Intelligent monitoring method and system based on drone

Abstract

The present application provides an intelligent monitoring method and system based on a drone, including determining, for each row of pixels in the second spectral image data, whether the spectral data of the previous pixel and the spectral data of the current pixel meet a replacement condition. If so, the spectral data of the previous pixel is used to replace the spectral data of the current pixel; if not, the spectral data of the current pixel is maintained unchanged, all pixels are traversed to obtain third spectral image data, and the third spectral image data is analyzed and processed to obtain water quality parameters in the monitored water area. Since a large amount of identical spectral data exists in the spectral image data, the amount of data processing can be reduced and the data processing efficiency can be improved during subsequent processing of the spectral data. There is no need to select a few bands from multiple bands of data for analysis and processing in order to reduce the amount of data, which can improve the accuracy of data analysis and processing.

Description

Technical Field

[0001] The present application relates to the field of environmental monitoring technology, and in particular to an intelligent monitoring method and system based on drones. Background Art

[0002] Due to the limitations of traditional ground-based monitoring methods, such as limited field of view and low efficiency, drone technology has been introduced into the field of environmental protection due to its advantages in high-altitude operation and is used to monitor illegal sewage discharge.

[0003] More specifically, drones are equipped with various sensors to collect river data and transmit it back to the ground in real time, where it is processed by a data processing center. Due to the large amount of data collected by drones, improving the data processing capabilities of ground-based data processing centers is an urgent issue. Summary of the Invention

[0004] The embodiments of the present application provide an intelligent monitoring method and system based on drones to improve the data processing efficiency of the background server.

[0005] Some embodiments of the present application provide a drone-based intelligent monitoring method, which is applied to a backend server and includes:

[0006] The UAV is scheduled based on the flight mission and the flight capability of the UAV, and a corresponding flight path is assigned to the UAV; so that the UAV flies along the flight path during the flight time, and controls the carried hyperspectral detector to collect first spectral image data; the flight mission includes at least one flight path and the flight time of the flight path;

[0007] receiving first spectral image data returned by the UAV, preprocessing the first spectral image data, and obtaining second spectral image data;

[0008] For the current pixel point in each row of pixels in the second spectral image data, determine whether the spectral data of the previous pixel point and the spectral data of the current pixel point meet a replacement condition; if so, replace the spectral data of the current pixel point with the spectral data of the previous pixel point; if not, maintain the spectral data of the current pixel point unchanged, traverse all pixels in each row of pixels in the second spectral image data, and obtain third spectral image data;

[0009] Extracting features from the third spectral image data, and classifying the extracted features to obtain water quality parameters of the water area corresponding to the third spectral image;

[0010] It is determined whether the sewage discharged by the monitored factory meets the water quality standard according to the water quality parameters of the water area corresponding to the third spectral image.

[0011] In the above technical solution, the drone collects first spectral image data along its flight path. The backend server preprocesses the first spectral image data to obtain second spectral image data. Subsequently, when the spectral data of two previous pixels in the second spectral image meet a replacement condition, the backend server replaces the spectral data, making the spectral data of the two adjacent pixels identical. This reduces the amount of data to be processed during subsequent spectral data processing, improving data processing efficiency. Furthermore, there is no need to select one or more bands from multiple bands for analysis and processing to reduce data volume, which improves the accuracy of subsequent data analysis and processing.

[0012] In a possible implementation, replacing the spectral data of the current pixel with the spectral data of the previous pixel specifically includes:

[0013] Obtaining the spectral data identifier associated with the previous pixel point, and establishing an association relationship between the current pixel point and the spectral data identifier associated with the previous pixel point; wherein a mapping relationship exists between the spectral data identifier and the shared spectral data, and the shared spectral data is stored in a shared storage area;

[0014] Accordingly, the spectral data of the current pixel is maintained unchanged, specifically including:

[0015] A new spectral data identifier is randomly generated, the spectral data of the current pixel point is used as the new shared spectral data, the new shared spectral data is stored in the shared storage area, and an association relationship is established between the current pixel point and the new spectral data identifier.

[0016] In the above technical solution, if the spectral data of two adjacent pixels are similar, the spectral data is replaced, resulting in a large amount of identical spectral data. To reduce data storage, a shared storage area is set up in the backend server to store shared spectral data. Each shared spectral data is assigned a corresponding spectral data identifier, and each pixel is associated with a spectral data identifier. This reduces the amount of spectral data storage and eliminates the need to repeatedly calculate the features of the same spectral data during subsequent processing, reducing the amount of data calculation.

[0017] In a possible implementation, feature extraction is performed on the third spectral image data, and the extracted features are classified to obtain water quality parameters of the water body corresponding to the third spectral image, specifically including:

[0018] Extracting features from each spectral data in the shared storage area, and classifying the extracted features to obtain water quality parameters corresponding to each spectral data;

[0019] The water quality parameters of the water area corresponding to each pixel point are obtained according to the spectral data identifier associated with each pixel point and the water quality parameters of the spectral data associated with each spectral data identifier.

[0020] In the above technical solution, the spectral data in the shared storage area is analyzed and processed to obtain water quality parameters reflected by the spectral data. Subsequently, the water quality parameters reflected by the spectral data are associated with each pixel through spectral data identification, thereby obtaining the water quality parameters of the water area corresponding to each pixel. Because the third spectral image data after the replacement process contains a large amount of identical data, it is not necessary to process the spectral data of each pixel. Only the spectral data in the shared storage area is processed, which can reduce the data processing volume and improve data processing efficiency.

[0021] In a possible implementation, the spectral data includes light intensities in multiple bands, and determining whether the spectral data of the previous pixel and the spectral data of the current pixel meet a replacement condition specifically includes:

[0022] Calculate the light intensity difference between the light intensity of each band of the previous pixel and the light intensity of the corresponding band of the current pixel to obtain the light intensity difference of multiple bands;

[0023] Calculate the average value of the light intensity difference of all bands, and calculate the variance of the light intensity difference of all bands;

[0024] If the average value of the light intensity difference is less than the preset average value threshold, and the variance of the light intensity difference is less than the preset variance threshold, it is determined that the replacement condition is met;

[0025] If the average value of the light intensity difference is greater than or equal to the preset average value threshold, or the variance of the light intensity difference is greater than or equal to the preset variance threshold, it is determined that the replacement condition is not met.

[0026] In the above technical solution, the spectral data of the current pixel point and the spectral data of the previous pixel point both include the intensities of multiple bands. The light intensity difference between the light intensity of each band of the previous pixel point and the light intensity of the corresponding band of the current pixel point is calculated. By performing statistical analysis on the light intensity difference, the similarity between the spectral data of the current pixel point and the spectral data of the previous pixel point can be obtained. Based on the similarity, it is determined whether to replace the spectral data. In this way, it can be ensured that the information can be retained after the spectral data is replaced, and no information loss will occur, thereby improving the data processing efficiency and the accuracy of data processing.

[0027] In a possible implementation, preprocessing the first spectral image data to obtain the second spectral image data specifically includes:

[0028] Denoising the first spectral image data to obtain denoised spectral image data;

[0029] performing baseline correction on the denoised spectral image data to obtain baseline-corrected spectral image data;

[0030] The baseline-corrected spectral image data is subjected to black-and-white correction to obtain third spectral image data.

[0031] In the above technical solution, by successively performing denoising, baseline correction, and black-and-white correction on the first spectral image data, the image quality of the third spectral image data can be improved.

[0032] In one possible implementation, before scheduling the drones based on the flight missions and the flight capabilities of the drones and assigning corresponding flight paths to the drones, the method further includes:

[0033] Obtain the pollution discharge standards of the factories to be monitored, the real-time pollution discharge data of the factories to be monitored, and the terrain data of the areas to be monitored;

[0034] The flight mission is generated according to the pollution discharge standards of the factories to be monitored, the real-time pollution discharge data of the factories to be monitored, and the terrain data of the areas to be monitored.

[0035] In the above technical solution, flight missions are generated based on obtaining the pollution discharge standards of the factories to be monitored, the real-time pollution discharge data of the factories to be monitored, and the terrain data of the areas to be monitored. The missions can be executed with real-time reference to the pollution discharge conditions of the factories to be monitored, and targeted monitoring can be carried out to improve monitoring efficiency and accuracy.

[0036] In a possible implementation, the real-time sewage discharge data includes current sewage discharge data and historical sewage discharge data, and both current sewage discharge data and historical sewage discharge data include sewage discharge time, sewage discharge volume, and sewage quality;

[0037] Accordingly, flight missions are generated based on the pollution discharge standards of the factories to be monitored, the real-time pollution discharge data of the factories to be monitored, and the terrain data of the areas to be monitored, including:

[0038] For each factory to be monitored, the discharge time coefficient is set according to the discharge time of the factory to be monitored and the required discharge time; the discharge water quality coefficient is set according to the discharge water quality and water quality requirements of the factory to be monitored; the discharge volume coefficient is set according to the discharge volume and maximum discharge volume of the factory to be monitored;

[0039] For each factory to be monitored, the trained emission prediction model is used to obtain the predicted emission volume for the day; and the emission prediction coefficient is set based on the predicted emission volume for the day and the actual emission volume in the emission data for the day;

[0040] For each factory to be monitored, the monitoring frequency weight of each factory to be monitored is generated according to the discharge time coefficient, discharge water quality coefficient, discharge volume coefficient and discharge prediction coefficient;

[0041] The terrain data of the area to be monitored and the weight of the number of monitoring times of each factory to be monitored are input into the mission generation model to obtain the flight mission generated by the mission generation model.

[0042] In the above technical solution, the discharge time system, discharge water quality coefficient, discharge volume coefficient and discharge prediction coefficient are calculated respectively according to the discharge time, discharge water quality, discharge volume and predicted discharge volume of the factory to be monitored, and the monitoring frequency weight of the factory to be monitored is calculated based on the above four coefficients. Compared with increasing the monitoring frequency for all factories to be monitored, increasing the monitoring frequency weight of the factory to be monitored in a targeted manner can reduce the flight distance of the drone and reduce the amount of data collection.

[0043] Some embodiments of the present application provide an intelligent monitoring system based on a drone, including a drone and a background server. The drone is equipped with a hyperspectral detector and an image sensor, and the background server is used to execute the intelligent monitoring method involved in the above embodiments.

[0044] Some embodiments of the present application provide a backend server, including: a memory, a processor;

[0045] Memory stores computer-executable instructions;

[0046] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above-mentioned embodiments and various possible implementations.

[0047] An embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the above-mentioned embodiment and various possible implementation methods.

[0048] An embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned embodiment and various possible implementation methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the present application and, together with the description, serve to explain the principles of the present application.

[0050] Figure 1 A schematic diagram of a scene of an area to be monitored provided in some embodiments of the present application;

[0051] Figure 2 A flow chart of the intelligent monitoring method based on drones provided in this application;

[0052] Figure 3A Schematic diagram of the replacement of spectral image data provided in this application Figure 1 ;

[0053] Figure 3B Schematic diagram of the replacement of spectral image data provided in this application Figure 2 ;

[0054] Figure 4 This is a schematic diagram of the spectral data of the previous pixel and the spectral data of the current pixel provided in this application.

[0055] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0056] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0057] To address these issues, in the embodiments provided herein, a drone equipped with a hyperspectral detector and image sensor conducts flight inspections over the monitored area, collecting spectral and two-dimensional image data of the river and sewage outlets. The collected spectral image data undergoes preprocessing, including denoising, baseline correction, and black-and-white correction, to improve image quality. Furthermore, the two-dimensional image data is used to identify suspended matter and oil pollution, providing a preliminary assessment of water quality.

[0058] For the preprocessed spectral image data, each row of pixels is traversed to determine whether the spectral data of adjacent pixels meets the replacement conditions. If the conditions are met, the spectral data of the previous pixel is used to replace the spectral data of the current pixel, thereby reducing the amount of data to be processed and improving processing efficiency. This step provides a more concise and efficient data foundation for subsequent feature extraction and classification by reducing redundant information in the spectral image data. Feature extraction is performed on the optimized spectral image data, and the extracted features are classified and processed using machine learning or deep learning algorithms to obtain water quality parameters. Based on the water quality parameters, it is determined whether the pollution discharge behavior of the monitored factory meets the water quality standards. If the standards are not met, an early warning mechanism is triggered, notifying relevant departments for further investigation and processing. In addition, in the embodiments provided in this application, the system includes two parts: an unmanned aerial vehicle and a backend server. The unmanned aerial vehicle is responsible for data acquisition and preliminary processing, while the backend server is responsible for in-depth data analysis, feature extraction, classification, monitoring and early warning tasks. Data is transmitted between the unmanned aerial vehicle and the backend server via wireless communication to ensure the real-time and accuracy of the monitoring tasks.

[0059] In summary, the embodiments of the present application significantly reduce redundant information in spectral image data through a spectral data replacement algorithm, thereby improving data processing efficiency. By combining data from a hyperspectral detector and image sensor, accurate assessments of water quality conditions are achieved, effectively improving the accuracy of monitoring and early warning systems. The system can adjust the drone's flight path and monitoring frequency based on actual needs, demonstrating strong flexibility and scalability. The entire monitoring process is intelligent and automated, reducing manual intervention and improving monitoring efficiency.

[0060] Figure 1 Schematic diagram of the scene of the area to be monitored provided in some embodiments of the present application, such as Figure 1 As shown, the specific application scenario of this application is to monitor sewage discharge from factories in a certain monitoring area. The circle represents the factory, which is distributed along the river, and the triangle represents the sewage outlet of the factory.

[0061] Drone technology, due to its high-altitude operational capabilities, has been introduced into the environmental protection field, where it is used to monitor illegal sewage discharge. More specifically, drones are equipped with various sensors to collect river data and transmit it back to the ground in real time, where it is processed by a data processing center. Due to the large amount of data collected by drones, improving the processing capabilities of ground-based data processing centers is a pressing issue.

[0062] The drone-based intelligent monitoring method and system provided in the present application, for spectral image data collected by the drone, uses the spectral data of the previous pixel point to replace the spectral data of the current pixel point when the spectral data of the previous pixel point and the spectral data of the current pixel point meet the replacement condition, so that the spectral data of two adjacent pixels are the same. Therefore, a large amount of identical spectral data is included in the processed spectral image data, which can reduce the data processing amount and improve data processing efficiency when the spectral data is subsequently processed.

[0063] The following specific embodiments describe in detail the technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0064] Some embodiments of the present application provide a drone-based intelligent monitoring system, comprising a drone and a backend server. During patrols in a monitored area, the drone collects river data from that area, for example, from each factory's sewage outlet. The data is then transmitted back to the backend server. The backend server uses an algorithm to analyze and process the river data to determine whether the sewage discharged from each factory's sewage outlet meets discharge standards.

[0065] Drones are equipped with various sensors for data collection, such as image sensors and hyperspectral detectors.

[0066] Image sensors are categorized into infrared and visible light sensors. Visible light sensors can capture river images when visibility is high, while infrared sensors can capture river images when visibility is low. River images can be used to determine the water quality at each sewage outlet. For example, this can determine whether there is floating garbage, foam, or oil in the water at the outlet.

[0067] Furthermore, due to the varying composition of river water, the absorption rate of light at different wavelengths varies. A hyperspectral detector is used to collect spectral image data at different wavelengths. Based on this data, the composition of the water at the sewage outlet is determined, and further, whether the sewage discharged from the outlet meets discharge standards.

[0068] like Figure 2 As shown, some embodiments of the present application provide an intelligent monitoring method based on a drone, which specifically includes the following steps:

[0069] S101. The backend server dispatches the drones based on the flight missions and the flight capabilities of the drones, and assigns corresponding flight paths to the drones.

[0070] A flight mission consists of at least one flight path and the flight time for each flight path. A drone's flight capability refers to its maximum flight distance and the recharge time it takes. The backend server schedules drones based on the flight time, length, and flight capability of each flight path, assigning them flight paths.

[0071] Specifically, the backend server first schedules the flight based on the pre-set flight mission and the drone's flight capabilities. A flight mission consists of at least one flight path and the flight time for each path. Factors considered in determining a drone's flight capabilities include its maximum flight range, speed, and battery life. The backend server comprehensively considers these factors and assigns the optimal flight path to the drone, ensuring it can fly along the designated path and complete the data collection mission within the specified timeframe.

[0072] For example, if the flight times of two flight paths differ by more than 4 hours, the drone's battery fully charges in 3 hours, and the maximum flight distance of the drone is greater than the distance of the flight paths, then two flight paths can be assigned to the drone.

[0073] S102: The UAV flies along the flight path during the flight time, and controls the carried hyperspectral detector to collect first spectral image data.

[0074] The drone is equipped with a hyperspectral detector, which is activated while the drone is flying along the flight path. The hyperspectral detector collects spectral image data near the river channel, which is recorded as first spectral image data. The first spectral image data includes spectral data of multiple pixels. The spectral data includes light intensity in multiple wavelength bands.

[0075] Specifically, after receiving a flight mission, the drone flies according to the assigned flight path and duration. During the flight, the drone's onboard hyperspectral detector begins to collect spectral image data of the river channel and the vicinity of the sewage outlet, i.e., the first spectral image data. Optionally, this data can include light intensity information at different wavelengths, reflecting the composition and concentration of the water.

[0076] S103: The backend server receives the first spectral image data returned by the drone, pre-processes the first spectral image data, and obtains second spectral image data.

[0077] More specifically, preprocessing includes denoising, baseline correction, and black-and-white correction, thereby improving image quality.

[0078] Denoising specifically involves smoothing algorithms or inverse methods. Smoothing algorithms include moving window averaging or least squares fitting. Baseline correction includes multivariate scattering correction or standard normal variate transformation. Black-white correction includes dark current correction and reflection correction.

[0079] In some embodiments, the first spectral image data is denoised to obtain denoised spectral image data, the denoised spectral image data is baseline corrected to obtain baseline-corrected spectral image data, and the baseline-corrected spectral image data is subjected to black-white correction to obtain third spectral image data.

[0080] Specifically, after receiving the first spectral image data returned by the drone, the backend server first performs preprocessing. This preprocessing includes steps such as denoising, baseline correction, and black-and-white correction. Denoising uses a filtering algorithm to remove noise from the image; baseline correction eliminates baseline drift in the spectral image; and black-and-white correction ensures appropriate brightness and contrast for subsequent processing. After preprocessing, the second spectral image data is obtained.

[0081] In the above technical solution, by successively performing denoising, baseline correction, and black-and-white correction on the first spectral image data, the image quality of the third spectral image data can be improved.

[0082] It is worth noting that the above solution has at least the following technical effects:

[0083] First, improve the quality of spectral image data: Denoising the first spectral image data effectively removes noise and interfering signals, resulting in purer spectral data. Subsequent baseline correction eliminates baseline drift, ensuring the accuracy and stability of the spectral data. Finally, black-and-white correction further adjusts the brightness and contrast of the spectral image, making it clearer and easier to read. These preprocessing steps collectively improve the quality of the spectral image data and provide a reliable foundation for subsequent data analysis and processing.

[0084] Second, it enhances the accuracy of data analysis: High-quality spectral image data is a prerequisite for accurately extracting water quality parameters. Preprocessing the first spectral image data, including denoising, baseline correction, and black-white correction, significantly reduces noise and errors, improving the accuracy of data analysis. This helps more accurately determine the types and concentrations of pollutants in water bodies, providing a more reliable basis for environmental protection and regulation.

[0085] Third, improving system robustness: In practical applications, spectral image data collected by drones is often affected by a variety of factors, such as lighting conditions, atmospheric conditions, and sensor performance. These factors can cause fluctuations and errors in the spectral data. Preprocessing the first spectral image data effectively reduces the impact of these external factors on data analysis, improving the robustness and stability of the system. This enables the system to maintain stable performance in a variety of complex environments, ensuring the smooth execution of monitoring tasks.

[0086] Fourth, optimize system resource utilization: High-quality spectral image data can reduce the time and computing resources required for subsequent data analysis and processing. Preprocessing the first spectral image data makes it easier to process and analyze, thereby reducing system resource consumption. This helps achieve higher system performance and longer battery life within limited hardware conditions, providing strong support for long-term, continuous monitoring of illegal sewage discharge.

[0087] To sum up, in the above scheme, the preprocessing steps of denoising, baseline correction and black and white correction are performed on the first spectral image data, which has multiple technical effects such as improving the quality of spectral image data, enhancing the accuracy of data analysis, improving system robustness and optimizing system resource utilization, thereby improving the overall performance and monitoring efficiency of the system.

[0088] S104: For each row of pixels in the second spectral image data, the backend server determines whether the spectral data of the previous pixel and the spectral data of the current pixel meet a replacement condition. If so, the spectral data of the previous pixel is used to replace the spectral data of the current pixel. If not, the spectral data of the current pixel is maintained unchanged, and all pixels in each row of pixels in the second spectral image data are traversed to obtain third spectral image data.

[0089] The previous pixel and the current pixel are adjacent pixels. If the difference between the spectral data of the previous pixel and the current pixel is large, the replacement condition is not met and the spectral data of the current pixel remains unchanged. If the difference between the spectral data of the previous pixel and the current pixel is small, the replacement condition is met and the spectral data of the adjacent pixel is used to replace the spectral data of the current pixel. This reduces the amount of data processing during feature extraction. The third spectral image data is obtained by traversing all pixels in each row of pixels in the second spectral image data.

[0090] Specifically, the backend server performs traversal processing for each row of pixels in the second spectral image data. For the current pixel, it is determined whether its spectral data meets the replacement condition with the previous pixel. The replacement condition is usually determined based on the similarity between the spectral data, for example, the difference between the light intensity of two pixels in each band is calculated, and the average value and variance of the difference are calculated. If the average value and variance of the difference are both less than the preset threshold, it is considered that the spectral data similarity is high and the replacement condition is met, and the spectral data of the previous pixel is used to replace the spectral data of the current pixel; if the condition is not met, the spectral data of the current pixel is maintained unchanged. After traversing all pixels, the optimized third spectral image data is obtained.

[0091] Figure 3A and Figure 3B is a replacement diagram for spectral image data, such as Figure 3A and Figure 3B As shown, for example, the second spectral image data includes 10 rows of pixels. For the first row of pixels, the pixel in the first row and first column does not have a previous pixel, so the spectral data of the pixel in the first row and first column remains unchanged. The pixel in the first row and second column is updated to the current pixel. A determination is made as to whether the difference between the pixel in the first row and first column and the pixel in the first row and second column is significant. If so, the spectral data of the pixel in the first row and second column remains unchanged. If the difference is small, the spectral data of the pixel in the first row and first column is used to replace the spectral data of the pixel in the first row and second column.

[0092] Update the pixel at row 1, column 3 to the current pixel. Determine whether the difference between the pixel at row 1, column 2 and the pixel at row 1, column 3 is significant. If so, keep the spectral data of the pixel at row 1, column 3 unchanged. If the difference is small, replace the spectral data of the pixel at row 1, column 3 with the spectral data of the pixel at row 1, column 2.

[0093] And so on, until all pixels in the first row are traversed.

[0094] For the pixel in row 2, row 2, column 1, there is no previous pixel, so the spectral data for the pixel in row 2, column 1 remains unchanged. The pixel in row 2, column 2 is updated as the current pixel. A check is performed to determine whether the difference between the pixel in row 2, column 1 and the pixel in row 2, column 2 is significant. If so, the spectral data for the pixel in row 2, column 2 remains unchanged. If the difference is small, the spectral data for the pixel in row 2, column 1 is used to replace the spectral data for the pixel in row 2, column 2.

[0095] Update the pixel at row 2, column 3 to the current pixel. Determine whether the difference between the pixel at row 2, column 2 and the pixel at row 2, column 3 is significant. If so, keep the spectral data of the pixel at row 2, column 3 unchanged. If the difference is small, replace the spectral data of the pixel at row 2, column 3 with the spectral data of the pixel at row 2, column 2.

[0096] And so on, until all pixels in the second row are traversed.

[0097] Through the above method, the traversal of all rows of pixels is completed.

[0098] S105 : The backend server extracts features from the third spectral image data, and classifies the extracted features to obtain water quality parameters of the water area corresponding to the third spectral image.

[0099] Among them, the existing feature extraction algorithm can be used to extract features from the third spectral image data to obtain feature data of the third spectral image data, and the existing classification algorithm can be used to classify the feature data of the third spectral image data to obtain water quality parameters of the water area corresponding to the third spectral image.

[0100] Since the spectral image data is collected by the drone as it flies along the flight path, the drone can collect multiple frames of first spectral image data. By processing these multiple frames of first spectral image data in the aforementioned manner, multiple frames of second spectral image data can be obtained. The multiple frames of second spectral image data are then processed to obtain multiple frames of third spectral image data. The water quality parameters corresponding to the water area in the third spectral image data can reflect the water quality parameters of the river along the flight path.

[0101] Specifically, feature extraction is performed on the third spectral image data to identify key features that reflect water quality. These features may include light intensity values ​​in specific wavelength bands, the shape and slope of the spectral curve, and other factors. Machine learning or deep learning algorithms are then used to classify these extracted features to obtain water quality parameters for the water area corresponding to the third spectral image. These parameters may include key water quality indicators such as dissolved oxygen content, chemical oxygen demand, and ammonia nitrogen content.

[0102] S106. The backend server determines whether the sewage discharged by the monitored factory meets the water quality standard based on the water quality parameters of the water area corresponding to the third spectral image.

[0103] If the water quality parameters of a certain area within the river do not meet the water quality standards, it is determined that the sewage discharge of at least one monitored factory located in that area does not meet the water quality standards. If the water quality parameters of a certain area within the river do meet the water quality standards, it is determined that the sewage discharge of the monitored factory located in that area does meet the water quality standards.

[0104] Specifically, based on the acquired water quality parameters, the backend server determines whether the monitored factory's discharge meets water quality standards. This step typically involves comparative analysis against pre-set water quality standards. If the water quality parameters exceed the standard range, the factory's discharge is deemed to be substandard, triggering an early warning mechanism. This mechanism may include sending an alarm to relevant departments and displaying an abnormality notification on the monitoring interface, enabling timely intervention and resolution.

[0105] In the above technical solution, a drone collects first spectral image data along its flight path. A backend server preprocesses the first spectral image data to obtain second spectral image data. Subsequently, when the spectral data of two adjacent pixels in the second spectral image data meet a replacement condition, the backend server replaces the spectral data, making the spectral data of the two adjacent pixels identical. This reduces the amount of data to be processed during subsequent spectral data processing, improving data processing efficiency. Furthermore, there is no need to select one or more bands from multiple bands for analysis and processing to reduce data volume, which improves the accuracy of subsequent data analysis and processing.

[0106] In one possible implementation, the spectral data includes light intensities in multiple bands, and determining whether the spectral data of the previous pixel and the spectral data of the current pixel meet a replacement condition specifically includes:

[0107] S21, calculating the light intensity difference between the light intensity of each wavelength band of the previous pixel and the light intensity of the corresponding wavelength band of the current pixel.

[0108] Figure 4 This is a schematic diagram of the spectrum data of the previous pixel and the spectrum data of the current pixel provided in this application, such as Figure 4 For example, the spectrum data for the previous pixel includes light intensity data for 10 bands, namely, the light intensity In10λ1 for band λ1, the light intensity In11λ2 for band λ2, ..., and the light intensity In19λ10 for band λ10. The spectrum data for the current pixel includes light intensity data for 10 bands, namely, the light intensity In20λ1 for band λ1, the light intensity In21λ2 for band λ2, ..., and the light intensity In29λ10 for band λ10.

[0109] S22. Calculate the light intensity difference of band λ1 as (In10λ1-In20λ1), calculate the light intensity difference of band λ2 as (In11λ2-In21λ2), ..., calculate the light intensity difference of band λ10 as (In19λ10-In29λ10), and obtain the light intensity differences of 10 bands.

[0110] S23. Calculate the average value of the light intensity difference of all bands, and calculate the variance of the light intensity difference of all bands.

[0111] The average value of the light intensity difference between the wavelength band λ1 and the wavelength band λ10 is calculated, and the variance of the light intensity difference between the wavelength band λ1 and the wavelength band λ10 is calculated.

[0112] S24: If the average value of the light intensity difference is greater than or equal to the preset average value threshold, or the variance of the light intensity difference is greater than or equal to the preset variance threshold, it is determined that the replacement condition is not met.

[0113] The variance and average of the intensity difference values ​​reflect the difference between the spectral data of the current pixel and the spectral data of the previous pixel. If both the average and variance of the intensity difference values ​​are large, it indicates that the spectral data of the current pixel differs significantly from the spectral data of the previous pixel, and the spectral data of the previous pixel cannot be used to replace the spectral data of the current pixel.

[0114] S25: If the average value of the light intensity difference is less than the preset average value threshold, and the variance of the light intensity difference is less than the preset variance threshold, it is determined that the replacement condition is met.

[0115] If the average value and the variance of the light intensity difference are relatively small, it means that the spectral data of the current pixel point is very similar to the spectral data of the previous pixel point. In this case, the spectral data of the current pixel point can be replaced by the spectral data of the previous pixel point, thereby reducing the data processing amount of subsequent feature extraction.

[0116] In the above technical solution, the spectral data of the current pixel point and the spectral data of the previous pixel point both include the intensities of multiple bands. The light intensity difference between the light intensity of each band of the previous pixel point and the light intensity of the corresponding band of the current pixel point is calculated. By performing statistical analysis on the light intensity difference, the similarity between the spectral data of the current pixel point and the spectral data of the previous pixel point can be obtained. Based on the similarity, it is determined whether to replace the spectral data. In this way, it can be ensured that the information can be retained after the spectral data is replaced, and no information loss will occur, thereby improving the data processing efficiency and the accuracy of data processing.

[0117] It is worth noting that the above solution has at least the following technical effects:

[0118] First, ensure the accuracy of spectral data: By calculating the intensity difference between each band of the previous pixel and the intensity of the corresponding band of the current pixel, and calculating the mean and variance of these differences, the similarity of the spectral data of the two pixels can be accurately assessed. Only when the mean and variance of the intensity difference are both less than a preset threshold, the spectral data is considered sufficiently similar to meet the replacement criteria. This method ensures that the replaced spectral data accurately reflects the information of the original data and avoids data distortion caused by the replacement.

[0119] Second, it improves data processing efficiency: In spectral image data, the spectral data of adjacent pixels often have high similarity. By setting appropriate mean and variance thresholds, pixels that meet the replacement criteria can be quickly screened, reducing unnecessary data processing. This approach significantly improves data processing efficiency, enabling the system to more quickly optimize and analyze spectral image data.

[0120] Third, the system's adaptability and robustness are enhanced. Different monitoring scenarios and lighting conditions can cause fluctuations in spectral data. By calculating the mean and variance of light intensity differences and using this information to determine replacement conditions, the system can automatically adapt to these changes. Even when spectral data fluctuates slightly due to external factors, the system can still accurately determine which pixels' spectral data can be replaced, thereby maintaining monitoring stability and accuracy. This adaptability and robustness enable the system to operate stably in a variety of complex environments.

[0121] Fourth, it supports more refined monitoring and analysis: By calculating and analyzing multi-band light intensity differences in spectral data, the system can capture more subtle changes in spectral features. These changes may reflect changes in the concentration of different pollutants in the water or subtle fluctuations in water quality parameters. By accurately determining whether replacement conditions are met, the system can retain this important characteristic information, providing more accurate and comprehensive data support for subsequent water quality parameter extraction and classification. This facilitates more refined monitoring and analysis, improving monitoring accuracy and reliability.

[0122] To sum up, in the above scheme, the technical measures taken to determine whether the spectral data of the previous pixel point and the spectral data of the current pixel point meet the replacement conditions have multiple technical effects such as ensuring the accuracy of spectral data, improving data processing efficiency, enhancing system adaptability and robustness, and supporting more refined monitoring and analysis, thereby improving the overall performance and monitoring accuracy of the system.

[0123] In some possible implementations, replacing the spectral data of the current pixel with the spectral data of the previous pixel may include:

[0124] S31 , obtaining the spectral data identifier associated with the previous pixel, and establishing an association relationship between the current pixel and the spectral data identifier associated with the previous pixel; a mapping relationship exists between the spectral data identifier and the shared spectral data, and the shared spectral data is stored in a shared storage area.

[0125] A shared storage area is reserved in the background server, and the shared storage area is used to store shared spectral data. A mapping relationship exists between the spectral data identifier and the shared spectral data, and a mapping relationship exists between the spectral data identifier and each pixel point.

[0126] If the spectral data of the current pixel point is highly similar to the spectral data of the previous pixel point, the replacement condition is met. The spectral data identifier associated with the previous pixel point is obtained, and an association relationship between the spectral data identifier associated with the current pixel point and the previous pixel point is established.

[0127] Figure 3A Schematic diagram of the replacement of spectral image data Figure 1 ,like Figure 3A As shown, spectral image data 100 includes 10 rows and 6 columns of pixels. For example, for the pixel in row 1, row 1, column 1, there is no previous pixel, so the spectral data is retained. A first spectral data identifier (identifier 1 in the figure) is randomly generated, and the spectral data of the pixel in row 1, column 1 is used as the first shared spectral data (data 1 in the figure). The first spectral data identifier is associated with the first shared spectral data, and the first shared spectral data is stored in shared storage area 200. The pixel in row 1, column 1 is associated with the first spectral data identifier.

[0128] Update the pixel in row 1, column 2 to the current pixel. Read the first shared spectral data corresponding to the first spectral data identifier from the shared storage area. Determine whether the first shared spectral data and the spectral data of the pixel in row 1, column 2 meet the replacement criteria. If the difference between the pixel in row 1, column 1 and the pixel in row 1, column 2 is small, associate the pixel in row 1, column 2 with the first spectral data identifier. This eliminates the need to store multiple copies of the same spectral data, reducing data storage requirements.

[0129] In some possible implementations, maintaining the spectral data of the current pixel unchanged includes:

[0130] S32 , randomly generate a new spectral data identifier, use the spectral data of the current pixel as new shared spectral data, store the new shared spectral data in a shared storage area, and establish an association relationship between the current pixel and the new spectral data identifier.

[0131] If the spectral data of the current pixel is less similar to the spectral data of the previous pixel, and the replacement condition is not met, new spectral data must be stored in the shared storage area. A new spectral data identifier is randomly generated, and the spectral data of the current pixel is used as the new shared spectral data. The new shared spectral data is stored in the shared storage area, and an association is established between the new spectral data identifier and the new shared spectral data. An association is also established between the current pixel and the new spectral data identifier.

[0132] Figure 3B Schematic diagram of the replacement of spectral image data Figure 2 ,like Figure 3BAs shown, for example: for the pixel in the first row, the pixel in the second column of the first row is updated as the current pixel, the first common spectral data corresponding to the first spectral data identifier is read from the common storage area, and it is determined whether the first common spectral data and the spectral data of the pixel in the second column of the first row meet the replacement condition. If the difference between the pixel in the first column of the first row and the pixel in the second column of the first row is relatively large, a second spectral data identifier (identifier 2 in the figure) is generated, and the spectral data of the pixel in the second column of the first row is used as the second common spectral data (data 2 in the figure). A mapping relationship between the second spectral data identifier and the second common spectral data is established, and a mapping relationship between the second spectral data identifier and the pixel in the second column of the first row is established.

[0133] In the above technical solution, if the spectral data of two adjacent pixels are similar, the spectral data is replaced, resulting in a large amount of identical spectral data. To reduce data storage, a shared storage area is set up in the backend server to store shared spectral data. Each shared spectral data is assigned a corresponding spectral data identifier, and each pixel is associated with a spectral data identifier. This reduces the amount of spectral data storage and eliminates the need to repeatedly calculate the features of the same spectral data during subsequent processing, reducing the amount of data calculation.

[0134] It is worth noting that the above solution has at least the following technical effects:

[0135] First, improving the consistency of spectral image data: In spectral image data, the spectral data of adjacent pixels are often highly similar, especially in conditions of stable water bodies and uniform illumination. By obtaining the spectral data identifier associated with the previous pixel and replacing the spectral data of the current pixel with that spectral data, redundant information in the spectral image can be effectively reduced, improving data consistency. This consistency not only facilitates subsequent feature extraction and classification, but also reduces the computational complexity of data processing to a certain extent, improving overall monitoring efficiency.

[0136] Second, optimizing spectral data storage and access: While maintaining the spectral data of the current pixel, a new spectral data identifier is randomly generated and stored as new shared spectral data in a shared storage area. This allows for centralized management and efficient access of spectral data. This storage method avoids duplicate data storage and saves storage space. Furthermore, by establishing an association between pixels and spectral data identifiers, the corresponding spectral data can be quickly located, improving the speed and efficiency of data access.

[0137] Third, it enhances system flexibility and scalability: This approach enables the system to flexibly handle changes in spectral image data. Whether replacing the current pixel's spectral data with the previous one, or maintaining the current pixel's spectral data unchanged and generating a new spectral data identifier, the system can adapt and optimize based on actual needs. This flexibility helps the system adapt to different monitoring scenarios and changing requirements, while also facilitating system expansion and upgrades.

[0138] Fourth, improving monitoring accuracy and reliability: By optimizing the processing and storage structure of spectral image data, the accuracy and reliability of monitoring are improved. Consistent and optimized spectral data provides a more accurate data foundation for subsequent feature extraction and classification processing, enabling more accurate judgment of whether the pollutant discharge behavior of the monitored plant meets water quality standards. Furthermore, centrally managed and efficiently accessed spectral data ensures the continuity and stability of the monitoring process, improving system reliability.

[0139] In summary, the technical measures taken to replace the spectral data of the current pixel with the spectral data of the previous pixel and to maintain the spectral data of the current pixel unchanged not only improve the consistency and storage efficiency of spectral image data, but also enhance the flexibility and scalability of the system, thereby improving the accuracy and reliability of monitoring.

[0140] In some possible implementations, feature extraction is performed on the third spectral image data, and the extracted features are classified to obtain water quality parameters of the water area corresponding to the third spectral image, specifically including:

[0141] S41 , extracting features from each spectral data in the shared storage area, and classifying the extracted features to obtain water quality parameters corresponding to each spectral data.

[0142] For the monitored substance, the absorption of light at different wavelengths is obtained for the substance at different concentration ranges. A characteristic wavelength is selected from the absorption at different wavelengths, and the absorption at this characteristic wavelength is used to represent the concentration of the monitored substance. In other words, a mapping relationship is established between the absorption at the characteristic wavelength and the concentration of the monitored substance.

[0143] For each spectral data, the wavelength corresponding to the absorption peak in the spectral data is extracted, and the wavelength corresponding to the absorption peak is compared with the characteristic wavelength of each monitored substance. If the wavelength corresponding to the absorption peak is the same as the characteristic wavelength of each monitored substance, it means that the water body contains the substance, and the substance content can be determined based on the mapping relationship between the intensity corresponding to the absorption peak and the absorption amount of the characteristic wavelength and the monitored substance.

[0144] For example, if substance A absorbs significantly at wavelength λm, and other monitored substances do not absorb at wavelength λm, then wavelength λm is used as the characteristic wavelength of substance A, and the absorbance of substance A at wavelength λm at different concentrations is obtained. By analyzing the data collected by the hyperspectral detector and determining that the absorption peak wavelength is near wavelength λm, it can be determined that substance A is present in the water, and the concentration of substance A can be determined based on the light intensity at the absorption peak wavelength.

[0145] S42 . Obtain the water quality parameter of the water area corresponding to each pixel point according to the spectral data identifier associated with each pixel point and the water quality parameter of the spectral data associated with each spectral data identifier.

[0146] The water quality parameters include the content of each monitored substance. By performing feature extraction and classification processing on each spectral data in the shared storage area, the content of the monitored substance corresponding to each spectral data is obtained, and then the water quality parameters corresponding to the spectral data are obtained.

[0147] The spectral data of each pixel point can be obtained from the shared storage area through the spectral data identifier. The water quality parameters of the spectral data are also stored in the shared storage area, and then the water quality parameters of the water area corresponding to each pixel point can be obtained from the shared storage area.

[0148] In the above technical solution, the spectral data in the shared storage area is analyzed and processed to obtain water quality parameters reflected by the spectral data. Subsequently, the water quality parameters reflected by the spectral data are associated with each pixel through spectral data identification, thereby obtaining the water quality parameters of the water area corresponding to each pixel. Because the third spectral image data after the replacement process contains a large amount of identical data, it is not necessary to process the spectral data of each pixel. Only the spectral data in the shared storage area is processed, which can reduce the data processing volume and improve data processing efficiency.

[0149] It is worth noting that the above solution has at least the following technical effects:

[0150] First, it improves data processing efficiency: By extracting and classifying features from each spectral data point in a shared storage area, rather than processing the spectral data of each pixel individually, the amount of data processing is significantly reduced. Since the spectral data of adjacent pixels are often similar, a large amount of identical spectral data will be present in the third spectral image data after replacement. This method avoids repeated calculation of features for identical spectral data, thereby improving data processing efficiency and enabling the system to respond more quickly to monitoring needs.

[0151] Second, data processing accuracy is enhanced: Feature extraction and classification are performed based on the entire spectral dataset, helping to capture more comprehensive spectral feature information. Centralized processing of spectral data in a shared storage area allows for a richer data sample to be used for feature extraction and model training, thereby improving classification accuracy. This increased accuracy enables more precise assessment of water quality parameters and reduces the probability of false positives and false negatives.

[0152] Third, optimized resource utilization: Spectral data is stored in a shared storage area and associated with pixels through spectral data identifiers, enabling centralized data management and efficient access. This storage method avoids duplicate data storage and redundant processing, optimizing system resource utilization. Furthermore, by reducing the amount of data to be processed, the demand for computing resources is also lowered, enabling the system to achieve higher performance even with limited hardware.

[0153] Fourth, it supports real-time monitoring and early warning. Efficient feature extraction and classification processing enable the system to more quickly obtain water quality parameters, enabling real-time monitoring and early warning capabilities. When abnormal water quality parameters are detected, the system can quickly trigger an early warning mechanism, notifying relevant departments to take countermeasures. This real-time monitoring and early warning capability is crucial for promptly detecting and preventing illegal sewage discharge.

[0154] In summary, through the above scheme, the technical measures taken in feature extraction and classification processing of the third spectral image data not only improve the data processing efficiency and accuracy, but also optimize resource utilization and support real-time monitoring and early warning functions, thereby jointly improving the overall performance and monitoring effect of the system.

[0155] In some embodiments, before receiving the first spectral image data returned by the drone, the method further includes:

[0156] S51. The backend server receives the two-dimensional image data returned by the drone, and performs suspended matter recognition and oil pollution recognition on the two-dimensional image data.

[0157] The drone is equipped with a hyperspectral detector and an image sensor. When monitoring a sewage outlet, the image sensor is first activated to collect smaller two-dimensional image data. Existing image recognition methods are then used to identify suspended solids and oil pollution in the two-dimensional image data.

[0158] S52: If no suspended matter or oil pollution is identified, the backend server sends a start instruction of the hyperspectral detector to the UAV, so that the hyperspectral detector carried by the UAV collects first spectral image data.

[0159] Among them, if suspended matter and oil pollution are not identified, it is necessary to analyze the composition of the water body at the sewage outlet. The hyperspectral detector collects the first spectral image data, and the background server analyzes the first spectral image data to obtain the water quality parameters of the water body at the sewage outlet.

[0160] S53: If suspended matter and oil pollution are identified, the background server controls the drone to continue to fly along the flight path to the next sewage outlet.

[0161] In the above technical solution, since the hyperspectral detector collects a large amount of data, to further reduce the data processing workload of the backend server, the drone is also equipped with an image sensor. The image sensor collects image data, which includes the grayscale values ​​of multiple pixels. This image data is analyzed using existing image processing methods to determine whether visible pollutants are present in the monitored waters. If no visible pollutants are detected, the hyperspectral detector is activated, reducing the amount of data collected and the data processing workload of the backend server, thereby improving monitoring efficiency.

[0162] It is worth noting that the above solution has at least the following technical effects:

[0163] First, reduce unnecessary data collection: Before receiving the first spectral image data returned by the drone, the image data is first identified for suspended solids and oil contamination. If no suspended solids or oil contamination is detected in the image, it indicates that the water quality in the area is likely clear and free of visible pollutants. At this point, the drone is instructed to activate its hyperspectral detector for detailed spectral image data collection. This effectively avoids unnecessary hyperspectral data collection in areas without obvious pollution, thereby reducing the amount of data collected and alleviating the processing burden on the backend server.

[0164] Second, it improves monitoring efficiency: By first identifying suspended matter and oil contamination in image data, areas of potential contamination can be quickly identified. For areas without obvious contamination, monitoring frequency can be reduced or hyperspectral data collection can be skipped, focusing limited monitoring resources on areas more likely to detect problems. This targeted monitoring strategy significantly improves monitoring efficiency, enabling the system to more quickly respond to and address actual pollution incidents.

[0165] Third, optimize resource utilization: Hyperspectral detectors typically collect large amounts of data, and processing this data requires significant computing resources and time. By preprocessing and recognizing the image data, we can avoid wasting resources in areas where hyperspectral data is not needed. This resource optimization strategy enables the system to achieve higher performance and longer battery life within limited hardware conditions.

[0166] Fourth, it improves monitoring accuracy: Although suspended matter and oil pollution identification using image data is a relatively crude monitoring method, it can provide a preliminary understanding of water quality. Combining this preliminary identification with detailed analysis of hyperspectral data can form a complementary monitoring system. If initial identification reveals no significant pollution, further hyperspectral data collection and analysis can provide a more accurate assessment of water quality, reducing the likelihood of false positives and missed reports.

[0167] To sum up, in the above scheme, adding the step of identifying suspended matter and oil pollution in the image data before receiving the first spectral image data returned by the drone has multiple technical effects such as reducing unnecessary data collection, improving monitoring efficiency, optimizing resource utilization, and improving monitoring accuracy, thereby jointly improving the overall performance and monitoring effectiveness of the system.

[0168] Some embodiments of the present application provide an intelligent monitoring method based on a drone, which specifically includes the following steps:

[0169] S601: The backend server obtains the pollution discharge standards of the factory to be monitored, the real-time pollution discharge data of the factory to be monitored, and the terrain data of the area to be monitored.

[0170] The pollutant discharge standards of the factory to be monitored include the required pollutant discharge time, the maximum pollutant discharge volume, and the water quality standard. The pollutant discharge standard is extracted from the pollutant discharge standard file by reading the pollutant discharge standard file.

[0171] Real-time sewage discharge data for the monitored factory includes discharge time, discharge volume, and sewage quality. Monitoring equipment is installed at the discharge outlet of the monitored factory. For example, a flow meter is installed to monitor the discharge time and volume. Another example is a water quality meter to monitor the sewage quality at the discharge outlet.

[0172] The topographic data of the area to be monitored includes the location of the factory to be monitored and the flow path of the river. The topographic data is obtained by collecting data on the area to be monitored through map surveying and mapping equipment.

[0173] S602: The backend server generates a flight mission according to the pollution discharge standard of the factory to be monitored, the real-time pollution discharge data of the factory to be monitored, and the terrain data of the area to be monitored.

[0174] The flight mission includes at least one flight path and the flight time of each flight path.

[0175] By analyzing and processing the real-time sewage discharge data of the monitored factories, we can determine whether the monitored factories are discharging sewage in accordance with regulations. For example, we can determine whether the monitored factories are discharging sewage within the required discharge time, whether the water quality of the monitored factories' sewage meets the water quality standards, and whether the sewage discharge volume of the monitored factories is less than the maximum discharge volume.

[0176] If the factory to be monitored is discharging pollutants in accordance with regulations, the number of flights in that area can be reduced when formulating flight missions. If the factory to be monitored is not discharging pollutants in accordance with regulations, the number of flights in that area can be increased when formulating flight missions.

[0177] In the above technical solution, by obtaining the pollution discharge standards of the factory to be monitored, the real-time pollution discharge data of the factory to be monitored, and the terrain data of the area to be monitored, a flight mission is generated based on the pollution discharge standards of the factory to be monitored, the real-time pollution discharge data of the factory to be monitored, and the terrain data of the area to be monitored. The mission can be executed with real-time reference to the pollution discharge situation of the factory to be monitored, and targeted monitoring can be carried out to improve monitoring efficiency and accuracy.

[0178] It is worth noting that the above solution has at least the following technical effects:

[0179] First, it improves the relevance and effectiveness of monitoring: By obtaining the pollution discharge standards and real-time pollution data of the monitored factories, as well as topographical data of the monitored areas, we can gain a comprehensive understanding of the factory's pollution discharge situation and the environmental characteristics of the monitored area. Flight missions generated based on this information can more accurately locate factories and areas where illegal emissions may occur, thereby improving the relevance and effectiveness of monitoring. This targeted monitoring strategy helps reduce ineffective monitoring and resource waste, and improves overall monitoring efficiency.

[0180] Second, optimize drone flight paths and task allocation: When generating flight missions, a comprehensive consideration of the pollutant emission standards, real-time pollution data, and terrain data of the monitored factories allows for more rational and efficient flight paths. This not only reduces drone flight distance and time, lowering energy consumption and costs, but also ensures that drones cover all key areas requiring monitoring. Furthermore, drone monitoring tasks are rationally allocated based on each factory's pollution emission profile and monitoring needs, avoiding duplication of effort and blind spots, thereby improving overall monitoring quality.

[0181] Third, improving the system's intelligence and automation: By introducing a data-driven mission generation mechanism, drone scheduling and flight path allocation are made more intelligent and automated. The system automatically adjusts flight missions and paths based on real-time pollution and terrain data, eliminating the need for manual intervention or frequent adjustments. This intelligent and automated scheduling approach not only improves the system's responsiveness and flexibility, but also reduces errors and risks caused by human error.

[0182] Fourth, the system's adaptability and scalability are enhanced: flight missions and paths can be flexibly adjusted to meet new monitoring needs. This adaptability and scalability ensures long-term stable operation and allows for continuous optimization and upgrades as the monitoring area expands and monitoring requirements increase. Furthermore, the system can be equipped with new functional modules or optimized algorithms based on actual needs, further enhancing system performance and monitoring effectiveness.

[0183] To sum up, in the above scheme, before scheduling drones based on flight missions and drones' flight capabilities, adding steps to obtain the pollution discharge standards, real-time pollution discharge data and terrain data of the factories to be monitored, and generating flight missions based on these data, has multiple technical effects such as improving the targetedness and effectiveness of monitoring, optimizing drone flight paths and task allocation, improving the intelligence and automation level of the system, and enhancing the adaptability and scalability of the system.

[0184] In one possible implementation, the real-time pollution data of the monitored factory includes historical pollution data and current pollution data. Both current pollution data and historical pollution data include the time of pollution discharge, the amount of pollution discharged, and the water quality of the wastewater. Accordingly, in step S602, the backend server generates a flight mission based on the pollution discharge standards of the monitored factory, the real-time pollution data of the monitored factory, and the terrain data of the monitored area. Specifically, the mission includes:

[0185] S71. For any factory to be monitored, the backend server determines whether the factory's pollutant discharge meets the discharge standard based on the factory's discharge standard and the current day's discharge data. If so, the process proceeds to S72. If not, the monitoring weight of the factory to be monitored is increased to the default value.

[0186] Among them, determining whether the pollutants discharged by the monitored factory meet the emission standards includes:

[0187] Determine whether the pollution discharge time of the monitored factory is within the required discharge time.

[0188] Determine whether the wastewater quality of the monitored factory meets the water quality requirements.

[0189] Determine whether the pollutant discharge of the monitored factory is less than or equal to the maximum pollutant discharge.

[0190] If any of the above items does not meet the requirements, it is determined that the pollutants discharged by the monitored factory do not meet the emission standards. If all three of the above items meet the requirements, it is determined that the pollutants discharged by the monitored factory meet the emission standards.

[0191] If the pollutants discharged by the monitored factory meet the emission standards, the system will further determine whether the monitored factory has the possibility of illegal discharge. If the pollutants discharged by the monitored factory do not meet the emission standards, the monitoring frequency weight of the monitored factory will be increased.

[0192] S72: If the pollutant discharge of the monitored factory meets the emission standards, the untrained pollutant discharge prediction model is trained using the historical pollutant discharge data to obtain a trained pollutant discharge prediction model. The trained pollutant discharge prediction model is used to obtain a predicted pollutant discharge amount for the day.

[0193] Among them, the historical pollution discharge data of each day in the past years are obtained in units of days, and the pollution discharge prediction model is trained with the historical pollution discharge data of each day in the past years. After receiving the date information, the trained pollution discharge prediction model can predict the pollution discharge volume of the day.

[0194] S73. Calculate the difference between the predicted pollutant discharge volume for the day and the actual pollutant discharge volume in the pollutant discharge data for the day, and determine whether the difference is greater than the preset difference threshold. If so, increase the monitoring times weight of the factory to be monitored; if not, maintain the monitoring times weight of the factory to be monitored at the default value.

[0195] If the difference between the current predicted and actual pollutant discharge is large, the plant under monitoring is likely to be illegally discharging pollutants, and monitoring of the plant under monitoring should be strengthened. If the difference between the current predicted and actual pollutant discharge is small, the risk of illegal discharge of pollutants by the plant under monitoring is small, and monitoring of the plant under monitoring can be reduced.

[0196] For any factory to be monitored, through step S71, step S72 and step S73, it is determined whether the monitoring times weight of the factory to be monitored needs to be updated, so that the monitoring times weight of each factory to be monitored can be obtained.

[0197] S74: Input the terrain data of the area to be monitored and the weight of the number of monitoring times of each factory to be monitored into the mission generation model to obtain the flight mission generated by the mission generation model.

[0198] Among them, the mission generation model can be obtained after training the machine learning model. The input parameters of the mission generation model include the terrain data of the area to be monitored and the weight of the monitoring times of each factory to be monitored. By giving the input parameters of the mission generation model, the mission generation model can output the flight mission.

[0199] If the weight of the monitoring times of a factory to be monitored is larger, the more times the factory to be monitored is monitored, and more flight paths include the factory to be monitored. If the weight of the monitoring times of a factory to be monitored is smaller, the fewer times the factory to be monitored is monitored, and fewer flight paths include the factory to be monitored.

[0200] For example, the monitored areas include Factory A, Factory B, and Factory C. Factory A has the highest weight for monitoring, while Factory B and Factory C have the lowest weight. A flight mission consists of two flight paths: the first path passes through Factory A and Factory B, and the second path passes through Factory A and Factory C.

[0201] In the above technical solution, it is first determined whether the pollution discharge data of a certain factory to be monitored on the day meets the pollution discharge standard. If not, the weight of the monitoring times of the factory to be monitored is increased. If the pollution discharge standard is met, the pollution discharge volume of the day is further predicted based on the historical pollution discharge data. It is determined whether the difference between the actual pollution discharge volume and the predicted volume on the day is large, and the weight of the monitoring times of the factory to be monitored is increased. On the contrary, if the difference is relatively small, the weight of the monitoring times of the factory to be monitored is maintained unchanged. By determining whether to maintain the monitoring times weight as the default value through two-level judgment conditions, more accurate monitoring can be achieved, thereby reducing the risk of the factory to be monitored controlling the pollution discharge data within the standard data through illegal discharge. Compared with increasing the monitoring times for all factories to be monitored, increasing the monitoring times weight of the factory to be monitored in a targeted manner can reduce the flight distance of the drone and the amount of data collected.

[0202] It is worth noting that the above solution has at least the following technical effects:

[0203] First, it improves the relevance and effectiveness of monitoring: The system first determines whether a plant's discharge meets regulatory requirements based on the plant's discharge standards and daily discharge data. For plants that don't meet these standards, their monitoring weight is immediately increased, ensuring these high-risk areas receive greater attention. This immediate response mechanism improves the relevance of monitoring, allowing resources to be quickly allocated to problem areas and effectively curbing the illegal discharge of wastewater.

[0204] Second, it optimizes monitoring resource allocation and improves monitoring efficiency: For factories that meet the day's pollution discharge regulations, the system does not immediately relax its vigilance. Instead, it further uses historical pollution discharge data to train a pollution discharge prediction model and predict the day's pollution discharge volume. By comparing the difference between predicted and actual pollution discharge volumes, the system can identify factories that are currently compliant but may pose potential risks. This data-based risk assessment method allows monitoring resources to be more rationally allocated to factories that truly require attention, avoiding resource waste and improving overall monitoring efficiency.

[0205] Third, the system's intelligence and automation are enhanced: The entire mission generation process relies heavily on data analysis and model predictions, eliminating the need for excessive human intervention. The system automatically collects, processes, and analyzes data, dynamically adjusts monitoring strategies based on the results, and generates corresponding missions. This intelligent and automated operation not only reduces errors caused by human error but also significantly improves the system's responsiveness and flexibility, making monitoring more efficient and accurate.

[0206] Fourth, improving data utilization efficiency and decision-making support capabilities: By fully utilizing both daily and historical pollution data, the system not only monitors the plant's pollution behavior in real time but also provides strong support for future monitoring through data analysis and model prediction. This data not only provides a basis for current monitoring tasks but also provides valuable information resources for decision-makers in formulating long-term environmental policies and regulatory measures. This data-driven decision-making support capability makes the system's monitoring work more scientific, rational, and effective.

[0207] To sum up, in the above scheme, flight missions are generated by comprehensively considering the pollution discharge standards of the factories to be monitored, real-time pollution discharge data, and terrain data of the areas to be monitored. This not only improves the pertinence and effectiveness of monitoring, optimizes the allocation of monitoring resources, but also enhances the intelligence and automation level of the system, improves data utilization efficiency and decision-making support capabilities, and thus provides a strong guarantee for effectively combating the illegal discharge of sewage.

[0208] In another possible implementation, S602, the backend server generates a flight mission based on the pollution discharge standards of the factory to be monitored, the real-time pollution discharge data of the factory to be monitored, and the terrain data of the area to be monitored, specifically including:

[0209] S81. For each factory to be monitored, set a discharge time coefficient according to the discharge time of the factory to be monitored and the required discharge time.

[0210] More specifically, it is determined whether the pollutant discharge time of the monitored plant is within the required discharge time. If so, the discharge time coefficient is set to a first value; if not, the discharge time coefficient is set to a second value, wherein the first value is smaller than the second value.

[0211] S82. For each factory to be monitored, set a discharge water quality coefficient according to the discharge water quality and water quality requirements of the factory to be monitored.

[0212] More specifically, it is determined whether the wastewater quality of the monitored factory meets the water quality standard. If so, the wastewater quality coefficient is set to a third value; if not, the wastewater quality coefficient is set to a fourth value. The third value is less than the fourth value.

[0213] S83. For each factory to be monitored, set a pollutant discharge coefficient according to the pollutant discharge volume and the maximum pollutant discharge volume of the factory to be monitored.

[0214] More specifically, it is determined whether the pollutant discharge of the monitored factory is less than or equal to the maximum pollutant discharge. If so, the pollutant discharge coefficient is set to the fifth value; if not, the pollutant discharge coefficient is set to the sixth value.

[0215] The fifth value is smaller than the sixth value.

[0216] S84. For each factory to be monitored, use the trained pollution emission prediction model to obtain the predicted pollution emission for the day, and set the emission prediction coefficient based on the predicted pollution emission for the day and the actual pollution emission in the pollution emission data for the day.

[0217] More specifically, the difference between the predicted pollutant discharge for the day and the actual pollutant discharge data for the day is calculated, and a determination is made as to whether the difference is greater than a preset difference threshold. If so, the emission prediction coefficient is set to the seventh value; otherwise, the emission prediction coefficient is set to the eighth value. The seventh value is less than the eighth value.

[0218] S85. For each factory to be monitored, generate a monitoring times weight for the factory to be monitored according to the discharge time coefficient, the discharge water quality coefficient, the discharge volume coefficient, and the discharge prediction coefficient.

[0219] More specifically, the weighted average of the discharge time coefficient, the discharge water quality coefficient, the discharge volume coefficient and the discharge prediction coefficient is performed to calculate the monitoring times weight of the factory to be monitored.

[0220] For any factory to be monitored, the monitoring times weight of the factory to be monitored is calculated through steps S81 to S85, so that the monitoring times weight of each factory to be monitored can be obtained.

[0221] S86: Input the terrain data of the area to be monitored and the weight of the number of monitoring times of each factory to be monitored into the mission generation model to obtain the flight mission generated by the mission generation model.

[0222] This step has been described in detail in the above embodiment and will not be repeated here.

[0223] In the above technical solution, the discharge time system, discharge water quality coefficient, discharge volume coefficient and discharge prediction coefficient are calculated respectively according to the discharge time, discharge water quality, discharge volume and predicted discharge volume of the factory to be monitored, and the monitoring frequency weight of the factory to be monitored is calculated based on the above four coefficients. Compared with increasing the monitoring frequency for all factories to be monitored, increasing the monitoring frequency weight of the factory to be monitored in a targeted manner can reduce the flight distance of the drone and reduce the amount of data collection.

[0224] It is worth noting that the above solution has at least the following technical effects:

[0225] First, it improves the refinement and personalization of monitoring: For each monitored plant, the system assigns coefficients (discharge time coefficient, discharge water quality coefficient, and discharge volume coefficient) based on multiple dimensions, including discharge time, discharge volume, and discharge quality. These coefficients reflect key aspects of a plant's discharge behavior. By comprehensively considering these coefficients, the system can more comprehensively assess a plant's discharge situation and generate personalized monitoring tasks based on the assessment results. This refined and personalized monitoring strategy enables the system to more effectively identify potential illegal discharges.

[0226] Second, enhancing predictive capabilities to proactively address potential risks: The system leverages a trained pollution prediction model, combined with the plant's real-time and historical data, to predict the day's discharge volume and set emission prediction coefficients accordingly. This step enables the system to proactively identify plants that are currently compliant but may pose potential risks. By increasing the weight of monitoring these plants, the system can proactively address potential risks and prevent the illegal discharge of wastewater.

[0227] Third, optimized mission generation improves monitoring efficiency: When generating missions, the system comprehensively considers the terrain data of the monitored area and the weighting of monitoring times for each factory. Terrain data provides a crucial reference for drone flight path planning, while monitoring weighting ensures that high-risk areas receive more attention. By optimizing mission generation, the system ensures that drones cover as many high-risk areas as possible within a limited timeframe, thereby improving overall monitoring efficiency.

[0228] Fourth, it improves data utilization efficiency and decision-making support capabilities: By analyzing real-time pollution data from multiple dimensions, the system not only improves the sophistication and personalization of monitoring, but also provides decision makers with richer and more accurate information support. Based on the system's multi-dimensional pollution data and the performance of monitoring tasks, decision makers can formulate more scientific and reasonable environmental protection policies and regulatory measures. This data-driven decision-making support capability makes the system's monitoring work more responsive to actual needs and more effective.

[0229] In summary, the above solution, through multi-dimensional analysis of real-time sewage discharge data and the generation of flight missions based on this data, not only improves the refinement and personalization of monitoring, enhances the system's predictive capabilities, but also optimizes the flight mission generation process and improves monitoring efficiency. This strategy also enhances data utilization efficiency and decision-making support capabilities, providing a strong guarantee for effectively combating illegal sewage discharge.

[0230] The backend server provided in this embodiment includes: at least one processor and a memory. Optionally, the device also includes a communication component. The processor, memory, and communication component are connected via a bus.

[0231] In a specific implementation process, at least one processor executes computer-executable instructions stored in a memory, so that the at least one processor performs the above method.

[0232] The specific implementation process of the processor can be found in the above method embodiment. Its implementation principle and technical effects are similar and will not be repeated here in this embodiment.

[0233] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASICs), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the present invention may be directly executed by a hardware processor or by a combination of hardware and software modules within the processor.

[0234] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage.

[0235] A bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be categorized as address buses, data buses, and control buses. For ease of illustration, the buses in the drawings of this application are not limited to just one bus or just one type of bus.

[0236] The present application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0237] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above method is implemented.

[0238] The readable storage medium may be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium may be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0239] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be an integral part of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0240] The division of units is merely a logical functional division; actual implementations may employ alternative divisions, such as combining or integrating multiple units or components into another system, or omitting or disabling certain features. Furthermore, any direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between devices or units, either through an interface, electrical, mechanical, or other means.

[0241] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0242] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0243] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0244] Those skilled in the art will appreciate that all or part of the steps in the above-described method embodiments can be implemented using hardware associated with program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0245] Finally, it should be noted that those skilled in the art will readily identify other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art not disclosed herein. The present invention is not limited to the precise structure described above and illustrated in the accompanying drawings, and various modifications and variations may be made without departing from the scope thereof. The scope of the present invention is limited solely by the appended claims.

Claims

1. An intelligent monitoring method based on drones, characterized in that: Applied to a backend server, the method includes: Obtaining the pollution discharge standards of the factories to be monitored, the real-time pollution discharge data of the factories to be monitored, and the topographic data of the areas to be monitored; the real-time pollution discharge data includes the current pollution discharge data and historical pollution discharge data, and the current pollution discharge data and historical pollution discharge data both include the pollution discharge time, actual pollution discharge volume, and sewage water quality; Generate a flight mission based on the pollution discharge standards of the factory to be monitored, the real-time pollution discharge data of the factory to be monitored, and the terrain data of the area to be monitored; specifically, the flight mission includes: For any factory to be monitored, determine whether the pollution discharge of the factory to be monitored meets the regulatory requirements based on the pollution discharge standards of the factory to be monitored and the pollution discharge data of the day; if not, increase the monitoring frequency weight of the factory to be monitored to the default value; If the pollution discharge of the monitored factory meets the regulatory requirements, the untrained pollution discharge prediction model is trained with the historical pollution discharge data to obtain a trained pollution discharge prediction model; the trained pollution discharge prediction model is used to predict the predicted pollution discharge volume for the day; Calculate the difference between the predicted pollutant discharge volume for the day and the actual pollutant discharge volume in the pollutant discharge data for the day, and determine whether the difference is greater than a preset difference threshold. If so, increase the monitoring times weight of the plant to be monitored; if not, maintain the monitoring times weight of the plant to be monitored at a default value. Inputting the terrain data of the area to be monitored and the weight of the number of monitoring times of each factory to be monitored into the task generation model to obtain a flight task generated by the task generation model; The UAV is scheduled based on the flight mission and the flight capability of the UAV, and a corresponding flight path is assigned to the UAV; the UAV is caused to fly along the flight path during the flight time, and the onboard hyperspectral detector is controlled to collect first spectral image data; the flight mission includes at least one flight path and the flight time of the flight path; receiving the first spectral image data returned by the UAV, and preprocessing the first spectral image data to obtain second spectral image data; For the current pixel point in each row of pixels in the second spectral image data, determining whether the spectral data of the previous pixel point and the spectral data of the current pixel point meet a replacement condition; if so, replacing the spectral data of the current pixel point with the spectral data of the previous pixel point; if not, maintaining the spectral data of the current pixel point unchanged, traversing all pixels in each row of pixels in the second spectral image data to obtain third spectral image data; Extracting features from the third spectral image data, and classifying the extracted features to obtain water quality parameters of the water area corresponding to the third spectral image; It is determined whether the sewage discharged by the monitored factory meets the water quality standard according to the water quality parameters of the water area corresponding to the third spectral image.

2. The intelligent monitoring method according to claim 1, characterized in that: Replacing the spectral data of the current pixel point with the spectral data of the previous pixel point specifically includes: Obtaining a spectral data identifier associated with the previous pixel, and establishing an association relationship between the current pixel and the spectral data identifier associated with the previous pixel; wherein a mapping relationship exists between the spectral data identifier and common spectral data, and the common spectral data is stored in a common storage area; Accordingly, the spectral data of the current pixel is maintained unchanged, specifically including: A new spectral data identifier is randomly generated, the spectral data of the current pixel is used as new shared spectral data, the new shared spectral data is stored in a shared storage area, and an association relationship is established between the current pixel and the new spectral data identifier.

3. The intelligent monitoring method according to claim 2, characterized in that: Extracting features from the third spectral image data and classifying the extracted features to obtain water quality parameters of the water area corresponding to the third spectral image specifically includes: Extracting features from each spectral data in the shared storage area, and classifying the extracted features to obtain water quality parameters corresponding to each spectral data; The water quality parameters of the water area corresponding to each pixel point are obtained according to the spectral data identifier associated with each pixel point and the water quality parameters of the spectral data associated with each spectral data identifier.

4. The intelligent monitoring method according to any one of claims 1 to 3, characterized in that: The spectral data includes light intensities in multiple bands. The determination of whether the spectral data of the previous pixel and the spectral data of the current pixel meet the replacement condition specifically includes: Calculate the light intensity difference between the light intensity of each band of the previous pixel and the light intensity of the corresponding band of the current pixel to obtain the light intensity difference of multiple bands; Calculate the average value of the light intensity difference of all bands, and calculate the variance of the light intensity difference of all bands; If the average value of the light intensity difference is less than a preset average value threshold, and the variance of the light intensity difference is less than a preset variance threshold, it is determined that the replacement condition is met; If the average value of the light intensity difference is greater than or equal to a preset average value threshold, or the variance of the light intensity difference is greater than or equal to a preset variance threshold, it is determined that the replacement condition is not met.

5. The intelligent monitoring method according to any one of claims 1 to 3, characterized in that: Preprocessing the first spectral image data to obtain second spectral image data specifically includes: Denoising the first spectral image data to obtain denoised spectral image data; performing baseline correction on the denoised spectral image data to obtain baseline-corrected spectral image data; The baseline-corrected spectral image data is subjected to black-and-white correction to obtain the third spectral image data.

6. The intelligent monitoring method according to claim 1, characterized in that: The real-time sewage discharge data includes the current day's sewage discharge data and historical sewage discharge data, and the sewage discharge data includes the sewage discharge time, sewage discharge volume and sewage quality; Accordingly, generating the flight mission according to the pollution discharge standard of the factory to be monitored, the real-time pollution discharge data of the factory to be monitored, and the terrain data of the area to be monitored specifically includes: For each factory to be monitored, the discharge time coefficient is set according to the discharge time of the factory to be monitored and the required discharge time; the discharge water quality coefficient is set according to the discharge water quality and water quality requirements of the factory to be monitored; the discharge volume coefficient is set according to the discharge volume and maximum discharge volume of the factory to be monitored; For each factory to be monitored, the trained emission prediction model is used to obtain the predicted emission volume for the day; and the emission prediction coefficient is set based on the predicted emission volume for the day and the actual emission volume in the emission data for the day; For each factory to be monitored, generating a monitoring times weight for each factory to be monitored according to the discharge time coefficient, the discharge water quality coefficient, the discharge volume coefficient and the discharge prediction coefficient; The terrain data of the area to be monitored and the weight of the number of monitoring times of each factory to be monitored are input into the task generation model to obtain the flight task generated by the task generation model.

7. An intelligent monitoring system based on drones, characterized in that: The method comprises a drone and a background server, wherein the drone is equipped with a hyperspectral detector and an image sensor, and the background server is used to execute the intelligent monitoring method according to any one of claims 1 to 6.

8. An electronic device, characterized in that: include: processor; as well as, a memory for storing executable instructions of the processor; Wherein, the processor is configured to execute the intelligent monitoring method according to any one of claims 1 to 6 by executing the executable instructions.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the intelligent monitoring method according to any one of claims 1 to 6.

Images