Water surface garbage monitoring method based on drone

Abstract

The present invention discloses a method for monitoring water surface garbage based on an unmanned aerial vehicle, which relates to the technical field of water surface garbage monitoring, and includes the following steps: the unmanned aerial vehicle route planning is based on water area information, and is generated by integrating multiple factors with an intelligent algorithm, setting parameters and optimizing them, and image acquisition relies on high-definition equipment, high-pixel photosensitivity, and real-time data transmission. Image preprocessing is performed at the ground center for denoising and correction, garbage identification and positioning are performed with the help of deep learning algorithms and coordinate conversion models, data storage and analysis are performed using distributed libraries for storage and multi-dimensional analysis, early warning and feedback exceed threshold early warning and there are cleaning suggestions and feedback mechanisms, and each step works closely together to achieve efficient and accurate water surface garbage monitoring, and improve the level of water surface garbage monitoring and management. The present invention achieves efficient and accurate, full-time wide-area monitoring of water surface garbage, provides accurate data support and scientific decision-making basis for management, can timely warn and promote collaborative management, effectively improve monitoring accuracy and cleaning efficiency, and ensure the quality of water area environment.

Description

Technical Field

[0001] The invention relates to the technical field of water surface garbage monitoring, and in particular to a water surface garbage monitoring method based on an unmanned aerial vehicle. Background Art

[0002] With the continuous expansion of human activities and the rapid development of the economy, the problem of water surface garbage pollution is becoming increasingly serious, which has had a great negative impact on aquatic ecosystems, water resource utilization, and water landscapes. Traditional methods for monitoring water surface garbage mainly rely on manual inspections and monitoring equipment at fixed monitoring points, but these methods have many limitations.

[0003] Manual inspections are inefficient and cost a lot of manpower, material resources and time. Monitoring personnel need to take a boat or conduct visual inspections on the shore, which not only limits the monitoring range, but also makes it difficult to conduct comprehensive and timely monitoring of large areas of water. For example, for large waters such as large lakes, river estuaries or coastal waters, manual inspections often take a long time to complete a monitoring task, resulting in garbage accumulation not being discovered and handled in a timely manner.

[0004] Although the monitoring equipment at fixed monitoring points can achieve a certain degree of automated monitoring, there are monitoring blind spots. These devices are usually installed in specific locations and can only monitor local areas around them. They cannot cover the entire water area and are prone to miss some spam information in remote or hidden areas. Moreover, fixed equipment is greatly affected by environmental factors. For example, in severe weather conditions (heavy rain, strong winds, dense fog, etc.), the monitoring effect will be greatly reduced or even fail to work properly.

[0005] In addition, traditional monitoring methods have difficulty obtaining high-resolution images of garbage on the water surface, and it is difficult to accurately determine detailed information such as the type, distribution density, and area size of the garbage. This results in a lack of accurate data support when formulating garbage cleaning plans and evaluating cleaning results, making it impossible to achieve targeted and efficient water surface garbage management.

[0006] With the rapid development of drone technology, it has the advantages of strong mobility, high flexibility, and the ability to carry a variety of monitoring equipment, providing a new solution for surface garbage monitoring. Drones can be used to quickly and efficiently monitor large areas of water in all directions, overcoming many of the shortcomings of traditional monitoring methods and obtaining richer and more accurate information on surface garbage, thus providing a strong data basis and technical guarantee for the effective management of surface garbage. Summary of the invention

[0007] The present invention proposes a water surface garbage monitoring method based on an unmanned aerial vehicle to solve the problems mentioned in the above-mentioned prior art.

[0008] In order to achieve the above object, the present invention adopts the following technical solution: a method for monitoring water surface garbage based on an unmanned aerial vehicle, comprising the following steps:

[0009] S1. Steps of drone route planning

[0010] According to the shape, size and surrounding environment information of the water surface area to be monitored, the intelligent algorithm is used to automatically generate the flight route of the drone. The intelligent algorithm adopts a path planning model that integrates multiple constraints and sets the wind direction influencing factor as W. f , the wind speed influence factor is W s , the no-fly zone impact factor is W n , the obstacle influence factor is W o , and its calculation formula is: Where P is the path planning rationality parameter;

[0011] Set the parameters of the drone's flight altitude H, 5≤H≤100 meters, speed 5≤V≤20 meters / second, and image acquisition interval 1≤H≤10 seconds;

[0012] S2. Image acquisition steps: The high-definition image acquisition device mounted on the drone uses a high-resolution optical lens and a high-sensitivity image sensor. The sensitivity of the sensor is automatically adjusted between ISO50 and ISO6400. The image acquisition device uses advanced photoelectric conversion technology. The calculation formula for its photoelectric conversion efficiency η is: Where N e is the effective number of electrons, N p is the number of incident photons;

[0013] When the UAV flies along the scheduled route, the image acquisition device automatically takes images of the water surface at set intervals and transmits the image data to the ground control center in real time. The data transmission adopts high-speed wireless transmission technology.

[0014] S3. Image preprocessing steps: After receiving the image data, the ground control center first performs denoising on the image. It uses an adaptive filtering algorithm to dynamically adjust the filtering parameters according to the local features of the image, remove salt and pepper noise and Gaussian noise in the image, and retain the edge and detail information of the image. The filter coefficient update formula of the adaptive filtering algorithm is: C n+1 =C n +μ×e n × n , where C n+1 is the next filtering coefficient, C n is the current filter coefficient, μ is the learning rate, e n is the error signal, x n is the input signal;

[0015] The denoised image is geometrically corrected. Based on the flight attitude data of the drone, including the heading angle α, pitch angle β, roll angle γ, and camera parameters, including focal length f and pixel size d, the image is corrected using the perspective transformation algorithm. The calculation of the perspective transformation matrix M is based on the following formula:

[0016]

[0017] Where m ij The value of is related to the parameters α, β, γ, f, and d, and is derived through complex geometric derivation;

[0018] S4, garbage identification and location step: Use deep learning-based image recognition algorithm to analyze the preprocessed images, and construct a neural network model with a training sample number of no less than 100,000, covering various types of water surface garbage images in different environments. Through model training, water surface garbage can be identified. The deep learning image recognition algorithm adopts a convolutional neural network and a recurrent neural network architecture. The convolutional neural network is used to extract the spatial features of the image, and the recurrent neural network is used to process the image sequence information.

[0019] For the identified garbage on the water surface, the image coordinate and geographic coordinate conversion model is used, combined with the real-time location information of the drone and the camera shooting angle, to calculate the geographical location of the garbage on the water surface, and record the area and shape information of the garbage. Suppose the pixel coordinates of the garbage in the image are (x p ,y p ), the latitude and longitude coordinates of the drone are (λ d , ), the pitch angle of the camera is θ, and the heading angle is Then the geographical coordinates of the garbage (λ g , ) is calculated as:

[0020]

[0021] Where R is the radius of the earth, and accurate positioning is achieved through this formula;

[0022] S5, data storage and analysis step: storing the identification results, geographical location, area, and shape data of the garbage in a database, which adopts a distributed storage architecture and has a data redundancy backup function;

[0023] Conduct statistical analysis on the stored data, including analysis of garbage distribution density, analysis of garbage type proportion, and analysis of garbage volume change trends over time, to provide data support and decision-making basis for water surface garbage management;

[0024] S6, early warning and feedback step: According to the preset garbage threshold parameters, when the garbage distribution density or garbage volume in the monitoring area exceeds the set threshold, the system automatically generates early warning information, which includes the location of the early warning area and the description of the garbage exceeding the standard, and is promptly pushed to relevant management departments and staff through SMS, email, system pop-up windows and other methods;

[0025] A feedback mechanism will be established. After receiving early warning information, relevant departments will feedback the implementation status and cleaning effect information of the cleanup action to the monitoring system. The system will correct and improve the monitoring data and analysis results based on the feedback information.

[0026] Furthermore, in the UAV route planning step, the intelligent algorithm uses a combination of genetic algorithm and simulated annealing algorithm. The genetic algorithm is first used for global search, and then the simulated annealing algorithm is used to locally optimize the search results. The crossover probability P in the genetic algorithm is c Set to 0.6-0.9, the mutation probability P m It is set to 0.01-0.1, the initial temperature T0 in the simulated annealing algorithm is set according to the problem scale, and the temperature reduction coefficient α is set to 0.9-0.99.

[0027] Furthermore, in the image acquisition step, the high-definition image acquisition device is also equipped with an infrared thermal imaging function to assist in identifying surface garbage at night or under low light conditions. The infrared thermal imaging device uses microbolometer focal plane array technology, and the response rate R of its detector is calculated as: R = V s / P, where V s is the detector output voltage, and P is the incident radiation power.

[0028] Furthermore, in the image preprocessing step, after geometric correction, the image is also subjected to contrast enhancement. The histogram equalization algorithm is used. Assume that the original gray level of the image is r k , the gray level after transformation is s k , the total number of pixels in the image is n, and the gray level is r k The number of pixels is n k , then the transformation function of histogram equalization is: This function is used to achieve contrast enhancement.

[0029] Furthermore, in the garbage identification and location step, the convolutional neural network used by the deep learning image recognition algorithm has a convolution kernel size of 3×3 or 5×5, a step size of 1 or 2, and a filling method of SAME or VALID. The feature extraction effect is optimized by adjusting these parameters.

[0030] Furthermore, in the data storage and analysis steps, the database supports remote access and sharing of data. Relevant scientific research institutions or environmental protection organizations can remotely obtain monitoring data within the authorized scope for conducting research and cooperation projects on water surface garbage management. Data sharing security uses encrypted transmission and identity authentication technology. The encryption algorithm uses the AES-256-bit encryption algorithm, and identity authentication uses digital certificate technology.

[0031] Furthermore, in the early warning and feedback step, the early warning information also includes cleaning route suggestions generated based on the garbage distribution. The cleaning route suggestions are generated using the shortest path algorithm based on the geographical distribution of the garbage, surrounding traffic, and the distribution of cleaning facilities. The shortest path algorithm uses the Dijkstra algorithm or the A* algorithm.

[0032] Furthermore, during the entire monitoring process, the drone's energy management system monitors the battery power B in real time. When B is lower than 20%, it automatically starts the low-battery return mode and flies back to the charging point according to the pre-planned shortest return path. The shortest return path planning adopts a path planning algorithm based on energy consumption optimization. This algorithm takes into account the distance factor while taking into account the energy consumption of the drone during flight, including energy consumption compensation when flying against the wind.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The intelligent algorithm is used to plan the drone route, which can adapt to various complex water environments, optimize the flight path, and improve monitoring efficiency. High-definition image acquisition equipment and advanced image preprocessing and recognition technology can accurately identify surface garbage with small positioning errors. It can accurately obtain various types of garbage information, including geographical location, area, shape, etc., to provide accurate target data for garbage cleaning, and the accuracy of monitoring data is improved by more than 30% compared with traditional methods.

[0035] The infrared thermal imaging function of the image acquisition equipment enables it to work normally at night or in low light conditions, breaking through time constraints and achieving full-time monitoring. The maneuverability of the drone enables it to easily cover a wide area of ​​water, avoiding the blind spots of fixed monitoring points, and the monitoring range can be expanded to several times that of traditional methods, greatly improving the understanding of the distribution of garbage on the water surface.

[0036] The system has powerful functions for storing and analyzing monitoring data, and can deeply analyze the garbage distribution density, type ratio, and trend over time, etc. These detailed data and intuitive analysis reports provide a scientific basis for decision-making on water surface garbage management, and help relevant departments to rationally allocate cleaning resources and formulate efficient cleaning plans, such as determining key cleaning areas based on garbage distribution density, which can increase cleaning efficiency by about 50%.

[0037] By presetting garbage thresholds and generating early warning information in a timely manner, relevant departments can be quickly notified to take cleanup actions to minimize the harm of garbage pollution. The feedback mechanism promotes the coordination of monitoring and governance, and can continuously optimize monitoring and governance strategies based on cleanup feedback, improve overall governance effects, and ensure continuous improvement of the water environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a schematic block diagram of a water surface garbage monitoring method based on an unmanned aerial vehicle proposed in the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0041] In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined. In addition, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, and it can be the internal connection of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0042] Reference Figure 1 :A method for monitoring water surface garbage based on drones, comprising the following steps:

[0043] The steps of UAV route planning are as follows: According to the shape, size and surrounding environment information of the water surface area to be monitored, the UAV flight route is automatically generated using an intelligent algorithm. The intelligent algorithm adopts a path planning model that integrates multiple constraints, assuming that the wind direction influencing factor is W. f , the wind speed influence factor is W s , the no-fly zone impact factor is W n , the obstacle influence factor is W o , and its calculation formula is: Among them, P is the rationality parameter of path planning. Through this model, various factors are comprehensively considered to ensure the scientificity and safety of route planning, so that the UAV can effectively avoid obstacles during flight, while minimizing energy consumption, and the flight path optimization rate is not less than 20%.

[0044] The parameters of the UAV flight altitude H, 5≤H≤100 m, speed 5≤V≤20 m / s and image acquisition interval 1≤H≤10 s were set to ensure the acquisition of clear and comprehensive water surface image information.

[0045] Image acquisition steps: The high-definition image acquisition equipment mounted on the drone uses a high-resolution optical lens and a high-sensitivity image sensor. The lens resolution is not less than 40 million pixels. The sensor's sensitivity can be automatically adjusted between ISO50 and ISO6400, which can adapt to image acquisition tasks under different lighting conditions. The color restoration of the collected images is not less than 90%, ensuring that the color, shape and other characteristics of the surface garbage are accurately presented. The image acquisition equipment uses advanced photoelectric conversion technology, and its photoelectric conversion efficiency η is calculated as follows: Where N e is the effective number of electrons, N p The number of incident photons is used to improve the image acquisition quality by optimizing the photoelectric conversion efficiency.

[0046] When the UAV is flying along the scheduled route, the image acquisition equipment automatically takes images of the water surface at set intervals and transmits the image data to the ground control center in real time. The data transmission uses high-speed wireless transmission technology with a transmission rate of no less than 50Mbps and a transmission delay of no more than 0.5 seconds, ensuring the timeliness and integrity of the image data.

[0047] Image preprocessing steps: After the ground control center receives the image data, it first performs denoising on the image, using an adaptive filtering algorithm to dynamically adjust the filtering parameters according to the local features of the image, effectively removing salt and pepper noise, Gaussian noise, etc. in the image, with a noise removal rate of no less than 95%, while retaining the edge and detail information of the image, with an edge information retention rate of no less than 90%. The filter coefficient update formula of the adaptive filtering algorithm is: C n+1 =C n +μ×en × n , where C n+1 is the next filtering coefficient, C n is the current filter coefficient, μ is the learning rate, e n is the error signal, x n For the input signal, adaptive denoising is achieved by continuously updating the filter coefficients.

[0048] The denoised image is geometrically corrected. Based on the UAV's flight attitude data (including heading angle α, pitch angle β, roll angle γ, etc.) and camera parameters (focal length f, pixel size d, etc.), the image is corrected using the perspective transformation algorithm. The correction error does not exceed 1 pixel, so that the image can accurately reflect the actual geographic spatial information of the water surface. The calculation of the perspective transformation matrix M is based on the following formula:

[0049]

[0050] Where m ij The value of is related to the parameters α, β, γ, f, and d and is derived through complex geometry to achieve accurate correction.

[0051] Garbage identification and location steps: Use a deep learning-based image recognition algorithm to analyze the preprocessed images, and build a neural network model with no less than 100,000 training samples, covering various types of surface garbage images in different environments. Through model training, it can accurately identify surface garbage with an accuracy rate of no less than 90%. The deep learning image recognition algorithm uses a convolutional neural network and a recurrent neural network architecture. The convolutional neural network is used to extract the spatial features of the image, and the recurrent neural network is used to process image sequence information to improve the ability to identify surface garbage of different shapes, sizes and distribution states. The model's recognition accuracy for complex garbage scenes is increased by no less than 10%.

[0052] For the identified garbage on the water surface, the image coordinate and geographic coordinate conversion model is used, combined with the real-time location information of the drone (obtained through the GPS / Beidou positioning system, with a positioning accuracy of 1 to 5 meters) and the camera shooting angle, to calculate the precise geographic location of the garbage on the water surface, with a positioning error of no more than 3 meters. At the same time, the area, shape and other related information of the garbage are recorded. Assume that the pixel coordinates of the garbage in the image are (x p ,y p ), the latitude and longitude coordinates of the drone are (λ d , ), the pitch angle of the camera is θ, and the heading angle is Then the geographical coordinates of the garbage (λ g , ) is calculated as:

[0053]

[0054] Where R is the radius of the earth, and accurate positioning is achieved through this formula;

[0055] Data storage and analysis steps: The identification results, geographic location, area, shape and other data of the garbage are stored in the database. The database adopts a distributed storage architecture and has data redundancy backup function. The data storage reliability is not less than 99.9%, and it can store at least 1 year of monitoring data history to facilitate subsequent query and analysis.

[0056] Statistical analysis is performed on the stored data, including garbage distribution density analysis (measured by the amount of garbage per unit area, with an analysis accuracy of no less than 95%), garbage type proportion analysis (capable of accurately distinguishing common garbage types such as plastic, metal, and wood, with a proportion analysis error of no more than 5%), and analysis of garbage volume change trends over time, etc., to provide data support and decision-making basis for water surface garbage management. The visualization of the analysis results is clear and intuitive, and it takes no more than 10 minutes to generate an analysis report.

[0057] Early warning and feedback steps: According to the preset garbage threshold parameters, when the garbage distribution density or garbage volume in the monitoring area exceeds the set threshold, the system automatically generates early warning information. The early warning information includes the location of the warning area, description of the garbage exceeding the standard, etc. It is promptly pushed to relevant management departments and staff through SMS, email, system pop-up windows and other methods. The push time does not exceed 1 minute to ensure that cleaning and management measures are taken in time.

[0058] A feedback mechanism will be established. After receiving early warning information, relevant departments can feed back information such as the implementation of cleanup actions and cleanup effects to the monitoring system. The system will correct and improve the monitoring data and analysis results based on the feedback information, continuously improve the coordination and effectiveness of monitoring and governance, and the accuracy of feedback information processing will be no less than 98%.

[0059] In the present invention, in the UAV route planning step, the intelligent algorithm adopts a combination of genetic algorithm and simulated annealing algorithm. The genetic algorithm is first used for global search, and then the search results are locally optimized by simulated annealing algorithm to further improve the rationality and efficiency of route planning. The total flight distance reduction rate after algorithm optimization is not less than 15%. The crossover probability P in the genetic algorithm c Set to 0.6-0.9, the mutation probability P m It is set to 0.01-0.1, the initial temperature T0 in the simulated annealing algorithm is set according to the problem scale, and the cooling coefficient α is set to 0.9-0.99. The algorithm performance can be improved by reasonably setting these parameters.

[0060] In the present invention, in the image acquisition step, the high-definition image acquisition device is also equipped with an infrared thermal imaging function, which can assist in identifying surface garbage at night or under low light conditions. The thermal imaging resolution is not less than 320×240 pixels, and the temperature resolution is not less than 0.1°C. It can effectively distinguish the temperature difference between garbage and the water surface background, and improve the accuracy and reliability of garbage identification. The infrared thermal imaging device adopts microbolometer focal plane array technology, and the response rate R of its detector is calculated as: R=V s / P, where V s is the detector output voltage, P is the incident radiation power, and the thermal imaging effect is enhanced by increasing the response rate.

[0061] In the present invention, in the image preprocessing step, after geometric correction, the image is also subjected to contrast enhancement processing, and a histogram equalization algorithm is used to increase the contrast between the garbage and the background in the image by no less than 50%, which is more conducive to the subsequent garbage identification and analysis operations. Assume that the original grayscale level of the image is r k , the gray level after transformation is s k , the total number of pixels in the image is n, and the gray level is r k The number of pixels is n k , then the transformation function of histogram equalization is: This function is used to achieve contrast enhancement.

[0062] In the present invention, in the garbage identification and positioning step, the convolution kernel size of the convolutional neural network used by the deep learning image recognition algorithm is 3×3 or 5×5, the step size is 1 or 2, and the filling method is SAME or VALID. The feature extraction effect is optimized by adjusting these parameters.

[0063] In the present invention, in the data storage and analysis steps, the database supports remote access and sharing functions of data. Relevant scientific research institutions or environmental protection organizations can remotely obtain monitoring data within the scope of authorization for conducting research and cooperation projects on water surface garbage management. Data sharing security adopts encrypted transmission and identity authentication technology to ensure that data is not illegally accessed and tampered with. The encryption algorithm adopts AES-256-bit encryption algorithm, and identity authentication adopts digital certificate technology. The authentication success rate is not less than 99%.

[0064] In the present invention, in the early warning and feedback step, the early warning information also includes a cleaning route suggestion generated based on the garbage distribution. The cleaning route suggestion is generated using the shortest path algorithm based on the geographical distribution of the garbage, the surrounding traffic, and the distribution of cleaning facilities. It can effectively guide the cleaning operation and improve the cleaning efficiency. The cleaning route optimization rate is not less than 25%. The shortest path algorithm uses the Dijkstra algorithm or the A* algorithm to achieve optimal route planning by constructing a suitable graph structure and weight setting.

[0065] In the present invention, during the entire monitoring process, the energy management system of the drone monitors the battery power B in real time. When B is lower than 20%, the low-battery return mode is automatically started, and the drone flies back to the charging point according to the pre-planned shortest return path. The shortest return path planning adopts a path planning algorithm based on energy consumption optimization. While considering the distance factor, the algorithm also takes into account the energy consumption of the drone during flight, such as energy consumption compensation when flying against the wind, etc., to ensure the safe return of the drone and reduce energy waste, and the return success rate is not less than 95%.

[0066] The above are only preferred specific implementation modes of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solutions and inventive concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for monitoring water surface garbage based on drones, characterized in that: The following steps are involved: S1. Steps of drone route planning According to the shape, size and surrounding environment information of the water surface area to be monitored, the intelligent algorithm is used to automatically generate the flight route of the drone. The intelligent algorithm adopts a path planning model that integrates multiple constraints and sets the wind direction influencing factor as W. f , the wind speed influence factor is W s , the no-fly zone impact factor is W n , the obstacle influence factor is W o , and its calculation formula is: Where P is the path planning rationality parameter; Set the flight altitude H of the drone, 5≤H≤100 meters, speed 5≤V≤20 meters / second, and image acquisition interval 1≤H≤10 seconds; S2. Image acquisition steps: The high-definition image acquisition device mounted on the drone uses a high-resolution optical lens and a high-sensitivity image sensor. The sensitivity of the sensor is automatically adjusted between ISO50 and ISO6400. The image acquisition device uses advanced photoelectric conversion technology. The calculation formula for its photoelectric conversion efficiency η is: Where N e is the effective number of electrons, N p is the number of incident photons; When the UAV flies along the scheduled route, the image acquisition device automatically takes images of the water surface at set intervals and transmits the image data to the ground control center in real time. The data transmission adopts high-speed wireless transmission technology. S3. Image preprocessing steps: After receiving the image data, the ground control center first performs denoising on the image. It uses an adaptive filtering algorithm to dynamically adjust the filtering parameters according to the local features of the image, remove salt and pepper noise and Gaussian noise in the image, and retain the edge and detail information of the image. The filter coefficient update formula of the adaptive filtering algorithm is: C n+1 =C n +μ×e n × n , where C n+1 is the next filtering coefficient, C n is the current filter coefficient, μ is the learning rate, e n is the error signal, x n is the input signal; The denoised image is geometrically corrected. Based on the flight attitude data of the drone, including the heading angle α, pitch angle β, roll angle γ, and camera parameters, including focal length f and pixel size d, the image is corrected using the perspective transformation algorithm. The calculation of the perspective transformation matrix M is based on the following formula: Where m ij The value of is related to the parameters α, β, γ, f, and d, and is derived through complex geometric derivation; S4, garbage identification and location step: Use deep learning-based image recognition algorithm to analyze the preprocessed images, and construct a neural network model with a training sample number of no less than 100,000, covering various types of water surface garbage images in different environments. Through model training, water surface garbage can be identified. The deep learning image recognition algorithm adopts a convolutional neural network and a recurrent neural network architecture. The convolutional neural network is used to extract the spatial features of the image, and the recurrent neural network is used to process the image sequence information. For the identified garbage on the water surface, the image coordinate and geographic coordinate conversion model is used, combined with the real-time location information of the drone and the camera shooting angle, to calculate the geographical location of the garbage on the water surface, and record the area and shape information of the garbage. Suppose the pixel coordinates of the garbage in the image are (x p ,y p ), the longitude and latitude coordinates of the drone are The pitch angle of the camera is θ, and the heading angle is The geographical coordinates of the garbage The calculation formula is: Where R is the radius of the earth, and accurate positioning can be achieved through this formula; S5, data storage and analysis step: storing the identification results, geographical location, area, and shape data of the garbage in a database, which adopts a distributed storage architecture and has a data redundancy backup function; Conduct statistical analysis on the stored data, including analysis of garbage distribution density, analysis of garbage type proportion, and analysis of garbage volume change trends over time, to provide data support and decision-making basis for water surface garbage management; S6, early warning and feedback step: According to the preset garbage threshold parameters, when the garbage distribution density or garbage volume in the monitoring area exceeds the set threshold, the system automatically generates early warning information, which includes the location of the early warning area and the description of the garbage exceeding the standard, and is promptly pushed to relevant management departments and staff through SMS, email, system pop-up windows and other methods; A feedback mechanism will be established. After receiving early warning information, relevant departments will feedback the implementation status and cleaning effect information of the cleanup action to the monitoring system. The system will correct and improve the monitoring data and analysis results based on the feedback information.

2. The method for monitoring water surface garbage based on drone according to claim 1 is characterized in that: In the UAV route planning step, the intelligent algorithm uses a combination of genetic algorithm and simulated annealing algorithm. The genetic algorithm is first used for global search, and then the simulated annealing algorithm is used to locally optimize the search results. The crossover probability P in the genetic algorithm is c Set to 0.6-0.9, the mutation probability P m It is set to 0.01-0.1, the initial temperature T0 in the simulated annealing algorithm is set according to the problem scale, and the temperature reduction coefficient α is set to 0.9-0.

99.

3. The method for monitoring water surface garbage based on drone according to claim 1 is characterized in that: In the image acquisition step, the high-definition image acquisition equipment is also equipped with infrared thermal imaging function to assist in identifying surface garbage at night or in low light conditions. The infrared thermal imaging equipment uses microbolometer focal plane array technology, and the response rate R of its detector is calculated as: R = V s / P, where V s is the detector output voltage, and P is the incident radiation power.

4. The method for monitoring water surface garbage based on drone according to claim 1 is characterized in that: In the image preprocessing step, after geometric correction, the image is also subjected to contrast enhancement. The histogram equalization algorithm is used. The original grayscale level of the image is set to r. k , the gray level after transformation is s k , the total number of pixels in the image is n, and the gray level is r k The number of pixels is n k , then the transformation function of histogram equalization is: This function is used to achieve contrast enhancement.

5. The method for monitoring water surface garbage based on drone according to claim 1 is characterized in that: In the garbage identification and location steps, the convolutional neural network used by the deep learning image recognition algorithm has a convolution kernel size of 3×3 or 5×5, a step size of 1 or 2, and a filling method of SAME or VALID. The feature extraction effect is optimized by adjusting these parameters.

6. The method for monitoring water surface garbage based on drone according to claim 1 is characterized in that: During the data storage and analysis steps, the database supports remote access and sharing of data. Relevant scientific research institutions or environmental protection organizations can remotely obtain monitoring data within the authorized scope for research and cooperation projects on water surface garbage management. The data sharing security adopts encrypted transmission and identity authentication technology. The encryption algorithm adopts AES-256-bit encryption algorithm, and identity authentication adopts digital certificate technology.

7. The method for monitoring water surface garbage based on drone according to claim 1 is characterized in that: In the early warning and feedback steps, the early warning information also includes cleaning route suggestions generated based on the garbage distribution. The cleaning route suggestions are generated using the shortest path algorithm based on the geographical distribution of the garbage, surrounding traffic and the distribution of cleaning facilities. The shortest path algorithm uses the Dijkstra algorithm or the A* algorithm.

8. The method for monitoring water surface garbage based on drone according to claim 1 is characterized in that: During the entire monitoring process, the drone's energy management system monitors the battery power B in real time. When B is lower than 20%, it automatically starts the low-battery return mode and flies back to the charging point according to the pre-planned shortest return path. The shortest return path planning adopts a path planning algorithm based on energy consumption optimization. This algorithm takes into account the distance factor while taking into account the energy consumption of the drone during flight, including energy consumption compensation when flying against the wind.

Images