A drone intelligent inspection method for particulate matter pollution prevention and control

Abstract

The application discloses an unmanned aerial vehicle intelligent inspection method for particulate matter pollution prevention and control, which comprises the following steps: S1, aerial photography is performed to acquire particulate matter pollution source images, and a pollution source image identification model is generated; S2, aerial photography is performed to acquire particulate matter pollution control facility images, and a pollution control facility image identification model is generated; S3, relevant data is collected, and a particulate matter pollution source and particulate matter pollution control facility list and distribution map are generated; S4, a terminal intelligent control platform plans an unmanned aerial vehicle inspection route; S5, when particulate matter concentration data of a monitoring site reaches a warning value, an inspection operation is started; S6, data recorded by the unmanned aerial vehicle inspection is transmitted to the terminal in real time; S7, according to the transmitted data, the terminal identifies pollution source types, pollution degrees and pollution control facility states, and forms a pollution source inspection list; and S8, the generated list is automatically sent to relevant personnel. The application has high target recognition precision and accurate positioning, improves the inspection efficiency of particulate matter pollution prevention and control, and reduces labor costs.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of particulate pollution prevention and control, and particularly relates to an unmanned aerial vehicle intelligent inspection method for particulate pollution prevention and control. BACKGROUND

[0002] With the development of social economy, the emission of atmospheric pollutants of industrial enterprises is increasing, which not only has an adverse effect on environmental quality, but also has an adverse effect on human health. The society and the government have an urgent need for improvement of atmospheric pollution prevention and control technology and equipment. At present, the patrol of atmospheric pollution sources mainly depends on on-site inspection or vehicle sailing inspection. Since the pollution sources are scattered and wide, manual inspection and vehicle sailing inspection are time-consuming and labor-consuming, the inspection area is limited, the intelligent degree is low, the inspection cost is high, and the timeliness of pollution response and disposal is poor. SUMMARY

[0003] (I) Technical problems solved

[0004] The purpose of the present application is to solve the problems of high consumption of manpower and material resources and low work efficiency in traditional atmospheric pollution on-site inspection work, and to save the consumption of manpower in particulate pollution prevention and control work by using the unmanned aerial vehicle intelligent inspection method, so as to greatly improve the work efficiency and intelligent level of the inspection work.

[0005] (II) Technical solutions

[0006] The application provides an unmanned aerial vehicle intelligent inspection method for particulate matter pollution prevention and control, which comprises the following steps: S1: particulate matter pollution source images are obtained by aerial photography of an unmanned aerial vehicle, the particulate matter pollution sources include construction site dust, road dust, bare soil and open-air burning, the pollution source images are subjected to denoising, enhancement, contrast adjustment and brightness adjustment, the image features of the construction site dust, road dust, bare soil and open-air burning are extracted based on an edge detection algorithm, an image feature database of the construction site dust, road dust, bare soil and open-air burning is established, and a pollution source image recognition model is generated based on a machine learning algorithm; S2: particulate matter pollution control facility images are obtained by aerial photography of the unmanned aerial vehicle, the particulate matter pollution control facilities include fence spraying, fog cannons and dust screens, the pollution control facility images are subjected to denoising, enhancement, contrast adjustment and brightness adjustment, the image features of the off and on states of the fence spraying and fog cannons and the image features of the covering state of the dust screen are extracted based on the edge detection algorithm, an image feature database of the off and on states of the fence spraying and fog cannons and the covering state of the dust screen is established, and a pollution control facility implementation state image recognition model is generated based on the machine learning algorithm; S3: particulate matter pollution source data and pollution source control facility data in a control area are collected, the collected data is sent to a data analysis system through a data communication system, the data analysis system corresponds to pollution source management departments and management person information, a particulate matter pollution source and pollution control facility list is exported, a particulate matter pollution source and particulate matter pollution control facility position distribution map is drawn, and the distribution is directly displayed through a three-dimensional space map; S4: based on the particulate matter pollution source distribution map, terrain elevation and meteorological environment parameters in the control area, an air quality model is measured and calculated by using selected air quality models, particulate matter pollution diffusion paths are calculated, particulate matter concentration data of each monitoring point in the area are collected to verify and optimize the air quality model, and an unmanned aerial vehicle inspection route is planned according to a three-dimensional map and the optimized pollution diffusion paths; S5: when particulate matter concentration data of a monitoring site monitored by the terminal intelligent control platform reaches a set early warning value, high value broadcasting is carried out, and an inspection instruction is sent to an unmanned aerial vehicle automatic hangar, the unmanned aerial vehicle automatic hangar receives the instruction, calls the unmanned aerial vehicle inspection route from the terminal intelligent control platform automatically, and starts the unmanned aerial vehicle inspection operation; S6: during the unmanned aerial vehicle inspection flight, particulate matter concentration data, meteorological environment parameters and terrain elevation data of the passing direction are monitored and recorded in real time, particulate matter concentration exceeding point positions are automatically photographed in all directions and from multiple angles, and the related data is transmitted to the terminal intelligent control platform in real time; S7: the terminal intelligent control platform identifies and judges the pollution source types, pollution degrees and operation states of the pollution control facilities by using the pollution source image recognition model and the particulate matter pollution control facility operation state image recognition model, marks the problem positions, and forms an inspection report; and S8: the terminal intelligent control platform automatically sends the generated inspection report to relevant personnel for processing.

[0007] Preferably, the image recognition model is improved in speed and accuracy by adjusting the network structure, optimizing the training parameters, enhancing the image data, introducing the attention module and extracting the image edge features based on the Yolov7 framework, and the image recognition training process specifically includes the following steps: S101: scaling and padding the collected images, and adjusting the size to 640*640; S102: standardizing the images, and normalizing the pixel value of the images to (0, 1); S103: denoising, enhancing, adjusting the contrast and brightness of the images, and randomly cropping, rotating and scaling the images to increase the number of training images; S104: extracting image features by using a convolutional neural network, combining the convolutional layer and the pooling layer, dynamically adjusting the number of convolutional layers, and finely extracting the shape, color, texture and edge features of the images; S105: inputting the image data into the Yolov7 model, starting the batch size of model training from 32, decreasing by 4 times in turn, starting the training round number from 220, decreasing by 20 times in turn, reducing the learning rate from 0.01 to 0.0001, dynamically evaluating the error of the image recognition result and the true label in the training process, and gradually establishing an optimized image recognition model.

[0008] Preferably, the particulate matter pollution source and pollution control facility list information includes the pollution source type, geographic location, emission mode, pollution control facility, management department and management person in charge.

[0009] Preferably, the terminal intelligent control platform displays the flight trajectory of the unmanned aerial vehicle in real time.

[0010] Preferably, in the step S6, the unmanned aerial vehicle performs omnidirectional and multi-angle high-definition shooting on the particulate matter concentration exceeding point, and transmits the related data to the terminal intelligent control platform in real time, and then continues to identify and lock a new target, so as to realize continuous inspection of multiple targets.

[0011] Preferably, the meteorological environment parameters include temperature, humidity, wind direction and wind speed.

[0012] Preferably, the terminal intelligent control platform compares the inspection results of the same pollution source in different time periods, forms a "look back" report, automatically generates a pollution source treatment "closed loop rate", and sends it to relevant personnel.

[0013] (Three) beneficial effects

[0014] The present application has the following beneficial effects:

[0015] (1) Based on the image automatic recognition technology, the pollution sources and control facilities in the inspection area are automatically identified.

[0016] (2) The UAV patrol route is targeted, improves the patrol efficiency and intelligent degree, and reduces the manpower and material cost in the patrol process.

[0017] (3) The terminal intelligent control platform can automatically generate a patrol report and send it to relevant personnel, thereby improving the response speed and management and control efficiency for pollution events. BRIEF DESCRIPTION OF DRAWINGS

[0018] Fig. 1 The flowchart of the embodiment of the present application.

[0019] Fig. 2 The patrol route map planned by the embodiment of the present application DETAILED DESCRIPTION

[0020] Embodiments of the present application will be described in more detail by referring to the accompanying drawings. Although certain embodiments of the present application are shown in the drawings, it is understood that the present application can be implemented in various forms and should not be interpreted as being limited to the embodiments set forth herein, but rather these embodiments are provided to make the present application more thorough and complete. It is understood that the drawings and embodiments of the present application are for exemplary purposes only and are not intended to limit the scope of the present application. It is understood that each step described in the method embodiment of the present application can be executed in different order and / or in parallel. In addition, the method embodiment can include additional steps and / or omit the execution of the steps shown. The scope of the present application is not limited in this respect.

[0021] As shown in Figs. 1-2 The present application proposes an unmanned aerial vehicle intelligent patrol method for particulate matter pollution prevention and control, comprising the following steps:

[0022] S1: Obtain particulate matter pollution source images by aerial photography of an unmanned aerial vehicle. The particulate matter pollution sources include construction dust, road dust, bare soil and open burning. The pollution source images are processed by denoising, enhancement, contrast adjustment and brightness adjustment. Image features of the construction dust, road dust, bare soil and open burning are extracted based on an edge detection algorithm. An image feature database of the construction dust, road dust, bare soil and open burning is established. A pollution source image recognition model is generated based on a machine learning algorithm.

[0023] S2: Obtain images of particulate matter pollution control facilities by aerial photography of unmanned aerial vehicles, the particulate matter pollution control facilities including fence spraying, fog cannons, and dust screens, and perform denoising, enhancement, contrast adjustment, and brightness adjustment on the images of the particulate matter pollution control facilities, extract image features of the off and on states of the fence spraying and fog cannons and the coverage state of the dust screens based on an edge detection algorithm, establish an image feature database of the off and on states of the fence spraying and fog cannons and the coverage state of the dust screens, and generate a pollution control facility implementation state image recognition model based on a machine learning algorithm;

[0024] S3: Collect particulate matter pollution source data and pollution source control facility data in the control area, send the collected data to a data analysis system through a data communication system, correspond to pollution source management departments and management person information, export a particulate matter pollution source and pollution control facility list including pollution source types, geographic locations, emission methods, pollution control facilities, management departments, and management person information, draw a particulate matter pollution source and particulate matter pollution control facility location distribution map, and visually display the distribution through a three-dimensional space map;

[0025] S4: The terminal intelligent control platform measures the particulate matter pollution diffusion path using a selected air quality model based on the particulate matter pollution source distribution map, terrain elevation, and meteorological environmental parameters in the control area, collects particulate matter concentration data of each monitoring point in the region to verify and optimize the air quality model, and plans a UAV inspection route according to the three-dimensional map and the optimized pollution diffusion path;

[0026] S5: When the terminal intelligent control platform detects that the particulate matter concentration data of the monitoring site reaches the set warning value, it performs high-value broadcasting and issues an inspection instruction to the UAV automatic hangar, the UAV automatic hangar automatically calls the UAV inspection route from the terminal intelligent control platform after receiving the instruction, and starts the UAV inspection operation;

[0027] S6: During the UAV inspection flight, real-time monitoring and recording of particulate matter concentration data, meteorological environmental parameters, and terrain elevation data in the passing direction are performed, full-range and multi-angle high-definition shooting of particulate matter concentration exceeding points is automatically performed, relevant data is transmitted to the terminal intelligent control platform in real time, and the inspection operation is continued, new targets are identified and locked, and continuous multi-target inspection is realized.

[0028] S7: The terminal intelligent control platform identifies and judges the pollution source types, pollution levels, and operation states of pollution control facilities according to the data transmitted by the UAV, using the pollution source image recognition model and the particulate matter pollution control facility operation state image recognition model, marks problem points, displays the flight trajectory of the UAV in real time, and forms an inspection report;

[0029] S8: The terminal intelligent control platform automatically sends the generated inspection report to relevant personnel for processing;

[0030] For example, when the particulate matter concentration monitored by a certain small micro station in the region exceeds 80% of the specified limit value of the station, the platform provides a high value reminder, calls up the list of particulate matter pollution sources in the region, and according to the environmental parameters such as air temperature, wind direction, wind speed, humidity, etc. at that time and place, uses the air quality model to simulate the diffusion trajectory of the pollutants in the region, preliminarily determines the area of the pollution source, plans the inspection route, and then issues an inspection instruction to the unmanned aerial vehicle. The unmanned aerial vehicle monitors, photographs and transmits the data of the particulate matter concentration exceeding the standard point in real time according to the route planned by the platform. The platform automatically identifies the image data, marks the pollution source type, geographical location, emission method, pollution control facilities, management department and management person information, generates an inspection report, and sends it to relevant personnel for processing. After the relevant personnel process the pollution source, the unmanned aerial vehicle on the platform starts to inspect the treatment effect of the pollution source. The treatment effect and the implementation of the relevant prevention and control measures are judged by using the data transmitted by the unmanned aerial vehicle, and the "look back" conclusion of the closed-loop or non-closed-loop treatment is automatically generated.

[0031] The content of the unmanned aerial vehicle inspection report is shown in Table 1.

[0032] Table 1 Unmanned aerial vehicle inspection report

[0033]

[0034] When the particulate matter concentration data monitored by each station in the region is within the limit value, the platform simulates the diffusion trajectory of the pollutants according to the existing list of particulate matter pollution sources in the region combined with meteorological environmental parameters, plans the preventive inspection route of the unmanned aerial vehicle, starts the routine inspection operation of the unmanned aerial vehicle, and realizes the early warning and control of pollution.

[0035] The terminal intelligent control platform compares the inspection results of the same pollution source at different time periods, further integrates the pollution source inspection problem discovery frequency and closed-loop rectification treatment into the dynamic control of the pollution source, automatically generates the pollution source treatment "closed-loop rate", and then optimizes the inspection route, realizes the integration, visualization and intelligentization of environmental quality monitoring and pollution source inspection and prevention, reduces manual intervention and time cost, and improves the fine management and intelligent decision-making level of atmospheric environmental protection.

[0036] In the statistical period, the pollution source treatment closed-loop rate = "look back" pollution source discovery frequency / inspection pollution source discovery frequency x 100%, the pollution frequency proportion of a certain place = inspection pollution source discovery frequency in a certain place / total inspection pollution source discovery frequency in the region x 100%, and the pollution source inspection and treatment statistical report in the region is shown in Table 2.

[0037] Table 2 Pollution source inspection and treatment statistical report in the region

[0038]

[0039]

[0040] The scheme is based on image automatic recognition technology, and automatically recognizes the pollution source and control facility situation in the inspection area. The unmanned aerial vehicle patrol route is targeted, which improves the patrol efficiency and intelligent degree, and reduces the human and material cost in the patrol process. The terminal intelligent control platform can automatically generate an inspection report and send it to the relevant personnel, which improves the response speed and control efficiency of the pollution event.

Claims

1. A drone intelligent inspection method for particulate matter pollution prevention and control, characterized in that: The following steps are involved: S1: Using drone aerial photography to obtain images of particulate matter pollution sources, including construction site dust, road dust, bare soil, and open-air burning, the pollution source images are subjected to denoising, enhancement, contrast adjustment, and brightness adjustment. Image features of construction site dust, road dust, bare soil, and open-air burning are extracted using an edge detection algorithm. An image feature database for these sources is established, and a pollution source image recognition model is generated using a machine learning algorithm. S2: Obtain images of particulate pollution control facilities through drone aerial photography. The particulate pollution control facilities include enclosure spraying, fog cannons, and dust screens. De-noise, enhance, adjust contrast, and adjust brightness of the images of the particulate pollution control facilities. Use an edge detection algorithm to extract image features of the enclosure spraying and fog cannons in the off and on states, as well as the image features of the dust screen coverage state. Establish an image feature database of the off and on states of enclosure spraying and fog cannons, as well as the dust screen coverage state. Generate an image recognition model for the implementation status of pollution control facilities based on a machine learning algorithm. S3: Collect data on particulate matter pollution sources and pollution control facilities within the control area, send the collected data to the data analysis system through the data communication system, and correspond it with the information of the pollution source management department and the person in charge to generate a list of particulate matter pollution sources and pollution control facilities, draw a location distribution map of particulate matter pollution sources and particulate matter pollution control facilities, and visually display the distribution through a three-dimensional spatial map; S4: The terminal intelligent control platform uses a selected air quality model to calculate the particle pollution diffusion path based on the particle pollution source distribution map, terrain elevation, and meteorological environmental parameters within the control area. It also collects particle concentration data from each monitoring point in the area to calibrate and optimize the air quality model. The platform then plans drone inspection routes based on the three-dimensional map and the optimized pollution diffusion path. S5: When the terminal intelligent control platform detects that the particulate matter concentration data at the monitoring site reaches the set warning value, it will broadcast a high value and issue an inspection instruction to the drone automatic hangar. After receiving the instruction, the drone automatic hangar will automatically call the drone inspection route from the terminal intelligent control platform and start the drone inspection operation; S6: During the drone inspection flight, it monitors and records the particle concentration data, meteorological environmental parameters and terrain elevation data in the direction of the route in real time. It automatically takes all-round, multi-angle high-definition photos of points where the particle concentration exceeds the standard, and transmits the relevant data in real time to the terminal intelligent control platform; S7: Based on the data transmitted by the drone, the terminal intelligent control platform uses pollution source image recognition models and particulate pollution control facility operation status image recognition models to identify the type of pollution source, pollution level, and the operation status of pollution control facilities, mark problem points, and generate an inspection report. S8: The terminal intelligent control platform automatically sends the generated inspection report to relevant personnel for processing.

2. According to the method for intelligent inspection by unmanned aerial vehicles for particulate matter pollution control in claim 1, the image recognition model improves the speed and accuracy of image recognition by adjusting the network structure, optimizing training parameters, performing image data enhancement processing, introducing an attention module, and extracting image edge features based on the Yolov7 framework. The image recognition training process specifically includes the following steps: S101: scaling and padding the collected images to adjust the size to 640×640; S102: performing normalization processing on the image, normalizing the pixel values ​​of the image to (0, 1); S103: Denoise, enhance, adjust contrast, and adjust brightness of the image, and randomly crop, rotate, and scale the image to increase the number of training images; S104: Use convolutional neural networks to extract image features, combine convolutional layers and pooling layers, dynamically adjust the number of convolutional layers, and refine the extraction of image shape, color, texture, and edge features; S105: Input the image data into the Yolov7 model. The batch size of the model training starts from 32 and decreases in multiples of 4. The number of training rounds starts from 220 and decreases in multiples of 20. The learning rate is reduced from 0.01 to 0.0001. During the training process, the error between the image recognition results and the true labels is dynamically evaluated to gradually establish an optimized image recognition model.

3. According to the method for intelligent inspection by drones for particulate matter pollution prevention and control according to claim 2, in step S3, the list of particulate matter pollution sources and pollution control facilities includes the type of pollution source, geographical location, emission method, pollution control facility, management department, and management responsible person.

4. According to the method for intelligent drone inspection for particulate matter pollution prevention and control described in claim 3, the terminal intelligent control platform displays the flight trajectory of the drone in real time.

5. According to the method for intelligent drone inspection for particulate matter pollution prevention and control according to claim 4, in step S6, the drone performs all-round, multi-angle high-definition photography of points where particulate matter concentration exceeds the standard, transmits the relevant data to the terminal intelligent control platform in real time, and then continues to identify and lock on new targets to achieve continuous inspection of multiple targets.

6. The method for intelligent inspection by drones for particulate matter pollution prevention and control according to claim 5, characterized in that: The meteorological environment parameters include temperature, humidity, wind direction, and wind speed.

7. The method for intelligent inspection by drones for particulate matter pollution prevention and control according to claim 6, characterized in that: The terminal intelligent control platform compares the inspection results of the same pollution source in different time periods, forms a "look back" report, automatically generates a "closed-loop rate" for pollution source treatment, and sends it to relevant personnel.

Images