Marine oil spill path planning system and method based on multi-source data fusion

Abstract

The invention discloses a marine oil spill path planning system and method based on multi-source data fusion, and belongs to the technical field of marine environment protection. According to the marine oil spill path planning method, real-time image data of an oil spill area is obtained through a multispectral camera carried by an unmanned aerial vehicle, oil spill thickness, flow velocity and oil contamination component data are collected through multiple sensors carried by a ship, meteorological data such as wind speed, wind direction and sea wave height are obtained, multi-source data are input into a reinforcement learning model, and a marine oil spill path is planned. And data fusion is carried out through dynamic weight distribution, a global optimal path is generated based on fused data, and ship-unmanned aerial vehicle cooperative operation is realized to execute an oil spill recovery task. The system comprises an unmanned aerial vehicle module, a ship module and a ground control module. Experiments show that compared with a ship oil cleaning method which does not adopt multi-source data fusion, the system and the method have the advantages that the path planning efficiency can be improved by 25%, the oil spill diffusion area can be reduced by more than 20%, the recovery efficiency can be improved by 30%, and the response time is less than or equal to 10 minutes.

Description

Technical Field

[0001] The present invention relates to the technical field of marine environmental protection, and more particularly to a marine oil spill path planning system and method based on multi-source data fusion. Background Art

[0002] Traditional oil spill response path planning relies on a single data source, such as satellite remote sensing or manual inspections, and has many drawbacks:

[0003] Satellite data has a long update cycle and cannot track oil spill dynamics in real time, resulting in delayed response;

[0004] Manual inspections are limited by weather and sea conditions, resulting in high missed inspection rates and insufficient accuracy;

[0005] The drone, ship and meteorological data are not effectively integrated, forming data islands, which makes path planning only reach local optimization rather than global optimization, resulting in low oil spill recovery efficiency. Summary of the Invention

[0006] 1. Technical problem to be solved by the invention

[0007] In response to the above-mentioned problems existing in the existing technology for marine oil spill treatment, the present invention provides a marine oil spill path planning system and method based on multi-source data fusion. Through a dynamic fusion mechanism, DQN path decision-making and collaborative control, a collaborative operation closed-loop system is formed, thereby improving the efficiency of oil spill recovery.

[0008] 2. Technical solution

[0009] In order to achieve the above object, the technical solution provided by the present invention is:

[0010] Dynamic fusion mechanism: UAV boundary identification, ship oil pollution parameters and meteorological diffusion prediction are weighted and integrated through Kalman filtering to ensure the accuracy and timeliness of data.

[0011] DQN path decision: With recycling efficiency, path length, and energy consumption as reward targets, the deep Q network model is used to output the global optimal path and realize intelligent path planning.

[0012] Collaborative control: The ground server adjusts the ship recovery route and drone monitoring area in real time, forming a collaborative closed-loop system to improve oil spill recovery efficiency.

[0013] 3. Beneficial effects

[0014] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects:

[0015] Response time is significantly shortened: Path planning response time is ≤10 minutes, while traditional methods are ≥2 hours, greatly improving emergency handling capabilities.

[0016] The recovery coverage rate has been significantly improved: the oil spill recovery efficiency has increased by 30%, while the traditional method average is only 60%, effectively improving the oil spill recovery efficiency.

[0017] The diffusion area is significantly reduced: the diffusion area is reduced, which reduces the risk of secondary pollution and better protects the marine environment.

[0018] Reduced energy consumption costs: Energy consumption costs were reduced by 18%, achieving optimal utilization of resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 : Architecture diagram of the marine oil spill path planning system based on multi-source data fusion in a specific embodiment;

[0020] Figure 2 : Flowchart of the ocean oil spill path planning algorithm based on multi-source data fusion in a specific embodiment;

[0021] Figure 3 : Comparison chart of simulation results of specific embodiments;

[0022] Figure 4 : Pie chart of contribution analysis of multi-source data modules in a specific embodiment.

[0023] Figure 5 : Comparative analysis diagram of path planning response time of specific embodiments. DETAILED DESCRIPTION

[0024] In order to further understand the content of the present invention, the invention is described in detail with reference to the accompanying drawings.

[0025] Example 1

[0026] The marine oil spill path planning system and method based on multi-source data fusion of this embodiment is applied to the oil spill treatment in the Bohai Sea area:

[0027] The marine oil spill path planning system includes a drone module, a ship module and a ground control module; Figure 1 As shown, the data interaction process between the UAV, ship and ground control modules is demonstrated, and the information transmission and collaborative working mechanism between the modules are clearly presented to form a collaborative operation closed-loop system.

[0028] UAV module: The UAV is equipped with a multispectral camera and an image transmission device. The multispectral camera is used to obtain real-time image data of the oil spill area;

[0029] Ship module: equipped with an oil-water separation device and a sensor array, including sensors for collecting oil spill thickness, flow rate and oil composition;

[0030] Ground control module: Integrates data fusion server and path planning algorithm. The data fusion server is used to fuse multi-source data, and the path planning algorithm is used to generate the global optimal path.

[0031] The specific steps are:

[0032] Data collection

[0033] Drones: Acquire 0.5m resolution images and extract the oil spill boundary through semantic segmentation. The high-resolution images clearly identify the boundaries of the oil spill area, providing accurate basic data for subsequent path planning.

[0034] Ship sensors: The various sensors installed on the ship collected data showing an oil spill thickness of 1.2 cm. After Kalman filtering, the accuracy was improved by 20%. Data on a flow rate of 0.8 m / s and oil spill composition were also obtained. Kalman filtering technology effectively reduced the noise in the sensor data and improved its reliability.

[0035] Meteorological station: Various sensors installed on the ship obtain meteorological data: wind speed 6, wave height 1.5m, wind direction northwest, and forecast that the oil spill spread area will increase by 15% within 2 hours. Accurate weather forecasts provide a basis for judging the oil spill spread trend.

[0036] Path generation and execution

[0037] like Figure 4 As shown in the figure, multi-source data consisting of real-time image data of the oil spill area, oil composition data, and meteorological data are input into the reinforcement learning model, and data fusion is performed through dynamic weight allocation; drone image recognition, ship sensor data, and meteorological forecast information contribute 20%, 35%, and 45% to the final path optimization respectively; the above data, after being fused by the dynamic data integration server, are input into the reinforcement learning training DQN model, and the total length of the output path is reduced by 25%. Figure 2 As shown in the figure, it includes data fusion, model training and path generation steps, and describes in detail the entire algorithm process from data collection to path generation. The model makes global optimal path decisions and generates a better recycling path by learning and optimizing multi-source data.

[0038] Based on data fusion, a global optimal path is generated, and the oil spill recovery task is performed through ship-UAV collaborative operation: the ship operates according to the planned path, and the recovery data is fed back to the DQN model in real time. More than 85% of the spilled oil is recovered within 6 hours, while the traditional method can only recover 60%. The significantly improved recovery efficiency demonstrates the superiority of the method of this embodiment.

[0039] Drones update the boundaries of the oil spill in real time and dynamically adjust the recovery area to ensure that ships always operate according to the actual distribution of the oil spill.

[0040] Experimental verification

[0041] Simulation comparison shows that after testing, the recovery efficiency of this embodiment reaches 92%, which is much higher than the 62% of the traditional method.

[0042] Energy consumption costs are reduced by 18%, thanks to the dynamic suppression of high-energy consumption paths by the γ coefficient in the reward function, achieving energy-saving operations.

[0043] from Figure 3 It can be clearly seen that the method of this embodiment is superior to the traditional method in three dimensions: recovery efficiency (62%: 92%), energy consumption cost (100%: 82%) and total path length (100%: 75%), which fully verifies the technical advantages of the method of this embodiment.

[0044] Example 2

[0045] The marine oil spill path planning system and method based on multi-source data fusion of this embodiment has the same basic equipment configuration and steps as Example 1, with the difference or improvement being that it is applied to the oil spill treatment in the Yellow Sea:

[0046] Drones: Equipped with multispectral cameras, they can obtain real-time image data of the oil spill area and identify the distribution range of thin oil layers through image analysis.

[0047] Ship sensors: Infrared oil film sensors collect oil spill thickness data. Because the oil layer is thin, most are less than 2mm thick. The sensor has a measurement range of 0-10mm and an accuracy of ±0.5mm, enabling accurate measurement of thin oil film thickness.

[0048] Meteorological data: The meteorological information obtained was wind speed level 3 and wave height 0.8m. The meteorological conditions were relatively stable and had little impact on the spread of oil spills.

[0049] Path planning and data fusion

[0050] Semantic segmentation is performed on drone images to extract the boundaries of thin oil layers. Since the features of thin oil layers are relatively unclear in images, a more sophisticated semantic segmentation algorithm is needed to improve the accuracy of boundary extraction.

[0051] The ship sensor data was subjected to Kalman filtering for noise reduction and further processed with a second-order low-pass filter (cut-off frequency of 2 Hz) to reduce data fluctuations.

[0052] The oil spill diffusion trend is predicted based on meteorological data. Due to stable meteorological conditions, the oil spill diffusion speed is predicted to be slow.

[0053] The processed data is fed into the DQN model. Taking into account the characteristics of thin oil layers, the weight coefficients α, β, and γ in the reward function R are dynamically adjusted, with greater emphasis on improving recovery efficiency. R = α·recovery efficiency + β·path length + γ·energy cost;

[0054] Among them, α, β, and γ are dynamically adjusted weight coefficients.

[0055] Collaborative job execution

[0056] The ship operates according to the generated global optimal path. During the operation, the oil suction plate speed is adjusted by the dynamic roll compensation algorithm. The basic speed is determined by the oil film thickness. When the oil film thickness δ≤2mm, V base =200rpm; calculate the compensation coefficient K by the roll angle θ comp , to achieve real-time adjustment of the rotation speed and ensure that the oil suction plate can still work efficiently when the ship rolls.

[0057] The drone monitors the changes in the thin oil layer in real time and feeds the data back to the ground control module, which dynamically adjusts the path and operation parameters based on the feedback data.

[0058] Technical Effects

[0059] After testing, the recovery rate of thin oil layers reached more than 72%, which is a significant improvement compared with traditional methods. This is mainly due to the accurate identification of thin oil layer characteristics and targeted path planning and operation adjustments.

[0060] The path planning response time is controlled within 10 minutes, meeting the time requirements for emergency handling.

[0061] The energy consumption cost is reduced by about 15% compared with traditional methods, achieving energy saving while ensuring recycling efficiency.

[0062] Example 3

[0063] The marine oil spill path planning system and method based on multi-source data fusion of this embodiment has the same basic equipment configuration and steps as Example 1, with the difference or improvement being that it is applied to the oil spill treatment in the East China Sea under complex meteorological conditions:

[0064] Data collection

[0065] Drones: Under complex weather conditions (wind speed level 8, wave height 2.5m), they overcame strong winds and other adverse factors to obtain real-time image data of the oil spill area. Due to the adverse weather conditions, the images may contain some blur and noise, requiring more rigorous subsequent processing.

[0066] Ship sensors: Collect data on oil spill thickness, flow rate, and oil composition. Under the influence of strong waves, the stability of sensor data is challenged and requires multiple noise reduction processes using technologies such as Kalman filtering.

[0067] Meteorological data: We obtain meteorological information on strong winds and high waves, and predict that the oil spill will spread quickly, and the direction of spread will be greatly affected by wind direction and ocean currents.

[0068] Data processing and path generation

[0069] Semantic segmentation of drone images requires a combination of multiple image processing techniques, such as denoising and enhancement, to improve the accuracy of oil pollution boundary extraction due to the poor image quality.

[0070] The ship sensor data was subjected to Kalman filtering for noise reduction and processed with a second-order low-pass filter (cut-off frequency of 2 Hz) to reduce the violent fluctuations in the sensor data caused by strong waves.

[0071] Based on complex meteorological data, a more accurate oil spill diffusion prediction model is established, fully considering the combined impact of factors such as wind speed, wind direction, wave height and ocean currents.

[0072] The fused data is input into the DQN model, and the weight coefficient is dynamically adjusted, paying more attention to the adaptability of the path and the timeliness of recovery.

[0073] Collaborative work and dynamic adjustment

[0074] The ship performs the recovery operation according to the generated path. During the operation, the ship's roll angle θ and oil film thickness δ are monitored in real time. When the roll angle is large, the dynamic roll compensation algorithm V set =V base ×K comp Adjust the oil suction plate speed to ensure the normal operation of the oil suction plate. The basic speed is calculated according to the oil film thickness. When δ>5mm, V base =350rpm.

[0075] The drone maintains a stable flight despite complex weather conditions, updating oil spill boundary data in real time and rapidly transmitting this data to the ground control module. Based on this updated data, the ground control module rapidly adjusts the vessel's recovery path and operating parameters, such as the suction disc's rotation speed and the vessel's travel speed.

[0076] Technical Effects

[0077] According to tests, under complex meteorological conditions, the method of this embodiment can still achieve a good recovery effect, with a recovery efficiency of more than 80%, which is significantly higher than the 50% of the traditional method under the same conditions.

[0078] The route planning response time is controlled within 10 minutes. Although the weather conditions are complex, the timeliness of the response is still guaranteed through efficient data processing and model optimization.

[0079] The diffusion area is reduced by more than 20% compared with traditional methods, effectively controlling the spread of oil spills and reducing further pollution to the marine environment.

[0080] from Figure 5 It can be seen that the response time of the method of this embodiment is significantly better than that of traditional methods in all stages of path planning, especially in the path generation and dynamic adjustment link, which only takes 2 minutes, greatly enhancing the emergency response capability under complex weather conditions.

[0081] Summary of technical effects of Examples 1-3:

[0082] The drone boundary identification, ship oil pollution parameters and meteorological diffusion prediction are weighted and integrated through Kalman filtering to form a dynamic fusion mechanism to ensure the accuracy and timeliness of the data.

[0083] Taking recycling efficiency, path length, and energy consumption as reward targets, the deep Q network model is used to output the global optimal path, form DQN path decision-making, and realize intelligent path planning.

[0084] The ground server adjusts the ship recovery route and drone monitoring area in real time, forming a collaborative control operation closed-loop system and improving the efficiency of oil spill recovery.

[0085] Through multi-source data fusion and reinforcement learning technology, intelligent planning and collaborative operation of marine oil spill paths are realized, and excellent performance is shown in different sea areas, different oil layer thicknesses and different meteorological conditions. It can improve path planning efficiency by 25%, reduce the oil spill diffusion area by more than 20%, increase recovery efficiency by 30%, and the response time is ≤10 minutes. It has good practical prospects and technical promotion value.

[0086] The above is a schematic description of the present invention and its embodiments, which is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure and manufacturing steps are not limited thereto. Therefore, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without creatively designing them, they shall fall within the scope of protection of the present invention.

Claims

1. A method for ocean oil spill path planning based on multi-source data fusion, characterized in that: The following steps are involved: (1) Obtain real-time image data of the oil spill area through the multispectral camera carried by the drone; (2) Collect data on oil spill thickness, flow rate, and oil composition through various sensors installed on the ship; (3) Sensors acquire meteorological data, including wind speed, wind direction, and wave height; (4) Multi-source data consisting of real-time image data of the oil spill area, oil pollution composition data, and meteorological data are input into the reinforcement learning model, and data fusion is performed through dynamic weight allocation; (5) Generate the global optimal path based on data fusion, and perform the oil spill recovery task through ship-UAV collaborative operation.

2. The planning method according to claim 1, characterized in that: The reinforcement learning model adopts a deep Q network, and its reward function is designed as: R = α·recovery efficiency + β·path length + γ·energy cost; Among them, α, β, and γ are dynamically adjusted weight coefficients.

3. The planning method according to claim 1, wherein: The data fusion step includes: (1) Perform semantic segmentation on drone images and extract the oil spill boundary; (2) Perform Kalman filtering to reduce noise on ship sensor data; (3) Predict the oil spill diffusion trend based on meteorological data.

4. The planning method according to claim 1, wherein: The ship sensor includes an infrared oil film sensor for collecting the thickness of the oil spill, with a measurement range of 0-10mm and an accuracy of ±0.5mm.

5. The planning method according to claim 1, characterized in that: During the collaborative operation, when the ship performs the recovery operation, the oil suction plate speed is adjusted by the dynamic roll compensation algorithm. The speed synthesis formula of the dynamic roll compensation algorithm is: V set =V base ×K comp , where V base is the basic speed, K comp is the compensation coefficient.

6. A system according to any one of claims 1 to 5, characterized in that: include: UAV module: equipped with a multispectral camera and an image transmission device. The multispectral camera is used to obtain real-time image data of the oil spill area; Ship module: equipped with an oil-water separation device and a sensor array, including sensors for collecting oil spill thickness, flow rate, oil pollution composition, and meteorological information; Ground control module: integrates a data fusion server and a path planning algorithm, wherein the data fusion server is used to fuse multi-source data, and the path planning algorithm is used to generate a global optimal path.

7. The system according to claim 6, characterized in that: When the data fusion server in the ground control module processes the ship sensor data, a second-order low-pass filter is used with a cut-off frequency of 2 Hz.

Images