps:
s11, carrying out horizontal scanning monitoring on the aerosol laser radar, and obtaining atmospheric extinction coefficient profiles of all angles of the horizontal scanning monitoring through an inversion algorithm of the laser radar monitoring;
s12, after horizontal scanning monitoring is finished, calculating atmospheric extinction coefficient data of a scanning result to obtain drawing parameters of an extinction coefficient scanning map;
and S13, drawing a scanning map with obvious color contrast based on the corresponding relation between the atmospheric extinction coefficient value and the color by combining drawing parameters of the extinction coefficient scanning map according to the spatial resolution and the scanning angle information of the atmospheric extinction coefficient profile.
Through the steps, the scanning monitoring result of the aerosol laser radar can be finally drawn into a square picture with 800 × 800 pixels, the data distance representing radar scanning is 6km, the picture format is a PNG format, the scanning map in the picture is a circle with the radius of 400 pixels, and partial angle image deletion can be caused according to the setting of different laser radar scanning angles.
According to the method and the device, the classification and image segmentation processing of the acquired scanning atlas is realized through machine learning, so that the drawing quality of the scanning atlas directly determines the effect of subsequent calculation and analysis. As shown in fig. 2, the drawing parameters of different extinction coefficient scanning maps are used to influence the drawing of the scanning map on the same scanning monitoring result, and it can be seen that the scanning maps obtained by using the drawing parameters of different extinction coefficient scanning maps are very different, and the setting of too small and too large drawing parameters of the extinction coefficient scanning maps may result in that the pollution hotspot information cannot be effectively distinguished, so that how to properly select the drawing parameters of the extinction coefficient scanning maps to draw the scanning maps is a very important link.
After the horizontal scanning monitoring is finished, calculating the atmospheric extinction coefficient data of the scanning result to obtain the drawing parameters of the extinction coefficient scanning atlas, wherein the drawing parameters comprise:
s121, forming a set of atmospheric extinction coefficient data in a certain height range (400m to 800m) in the atmospheric extinction coefficient profiles of all angles through horizontal scanning monitoring, and sequencing;
s122, taking out a set number (100) of atmospheric extinction coefficient data positioned in the middle of the sequence from the set, and calculating the average value of the atmospheric extinction coefficient data to obtain an atmospheric extinction coefficient intensity median;
and S123, dividing the atmospheric extinction coefficient intensity median into numerical value gear intervals, and correspondingly giving drawing parameters of the extinction coefficient scanning spectrum of the numerical value gear intervals by combining the system condition and the actual experience of the aerosol laser radar.
(2) And calling a radar scanning spectrum classification model (calling a pb model file of the radar scanning spectrum classification model by executing a Python program), and classifying the drawn scanning spectrum by the radar scanning spectrum classification model to obtain a scanning spectrum classification result.
The method comprises the following steps of establishing a radar scanning spectrum classification data set, and performing model training on a radar scanning spectrum classification model by using the radar scanning spectrum classification data set, and specifically comprises the following steps:
s21, collecting a radar scanning map classification training data set and a radar scanning map classification verification data set, and classifying and labeling the radar scanning map classification training data set;
s22, setting global parameters of model training;
s23, reading training images in the radar scanning atlas classification training data set, converting the training images into floating point arrays, normalizing the training images, and setting labels for the training images;
s24, carrying out image amplification on the training images, and expanding the size of the radar scanning map classification training data set (in the application, only the rotation type image amplification is carried out, and the scanning maps can generate similar image characteristics at different angles);
s25, calling a radar scanning map classification model, and performing model training by using a radar scanning map classification training data set;
s26, continuously adjusting global parameters of model training to obtain the optimal recognition rate, and storing the optimal radar scanning spectrum classification model into a h5 model file of a keras framework;
s27, loading the h5 model file, verifying by using a radar scanning atlas classification verification data set, and analyzing the classification effect of the model;
and S28, exporting the h5 model file into a pb-format model file for subsequent calling.
In the technical scheme, the radar scanning atlas classification model is a ResNet50 residual error network model based on a Convolutional Neural Network (CNN) algorithm in a deep learning framework Tensflow.
1) Gather radar scanning map classification training data set and radar scanning map classification verification data set to carry out categorised mark to radar scanning map classification training data set, include:
the scanning maps under various conditions need to be collected, the scanning maps which contain various normal, abnormal and defects in the daily use of the laser radar are included as much as possible, and a plurality of scanning maps obtained in different climatic condition areas, different pollution characteristic areas, different altitude latitude and longitude areas and different equipment operation states can be collected;
collecting 1500-3000 scanning maps as a radar scanning map classification training data set according to needs, and classifying and labeling the scanning maps in the training data set (as shown in fig. 4) so as to distinguish a normal scanning map from an abnormal scanning map in subsequent machine learning, wherein as shown in fig. 3, main classification marks comprise a normal scanning map, an equipment fault map (signal saturation), a noise abnormal map (excessive noise) and a weather abnormal map (cloud and fog influence);
and collecting 500-1000 scanning maps as a radar scanning map classification verification data set according to requirements so as to verify the scanning map classification result of the model in subsequent machine learning.
2) Global parameters for model training, including:
the input image size (images with 800 × 800 pixels are used collectively in this application), the image storage path, the number of training cycles, the learning rate, the number of scan pattern categories (4), and the gradient decreasing batch size.
(3) And calling a radar scanning spectrum pollution hot spot region identification model based on the scanning spectrum classification result (calling a pb model file of the radar scanning spectrum pollution hot spot region identification model by executing a Python program), and performing pollution hot spot region identification on the drawn scanning spectrum through the radar scanning spectrum pollution hot spot region identification model to obtain a pollution hot spot region identification result.
A. Calling a radar scanning spectrum pollution hot spot region identification model based on a scanning spectrum classification result, wherein the model comprises the following steps:
and when the classification result of the scanning map is a normal scanning map, calling a radar scanning map pollution hot spot region identification model.
B. Carrying out pollution hot spot area identification on the drawn scanning spectrum through a radar scanning spectrum pollution hot spot area identification model to obtain a pollution hot spot area identification result, wherein the method comprises the following steps:
establishing a radar scanning spectrum pollution hot spot area identification data set, and performing model training on a radar scanning spectrum pollution hot spot area identification model by using the radar scanning spectrum pollution hot spot area identification data set, wherein the model training specifically comprises the following steps:
s31, collecting a radar scanning spectrum pollution hotspot region identification training data set and a radar scanning spectrum pollution hotspot region identification verification data set;
s32, reading a training image in the radar scanning spectrum pollution hot spot area recognition training data set, carrying out pollution hot spot area labeling on the training image, and converting the training image and the verification image into a target data format required by model training;
s33, setting global parameters of model training;
s34, calling a radar scanning map pollution hot spot area recognition model, performing model training by using a radar scanning map pollution hot spot area recognition training data set, and obtaining a final check point (CheckPoint) file after the training is completed;
s35, loading a check point file, identifying and verifying a verification data set by utilizing a radar scanning map pollution hotspot region, and analyzing a model identification effect;
and S36, exporting the checkpoint file into a model file in a pb format for subsequent calling.
In the technical scheme, the radar scanning map pollution hot spot region identification model is a DeepLabV3+ model using an image semantic segmentation algorithm in a deep learning framework TensorFlow.
1) Collecting a radar scanning spectrum pollution hot spot region identification training data set and a radar scanning spectrum pollution hot spot region identification verification data set, wherein the collection comprises the following steps:
various normal scanning spectrums need to be acquired, including normal scanning spectrums of a pollution-free hot spot area, a single pollution hot spot area and a plurality of pollution hot spot areas as far as possible, and including normal scanning spectrums with lower background values and higher background values;
collecting 800-1500 normal scanning maps as a radar scanning map pollution hotspot area identification training data set according to needs, and marking the pollution hotspot area on the scanning maps in the training data set (as shown in fig. 6) so as to identify and extract the pollution hotspot area in the normal scanning maps in the subsequent machine learning, wherein as shown in fig. 5, the main pollution types of the pollution hotspot area comprise point pollution, linear pollution, planar pollution and no pollution;
collecting 300-500 normal scanning maps as a radar scanning map identification verification data set for the pollution hot spot area according to requirements, so as to verify the identification result of the pollution hot spot area of the model in subsequent machine learning.
2) Labeling a pollution hot spot region of a training image, and converting the training image and a verification image into a target data format required by model training, wherein the method comprises the following steps:
firstly, labeling the pollution hot spot region by using a labelme labeling tool one by one training image, and generating a corresponding image labeling information JSON file;
generating a PASCAL VOC data set by converting the training images and the corresponding JSON files through data conversion codes in a labelme marking tool, wherein the images in the segmentationClassPNG folder are marked training images;
because the deep lab model is trained by using a single-channel labeling image (namely, a grayscale image), the color labeling training image in the segmentationclass png folder in the previous step needs to be converted into a grayscale image;
and converting the original training image and the gray-scale image generated in the previous step into a TFrecord standard data format used by a TensorFlow frame, and simultaneously processing the verification images in the identification and verification data set of the radar scanning map pollution hot spot region together with the training image, and uniformly converting the verification images into the TFrecord standard data format.
3) Global parameters for model training, including:
number of training cycles, learning rate, number of pollution types (4), learning rate strategy, etc.
(4) Calculating relevant information of the polluted hot spot area according to the identification result of the polluted hot spot area, and comprehensively displaying the scanning spectrum and the polluted hot spot area (as shown in fig. 7), specifically comprising:
s41, analyzing the recognition result of the pollution hot spot region, and reserving the recognition result of the pollution hot spot region with certain confidence coefficient;
s42, determining a polluted hot spot area according to the mask layer two-dimensional array returned by the identification result of the polluted hot spot area;
s43, performing contour polygon calculation based on the two-dimensional mask layer array of each pollution hot spot area to obtain a contour polygon coordinate array of each pollution hot spot area;
s44, calculating the longitude and latitude of the outline polygon according to the outline polygon coordinate array of each pollution hot spot area by combining the longitude and latitude information of the radar scanning central point, and calculating the area and the gravity center position of the pollution hot spot area;
s45, carrying out comprehensive display of a scanning map and the pollution hot spot area in the platform in combination with an electronic map, carrying out contour plotting on each pollution hot spot area, and displaying the pollution type, the pollution area and the pollution gravity center position of the pollution hot spot area in detail.
The method for determining the polluted hot spot area according to the mask layer two-dimensional array returned by the identification result of the polluted hot spot area comprises the following steps:
if the identification result of the polluted hotspot area returns a group of mask layer two-dimensional arrays, the mask layer two-dimensional arrays correspond to a polluted hotspot area, namely the polluted hotspot area comprises pixel positions covered by the polluted hotspot area;
if the identification result of the polluted hot spot region returns a plurality of mask layer two-dimensional arrays, calculating whether the polluted hot spot regions are overlapped, and combining the overlapped polluted hot spot regions to obtain a plurality of non-overlapped polluted hot spot regions.
Under different pollution conditions, the radar scanning spectrum can present different pollution effects, and has point-shaped pollution, linear pollution and planar pollution, and can be classified according to the shapes of pollution hot spot areas, so that a user is guided to perform corresponding treatment on different types of pollution. For example, spot pollution is generally sporadic or small-sized dust, incineration, and the like, linear pollution is generally overhead discharge pollution, large-sized incineration, and the like, and planar pollution is generally large-area integral and collective pollution. As shown in fig. 8, for the pollution hot spot areas of different pollution types, the technical scheme adopted in the present application can effectively identify and accurately determine the boundary contour information thereof.
Because the scanning radius of the aerosol laser radar is wider, the covered scanning area is larger, and therefore, in some cases, a plurality of pollution hot spot areas with different pollution types may exist in the same radar scanning spectrum at the same time. As shown in fig. 9, the technical solution adopted in the present application can effectively identify a plurality of pollution hot spot areas with different pollution types in the same radar scanning spectrum.
In addition, because the situation that obstacles shield can occur in the working process of the aerosol laser radar, the radar scanning spectrum is not continuous under the situations, and the pollution hot spot area appearing in the discontinuous scanning spectrum is effectively identified, so that certain difficulty is increased. As shown in fig. 10, the technical solution adopted in the present application can effectively identify a plurality of pollution hot spot regions of different pollution types in a non-continuous scanning spectrum.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. An aerosol laser radar horizontal scanning pollution hotspot area identification method is characterized by comprising the following steps: the method comprises the following steps:
s1, radar scanning monitoring data are obtained and a scanning map with obvious color contrast is drawn;
s2, calling a radar scanning map classification model, and classifying the drawn scanning map through the radar scanning map classification model to obtain a scanning map classification result;
s3, calling a radar scanning map pollution hot spot region identification model based on the scanning map classification result, and performing pollution hot spot region identification on the drawn scanning map through the radar scanning map pollution hot spot region identification model to obtain a pollution hot spot region identification result;
and S4, calculating relevant information of the pollution hot spot area according to the identification result of the pollution hot spot area, and comprehensively displaying the scanning map and the pollution hot spot area.
2. The aerosol lidar horizontal scanning pollution hot spot area identification method according to claim 1, wherein: the method comprises the following steps of S1, acquiring radar scanning monitoring data, and drawing a scanning map with obvious color contrast, wherein the scanning map comprises the following steps:
s11, carrying out horizontal scanning monitoring on the aerosol laser radar, and obtaining atmospheric extinction coefficient profiles of all angles of the horizontal scanning monitoring through an inversion algorithm of the laser radar monitoring;
s12, after the horizontal scanning monitoring is finished, calculating the atmospheric extinction coefficient data of the scanning result to obtain drawing parameters of an extinction coefficient scanning map;
and S13, drawing a scanning map with obvious color contrast based on the corresponding relation between the atmospheric extinction coefficient value and the color by combining drawing parameters of the extinction coefficient scanning map according to the spatial resolution and the scanning angle information of the atmospheric extinction coefficient profile.
3. The aerosol lidar horizontal scanning pollution monitoring hotspot area identification method according to claim 2, wherein: after the horizontal scanning monitoring in the step S12 is finished, calculating the atmospheric extinction coefficient data of the scanning result of this time to obtain drawing parameters of the extinction coefficient scanning map, including:
s121, forming a set of atmospheric extinction coefficient data in a certain height range in the atmospheric extinction coefficient profiles of all angles through horizontal scanning monitoring, and sequencing;
s122, taking out a set number of atmospheric extinction coefficient data positioned in the middle of the sequence from the set, and calculating the average value of the atmospheric extinction coefficient data to obtain the median value of atmospheric extinction coefficient intensity;
and S123, dividing the atmospheric extinction coefficient intensity median into numerical value gear intervals, and correspondingly giving drawing parameters of the extinction coefficient scanning spectrum of the numerical value gear intervals by combining the system condition and the actual experience of the aerosol laser radar.
4. The aerosol lidar horizontal scanning pollution monitoring hotspot area identification method of claim 1, wherein: and S2, calling a radar scanning spectrum classification model, classifying the scanning spectrum obtained by drawing through the radar scanning spectrum classification model, and obtaining a scanning spectrum classification result, wherein the method comprises the following steps:
establishing a radar scanning pattern classification data set, and performing model training on a radar scanning pattern classification model by using the radar scanning pattern classification data set, wherein the method specifically comprises the following steps:
s21, collecting a radar scanning map classification training data set and a radar scanning map classification verification data set, and classifying and labeling the radar scanning map classification training data set;
s22, setting global parameters of model training;
s23, reading training images in the radar scanning atlas classification training data set, converting the training images into floating point arrays, normalizing the training images, and setting labels of the training images;
s24, carrying out image amplification on the training image, and expanding the size of a radar scanning spectrum classification training data set;
s25, calling a radar scanning spectrum classification model, and performing model training by using a radar scanning spectrum classification training data set;
s26, continuously adjusting global parameters of model training to obtain the optimal recognition rate, and storing the optimal radar scanning spectrum classification model into a model file;
s27, loading a model file, verifying by using a radar scanning map classification verification data set, and analyzing a model classification effect;
and S28, exporting the model file for subsequent calling.
5. The aerosol lidar horizontal scanning pollution hot spot area identification method according to claim 4, wherein the method comprises the following steps: the radar scanning atlas classification model is a ResNet50 residual error network model based on a Convolutional Neural Network (CNN) algorithm in a deep learning framework TensorFlow.
6. The aerosol lidar horizontal scanning pollution hot spot area identification method according to claim 1, wherein: s3, carrying out pollution hot spot area identification on the drawn scanning spectrum through the radar scanning spectrum pollution hot spot area identification model to obtain a pollution hot spot area identification result, wherein the method comprises the following steps:
establishing a radar scanning spectrum pollution hot spot area identification data set, and performing model training on a radar scanning spectrum pollution hot spot area identification model by using the radar scanning spectrum pollution hot spot area identification data set, wherein the model training specifically comprises the following steps:
s31, collecting a radar scanning spectrum pollution hotspot region identification training data set and a radar scanning spectrum pollution hotspot region identification verification data set;
s32, reading a training image in the radar scanning spectrum pollution hot spot area recognition training data set, carrying out pollution hot spot area labeling on the training image, and converting the training image and the verification image into a target data format required by model training;
s33, setting global parameters of model training;
s34, calling a radar scanning spectrum pollution hot spot area recognition model, performing model training by using a radar scanning spectrum pollution hot spot area recognition training data set, and obtaining a final check point file after the training is completed;
s35, loading a check point file, identifying and verifying a verification data set by utilizing a radar scanning map pollution hotspot region, and analyzing a model identification effect;
and S36, exporting the checkpoint file for subsequent calling.
7. The aerosol lidar horizontal scanning pollution monitoring hotspot area identification method of claim 6, wherein: the radar scanning map pollution hotspot area identification model is a DeepLabV3+ model using an image semantic segmentation algorithm in a deep learning framework Tensflow.
8. The aerosol lidar horizontal scanning pollution hot spot area identification method according to claim 6, wherein: and S3, calling a radar scanning spectrum pollution hotspot region identification model based on the scanning spectrum classification result, wherein the model comprises the following steps:
and when the classification result of the scanning atlas is a normal scanning atlas, calling a radar scanning atlas pollution hot spot region identification model.
9. The aerosol lidar horizontal scanning pollution monitoring hotspot area identification method of claim 1, wherein: s4, calculating relevant information of the pollution hot spot area according to the identification result of the pollution hot spot area, and comprehensively displaying the scanning map and the pollution hot spot area, wherein the method comprises the following steps:
s41, analyzing the recognition result of the pollution hot spot region, and reserving the recognition result of the pollution hot spot region with certain confidence coefficient;
s42, determining a polluted hot spot area according to the mask layer two-dimensional array returned by the identification result of the polluted hot spot area;
s43, performing contour polygon calculation based on the two-dimensional mask layer array of each pollution hot spot area to obtain a contour polygon coordinate array of each pollution hot spot area;
s44, calculating the longitude and latitude of the outline polygon according to the outline polygon coordinate array of each pollution hot spot area by combining the longitude and latitude information of the radar scanning central point, and calculating the area and the gravity center position of the pollution hot spot area;
s45, carrying out comprehensive display of a scanning map and the pollution hot spot area in the platform in combination with an electronic map, carrying out contour plotting on each pollution hot spot area, and displaying the pollution type, the pollution area and the pollution gravity center position of the pollution hot spot area in detail.
10. The aerosol lidar horizontal scanning pollution monitoring hotspot area identification method of claim 9, wherein: determining the polluted hot spot area according to the mask layer two-dimensional array returned by the identification result of the polluted hot spot area in the step S42, wherein the step comprises the following steps:
if the identification result of the polluted hotspot area returns a group of mask layer two-dimensional arrays, the mask layer two-dimensional arrays correspond to a polluted hotspot area, namely the polluted hotspot area comprises pixel positions covered by the polluted hotspot area;
if the identification result of the polluted hot spot region returns a plurality of mask layer two-dimensional arrays, calculating whether the polluted hot spot regions are overlapped, and combining the overlapped polluted hot spot regions to obtain a plurality of non-overlapped polluted hot spot regions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310061211.5A CN115792919B (en) | 2023-01-19 | 2023-01-19 | Method for identifying polluted hot spot area through horizontal scanning monitoring of aerosol laser radar |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310061211.5A CN115792919B (en) | 2023-01-19 | 2023-01-19 | Method for identifying polluted hot spot area through horizontal scanning monitoring of aerosol laser radar |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115792919A true CN115792919A (en) | 2023-03-14 |
| CN115792919B CN115792919B (en) | 2023-05-16 |
Family
ID=85429782
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310061211.5A Active CN115792919B (en) | 2023-01-19 | 2023-01-19 | Method for identifying polluted hot spot area through horizontal scanning monitoring of aerosol laser radar |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115792919B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117147778A (en) * | 2023-10-31 | 2023-12-01 | 航天宏图信息技术股份有限公司 | Method and device for tracing and monitoring atmospheric pollutants, electronic equipment and storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110245642A (en) * | 2019-06-21 | 2019-09-17 | 北京市市政工程研究院 | A kind of radar spectrum recognition method and system based on deep learning |
| CN110261271A (en) * | 2019-07-03 | 2019-09-20 | 安徽科创中光科技有限公司 | A kind of horizontal pollution based on laser radar big data is traced to the source artificial intelligence identifying system |
| CN110441777A (en) * | 2019-07-11 | 2019-11-12 | 中山大学 | A kind of inversion method of the aerosol Vertical Profile based on laser radar |
| CN111985567A (en) * | 2020-08-21 | 2020-11-24 | 河北先河环保科技股份有限公司 | Automatic pollution source type identification method based on machine learning |
| CN112485319A (en) * | 2020-10-19 | 2021-03-12 | 上海畅合智能技术有限公司 | Atmospheric environment navigation monitoring method, information processing method and device and monitoring vehicle |
| WO2021184630A1 (en) * | 2020-03-19 | 2021-09-23 | 平安国际智慧城市科技股份有限公司 | Method for locating pollutant discharge object on basis of knowledge graph, and related device |
| CN114969544A (en) * | 2022-06-16 | 2022-08-30 | 中国平安人寿保险股份有限公司 | Hot data-based recommended content generation method, device, equipment and medium |
| CN115032653A (en) * | 2022-05-13 | 2022-09-09 | 中国科学院合肥物质科学研究院 | Method for detecting atmospheric characteristics of geometric factor region by using scanning laser radar |
| CN115356748A (en) * | 2022-09-29 | 2022-11-18 | 江西财经大学 | Method and system for extracting atmospheric pollution information based on laser radar observation result |
-
2023
- 2023-01-19 CN CN202310061211.5A patent/CN115792919B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110245642A (en) * | 2019-06-21 | 2019-09-17 | 北京市市政工程研究院 | A kind of radar spectrum recognition method and system based on deep learning |
| CN110261271A (en) * | 2019-07-03 | 2019-09-20 | 安徽科创中光科技有限公司 | A kind of horizontal pollution based on laser radar big data is traced to the source artificial intelligence identifying system |
| CN110441777A (en) * | 2019-07-11 | 2019-11-12 | 中山大学 | A kind of inversion method of the aerosol Vertical Profile based on laser radar |
| WO2021184630A1 (en) * | 2020-03-19 | 2021-09-23 | 平安国际智慧城市科技股份有限公司 | Method for locating pollutant discharge object on basis of knowledge graph, and related device |
| CN111985567A (en) * | 2020-08-21 | 2020-11-24 | 河北先河环保科技股份有限公司 | Automatic pollution source type identification method based on machine learning |
| CN112485319A (en) * | 2020-10-19 | 2021-03-12 | 上海畅合智能技术有限公司 | Atmospheric environment navigation monitoring method, information processing method and device and monitoring vehicle |
| CN115032653A (en) * | 2022-05-13 | 2022-09-09 | 中国科学院合肥物质科学研究院 | Method for detecting atmospheric characteristics of geometric factor region by using scanning laser radar |
| CN114969544A (en) * | 2022-06-16 | 2022-08-30 | 中国平安人寿保险股份有限公司 | Hot data-based recommended content generation method, device, equipment and medium |
| CN115356748A (en) * | 2022-09-29 | 2022-11-18 | 江西财经大学 | Method and system for extracting atmospheric pollution information based on laser radar observation result |
Non-Patent Citations (2)
| Title |
|---|
| CHANG SHU 等: "AN AEROSOL LIDAR DATA VISUALIZATION METHOD BASED ON ARCGIS", 《THE INTERNATIONAL ARCHIVES OF THE PHOTOGRAMMETRY, REMOTE SENSING AND SPATIAL INFORMATION SCIENCES》 * |
| 李明 等: "基于输出信杂噪比的机载雷达训练样本选择算法", 《电子科技大学学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117147778A (en) * | 2023-10-31 | 2023-12-01 | 航天宏图信息技术股份有限公司 | Method and device for tracing and monitoring atmospheric pollutants, electronic equipment and storage medium |
| CN117147778B (en) * | 2023-10-31 | 2024-01-05 | 航天宏图信息技术股份有限公司 | Method and device for tracing and monitoring atmospheric pollutants, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115792919B (en) | 2023-05-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110263717B (en) | Method for determining land utilization category of street view image | |
| WO2020093630A1 (en) | Antenna downward inclination angle measurement method based on multi-scale deep semantic segmentation network | |
| CN108648169B (en) | Method and device for automatically identifying defects of high-voltage power transmission tower insulator | |
| CN112149543B (en) | Building dust recognition system and method based on computer vision | |
| CN105373135A (en) | Method and system for guiding airplane docking and identifying airplane type based on machine vision | |
| CN108355987A (en) | A kind of screen printing of battery quality determining method based on piecemeal template matches | |
| CN104008550A (en) | Wafer surface defect feature analysis method and system and wafer surface detect feature classification method and system | |
| KR102346676B1 (en) | Method for creating damage figure using the deep learning-based damage image classification of facility | |
| CN102974551A (en) | Machine vision-based method for detecting and sorting polycrystalline silicon solar energy | |
| CN114241326B (en) | Progressive intelligent production method and system for ground feature elements of remote sensing images | |
| CN113240623B (en) | Pavement disease detection method and device | |
| CN107527037A (en) | Blue-green algae identification and analysis system based on unmanned aerial vehicle remote sensing data | |
| CN114331986A (en) | Dam crack identification and measurement method based on unmanned aerial vehicle vision | |
| CN113378754B (en) | Bare soil monitoring method for construction site | |
| CN112750113B (en) | Glass bottle defect detection method and device based on deep learning and linear detection | |
| CN115792919B (en) | Method for identifying polluted hot spot area through horizontal scanning monitoring of aerosol laser radar | |
| CN116824347A (en) | Road crack detection method based on deep learning | |
| CN113592839A (en) | Distribution network line typical defect diagnosis method and system based on improved fast RCNN | |
| CN113554667A (en) | Three-dimensional displacement detection method and device based on image recognition | |
| CN115115954A (en) | Intelligent identification method for pine nematode disease area color-changing standing trees based on unmanned aerial vehicle remote sensing | |
| CN115239002A (en) | New energy distributed photovoltaic power station monitoring and early warning system and method | |
| CN117953048A (en) | Swivel bridge attitude monitoring system and method based on computer vision | |
| CN117422677A (en) | Method, device and system for detecting image defects of power line for airborne terminal | |
| CN115372281A (en) | Monitoring system and method for physical structure and chemical composition of soil | |
| CN112580504B (en) | Tree species classification counting method and device based on high-resolution satellite remote sensing image |
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 | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A Method for Identifying Pollution Hotspot Areas in Aerosol Lidar Horizontal Scanning Monitoring Effective date of registration: 20230921 Granted publication date: 20230516 Pledgee: Anhui pilot Free Trade Zone Hefei area sub branch of Huishang Bank Co.,Ltd. Pledgor: Hefei Zhongke Guangbo Quantum Technology Co.,Ltd. Registration number: Y2023980057852 |