Intelligent target supervision method and system based on data analysis

Abstract

The invention provides an intelligent target supervision method and system based on data analysis. The method comprises the steps that visible light and thermal infrared image sequences are synchronously collected through a coastline monitoring camera, and plastic garbage contour features and temperature distribution differences are recognized; segmenting the contour, extracting displacement to generate an ocean current trajectory prediction path, arranging a buoy collection device, and driving surface water flow to converge by using a propeller to guide garbage to move along the path to form a dynamic accumulation area; plastic garbage difficult to decompose is recognized and marked based on visible light reflection difference, offset is calculated in combination with temperature difference to generate a path correction instruction, the track of a collecting device is adjusted to preferentially intercept a target, and real-time images are synchronized to update closed-loop supervision. According to the technical scheme, intelligent recognition, dynamic gathering and precise interception closed-loop supervision of marine plastic garbage difficult to decompose are achieved, the pollution treatment efficiency is improved, and secondary diffusion is restrained.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent target supervision, and in particular to an intelligent target supervision method and system based on data analysis. BACKGROUND

[0002] In the dynamic distribution tracking scene of island plastic waste, it is necessary to achieve the following: real-time monitoring of the migration path of plastic waste driven by tides, monsoons and ocean currents, overcoming the problem of missing pollution hotspots caused by limited coverage of traditional manual patrols; fusing multi-modal sensing data and ocean dynamics models to dynamically build a spatio-temporal evolution atlas of waste accumulation areas to support intelligent scheduling of cleaning vessels; designing a plastic feature enhancement algorithm that is resistant to complex optical interference, which can accurately distinguish between floating plastics and natural objects such as seaweed and foam in strong wave shielding and salt spray lens environments, reducing the misjudgment rate and improving the tracing efficiency.

[0003] The current mainstream solution uses a multi-source federated learning framework and a spatio-temporal causal graph convolution network collaborative model: a distributed sensing network is constructed by deploying a cluster of unmanned aerial vehicles equipped with hyperspectral imagers and near-shore buoy arrays to collect plastic distribution data; cross-device multi-modal data feature extraction and encrypted parameter sharing are realized based on the federated learning architecture, combined with ocean current vector field data provided by satellites, and a spatio-temporal causal graph convolution network is used to model the spatio-temporal correlation of waste migration; an edge-end lightweight model is deployed, a polarized light fusion algorithm is used to suppress wave interference, and reinforcement learning is used to dynamically optimize the cleaning path planning strategy to realize real-time decision-making of the cloud monitoring platform.

[0004] However, the encrypted parameter sharing mechanism of the federated learning framework is vulnerable to malicious node gradient reverse attacks, leading to the risk of leakage of plastic distribution privacy data; the spatio-temporal graph convolution network has insufficient generalization ability for the spectral features of new microplastics (such as biodegradable fragments), mistakenly identifying coral secretions as plastic pollutants; the polarized light algorithm fails in heavy rain or fog due to a sudden drop in penetration rate, resulting in feature misjudgment in wave shielding areas; the compression of the edge-end model leads to a sharp increase in the detection missing rate of small plastic fragments in the long-tail distribution; the federated gradient bias caused by device heterogeneity amplifies the spatio-temporal graph distortion when the ocean current suddenly changes, and the spatio-temporal coupling degree of cleaning path planning and dynamic pollution hotspots continues to decline. SUMMARY

[0005] The present application provides an intelligent target supervision method and system based on data analysis to solve the problem of inaccurate plastic waste identification and disconnection between collection and regulation in the prior art, resulting in low interception efficiency.

[0006] In a first aspect, the present application provides an intelligent target supervision method based on data analysis, comprising:

[0007] A continuous image sequence containing visible light and thermal infrared bands is captured by a monitoring camera distributed along the coastline, and the contour features of plastic garbage under visible light and the temperature distribution difference in thermal infrared images in the continuous image sequence are identified;

[0008] The contour boundary of the plastic garbage is segmented according to the contour features, and the position change of the contour boundary in the continuous image sequence is obtained to generate a trajectory prediction path of the plastic garbage moving on the sea surface along the ocean current;

[0009] A buoy-type collection device is arranged in the sea area covered by the trajectory prediction path, the collection device causes the propeller to generate a surface water flow converging to the storage bin direction of the collection device, and the plastic garbage is guided to move along the direction of the trajectory prediction path by the surface water flow to form a dynamic aggregation area;

[0010] In the dynamic aggregation area, the reflection characteristic difference of the surface of the plastic garbage under visible light is analyzed, and the difficult-to-decompose plastic garbage with a preset degradation attribute is identified and marked as a supervision target according to the reflection characteristic difference;

[0011] Based on the temperature distribution difference of the supervision target in the dynamic aggregation area, the offset of the supervision target is calculated, and path correction instructions are generated according to the offset to correct the moving trajectory of the collection device, so that the collection device preferentially intercepts the difficult-to-decompose plastic garbage with the corrected moving trajectory, and completes the target supervision closed loop based on the real-time data update of the continuous image sequence.

[0012] Optionally, based on the temperature distribution difference of the supervision target in the dynamic aggregation area, the offset of the supervision target is calculated, and path correction instructions are generated according to the offset to correct the moving trajectory of the collection device, so that the collection device preferentially intercepts the difficult-to-decompose plastic garbage with the corrected moving trajectory, and completes the target supervision closed loop based on the real-time data update of the continuous image sequence, including:

[0013] According to the temperature distribution difference of the supervision target in the dynamic aggregation area, the position coordinates of the supervision target with a temperature distribution difference greater than a preset value in the thermal infrared image are obtained, and the current position coordinates of the collection device are obtained;

[0014] The position coordinates of the supervision target and the current position coordinates of the collection device are compared in space to calculate the moving direction deviation angle and deviation distance of the supervision target relative to the collection device;

[0015] The lateral offset and the longitudinal offset of the supervision target in the moving direction of the collection device are calculated based on the moving direction deviation angle and the deviation distance, a path correction instruction is generated in combination with the superposition relationship of the lateral offset and the longitudinal offset, and the path correction instruction includes an amplitude parameter and a direction parameter;

[0016] The rotational speed ratio of the propeller of the collection device is adjusted according to the amplitude parameter and the direction parameter of the path correction instruction, the propeller in the corresponding direction of the lateral offset generates a reverse compensation thrust, the propeller thrust output in the corresponding direction of the longitudinal offset matches a preset ratio, and the moving track of the collection device is corrected;

[0017] The collection device preferentially intercepts the difficult-to-decompose plastic garbage in the corrected moving track, and the moving track of the collection device is continuously adjusted based on the real-time data of the continuous image sequence, so as to complete the closed-loop interception of the difficult-to-decompose plastic garbage.

[0018] Optionally, the lateral offset and the longitudinal offset of the supervision target in the moving direction of the collection device are calculated based on the moving direction deviation angle and the deviation distance, a path correction instruction is generated in combination with the superposition relationship of the lateral offset and the longitudinal offset, and the path correction instruction includes an amplitude parameter and a direction parameter, including:

[0019] The lateral offset and the longitudinal offset of the supervision target in the moving direction of the collection device are calculated based on the moving direction deviation angle and the deviation distance, a path correction instruction is generated in combination with the superposition relationship of the lateral offset and the longitudinal offset, and the path correction instruction includes an amplitude parameter and a direction parameter, including:

[0020] The direction correction instruction is generated according to the relative relationship of the lateral offset and the longitudinal offset, and the amplitude correction instruction of the path correction instruction is generated according to the overall offset degree of the lateral offset and the longitudinal offset;

[0021] The direction correction instruction and the amplitude correction instruction are fused to obtain the path correction instruction including the amplitude parameter and the direction parameter.

[0022] Optionally, the rotational speed ratio of the propeller of the collection device is adjusted according to the amplitude parameter and the direction parameter of the path correction instruction, the propeller in the corresponding direction of the lateral offset generates a reverse compensation thrust, the propeller thrust output in the corresponding direction of the longitudinal offset matches a preset ratio, and the moving track of the collection device is corrected, including:

[0023] The thrust intensity of the propeller is determined according to the amplitude parameter of the path correction instruction, and the thrust intensity corresponds to the numerical value of the amplitude parameter;

[0024] According to the direction parameter of the path correction instruction, the direction of the propeller thrust on the side corresponding to the lateral offset is adjusted to be a reverse compensation thrust, the reverse compensation thrust is opposite to the deviation angle of the moving direction, so as to offset the influence of the lateral offset on the moving track;

[0025] According to the preset proportion, the strength of the propeller thrust in the direction corresponding to the longitudinal offset is adjusted, so that the propeller thrust output in the direction corresponding to the longitudinal offset matches the amplitude parameter, so as to correct the longitudinal offset;

[0026] The adjustment process of the direction of the propeller thrust and the strength of the propeller thrust is repeated to correct the moving track of the collection device.

[0027] Optionally, the contour boundary of the plastic garbage is segmented according to the contour feature, and the position change of the contour boundary in the continuous image sequence is obtained to generate a trajectory prediction path of the plastic garbage moving on the sea surface under the action of the ocean current, including:

[0028] The contour boundary of the plastic garbage is segmented according to the shape closure and edge continuity of the contour feature;

[0029] The position change of the same contour boundary in the continuous image sequence is tracked, and the moving direction and displacement of the contour boundary are obtained according to the position change;

[0030] An ocean current moving trend is obtained, the ocean current moving trend is corrected according to a preset correction proportion to generate a predicted moving direction of the contour boundary;

[0031] According to the predicted moving direction, a trajectory prediction path of the plastic garbage under the action of the ocean current is generated, and the sea area range covered by the trajectory prediction path is determined by the displacement of the contour boundary.

[0032] Optionally, a buoy-type collection device is arranged in the sea area range covered by the trajectory prediction path, the collection device causes the propeller to generate a surface water flow converging in the direction of the storage bin of the collection device, and the plastic garbage is guided to move along the direction of the trajectory prediction path by the surface water flow to form a dynamic gathering area, including:

[0033] A buoy-type collection device is arranged in the sea area range covered by the trajectory prediction path, the storage bin of the collection device is oriented in the extension direction of the trajectory prediction path, and the propeller of the collection device is started to rotate in the direction of the storage bin to generate a surface water flow converging in the direction of the storage bin;

[0034] The rotation direction and thrust distribution of the propeller are adjusted so that the flow direction of the surface water flow is consistent with the extension direction of the trajectory prediction path;

[0035] Plastic garbage in the sea area covered by the trajectory prediction path is gathered in the direction of the trajectory prediction path to form a dynamic gathering area.

[0036] Optionally, in the dynamic gathering area, differences in reflection characteristics of the plastic garbage under visible light in the continuous image sequence are analyzed, and plastic garbage with a preset degradation attribute is identified according to the differences in reflection characteristics and marked as a supervision target, including:

[0037] The reflection intensity values of the surfaces of the plastic garbage under visible light in the continuous image sequence in the dynamic gathering area are extracted, and different materials of the plastic garbage result in different reflection intensity values of the surfaces of the plastic garbage under visible light;

[0038] The reflection intensity values of different plastic garbage are compared to obtain differences in reflection characteristics of different plastic garbage;

[0039] According to the differences in reflection characteristics, the reflection intensity values are compared with reflection characteristics corresponding to a preset degradation attribute, the reflection characteristics are reflection intensity values, and plastic garbage with a reflection intensity value matching a degradation attribute of difficult-to-decompose plastic is screened out;

[0040] The matched plastic garbage is regarded as difficult-to-decompose plastic garbage, and the difficult-to-decompose plastic garbage is marked as a supervision target.

[0041] In a second aspect, the present application provides an intelligent target supervision system based on data analysis, including:

[0042] A capture module is configured to capture a continuous image sequence containing a visible light and a thermal infrared wave band through a monitoring camera distributed along a coastline, and identify differences in contour features of plastic garbage under visible light and temperature distribution in a thermal infrared image in the continuous image sequence;

[0043] A segmentation module is configured to segment a contour boundary of the plastic garbage according to the contour features, and obtain a position change of the contour boundary in the continuous image sequence to generate a trajectory prediction path of the plastic garbage moving on the sea surface along an ocean current;

[0044] A layout module is configured to layout a buoy-type collection device in a sea area covered by the trajectory prediction path, the collection device causes a surface water flow generated by a propeller to converge in a direction of a storage bin of the collection device, and the plastic garbage is guided by the surface water flow to move in a direction of the trajectory prediction path to form a dynamic gathering area;

[0045] an analysis module configured to analyze, in the dynamic aggregation area, a difference in reflection characteristics of the plastic garbage surface under visible light in the continuous image sequence, and identify and mark as a supervision target the difficult-to-decompose plastic garbage having a preset degradation attribute according to the difference in reflection characteristics;

[0046] a calculation module configured to calculate a displacement of the supervision target based on a temperature distribution difference of the supervision target in the dynamic aggregation area, and generate a path correction instruction to correct a moving track of the collection device according to the displacement, so that the collection device preferentially intercepts the difficult-to-decompose plastic garbage in the corrected moving track, and complete a target supervision closed loop based on real-time data update of the continuous image sequence.

[0047] The application can realize multispectral collaborative detection of marine plastic garbage by capturing a continuous image sequence containing visible light and thermal infrared wave band through a coastline distribution monitoring camera, and identifying a difference between a contour feature of the plastic garbage under visible light and a temperature distribution in a thermal infrared image in the continuous image sequence. The application can accurately predict the drifting direction of the plastic garbage by segmenting the contour boundary of the plastic garbage according to the contour feature, and obtaining a position change of the contour boundary in the continuous image sequence to generate a trajectory prediction path of the plastic garbage moving on the sea surface along the ocean current. The application can realize active aggregation and interception of plastic garbage by arranging a buoy-type collection device in a sea area covered by the trajectory prediction path, so that the collection device makes the propeller generate a surface water flow converging to the direction of the storage bin of the collection device, and guides the plastic garbage to move along the direction of the trajectory prediction path through the surface water flow to form a dynamic aggregation area. The application can accurately screen difficult-to-decompose plastics with high environmental risk by analyzing, in the dynamic aggregation area, a difference in reflection characteristics of the plastic garbage surface under visible light in the continuous image sequence, and identifying and marking as a supervision target the difficult-to-decompose plastic garbage having a preset degradation attribute according to the difference in reflection characteristics. The application can realize accurate capture and dynamic tracking of difficult-to-decompose plastics, and form an intelligent marine plastic pollution control system by calculating a displacement of the supervision target based on a temperature distribution difference of the supervision target in the dynamic aggregation area, generating a path correction instruction to correct a moving track of the collection device according to the displacement, so that the collection device preferentially intercepts the difficult-to-decompose plastic garbage in the corrected moving track, and completing a target supervision closed loop based on real-time data update of the continuous image sequence.

[0048] Further, by analyzing the temperature distribution difference of the monitoring target in the thermal infrared image, the spatial coordinates of the hard-to-degrade plastic garbage are accurately located, and the offset angle and distance are calculated combined with the real-time position of the collection device, so that high-precision dynamic tracking of the target object can be realized. By decomposing the lateral and longitudinal offset amounts and generating path correction instructions containing amplitude and direction parameters, the propeller speed ratio can be intelligently adjusted to generate a reverse compensation thrust, so that the collection device can quickly respond and correct the moving track. Based on the real-time image data, the propeller thrust output ratio is continuously adjusted to form a closed-loop interception system for hard-to-degrade plastic garbage, and finally the precise capture and dynamic removal of hard-to-degrade plastic are realized, effectively improving the pertinence and operation efficiency of marine plastic pollution governance.

[0049] These aspects or other aspects of the present application will be more apparent in the following description of the embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0051] Figure 1 A flow chart of an intelligent target monitoring method based on data analysis provided by the present application is shown;

[0052] Figure 2 A scene diagram of an intelligent target monitoring method based on data analysis provided by the present application is shown;

[0053] Figure 3 A structural schematic diagram of an intelligent target monitoring method based on data analysis provided by the present application is shown. DETAILED DESCRIPTION

[0054] In order to make those skilled in the art better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application.

[0055] In some of the flowcharts described in the specification and claims of the present application and in the above description of the drawings, a plurality of operations are included which occur in a particular order, but it should be clearly understood that the operations can be performed in an order other than that in which they appear herein or in parallel, and the serial numbers of the operations, such as 101, 102, etc., are merely used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, the flowcharts can include more or fewer operations, and the operations can be executed in sequence or in parallel. It should be noted that the descriptions herein, such as "first", "second", etc., are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit the "first" and "second" to be different types.

[0056] Researchers found that existing methods for monitoring marine plastic waste rely on single-spectrum data or static recycling strategies, making it difficult to dynamically track ocean current displacement and accurately identify degradation properties, resulting in low waste aggregation efficiency, difficulty in decomposing target supervision. Based on this, a method for dynamically tracking and preferentially intercepting marine plastic waste is provided, which can realize the accurate recycling of difficult-to-decompose waste through multi-spectrum feature collaborative analysis and closed-loop path correction. The technical solution of the present application can be applied to marine plastic pollution control, nearshore ecological protection and intelligent waste recycling scenarios.

[0057] The entire research and development process embodies the intelligent collaborative mechanism of multi-spectrum data fusion and dynamic closed-loop feedback, aiming to overcome the defects of single data dimension, large trajectory prediction deviation and lack of target classification supervision in existing solutions. Through complementary identification of visible light and thermal infrared features, the accuracy of plastic waste contour segmentation and degradation property identification is improved; combined with trajectory prediction driven by ocean current displacement and dynamic guidance of propeller water flow, the controllability of the waste aggregation area is enhanced; based on the calculation of offset amount and real-time path correction according to temperature distribution difference, the adaptability of traditional fixed recycling devices to complex sea conditions is broken through, realizing the preferential interception and supervision of difficult-to-decompose plastics, and providing a precise and efficient management solution for marine ecological protection.

[0058] The technical solutions in the embodiments of the present application will be described clearly and completely in the embodiments of the present application combined with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0059] Figure 1 A flowchart of an intelligent target supervision method based on data analysis is provided for the embodiments of the present application, as shown in Figure 1 The method comprises:

[0060] 101、capture a continuous image sequence containing both visible light and thermal infrared bands by a monitoring camera deployed along the coastline, and identify the contour feature of plastic waste in the visible light and the temperature distribution difference in the thermal infrared image in the continuous image sequence;

[0061] In this step, the monitoring camera refers to a camera device installed along the coastline for environmental monitoring. The visible light band refers to the range of electromagnetic spectrum that can be perceived by human eyes. The thermal infrared band refers to the range of electromagnetic spectrum that characterizes the thermal radiation of objects. The continuous image sequence refers to a set of images taken in time sequence. The contour feature refers to the visual feature of the boundary of the shape of an object. The temperature distribution difference refers to the difference in surface temperature of different regions.

[0062] In the embodiments of the present application, first, a continuous image sequence containing both visible light and thermal infrared bands is synchronously captured by a monitoring camera deployed along the coastline. The visible light image is used to capture the morphological features of plastic waste, and the thermal infrared image records the surface temperature distribution. Second, edge enhancement processing is performed on the visible light image, and an adaptive threshold segmentation algorithm is used to extract the contour feature of plastic waste. At the same time, the temperature distribution difference between the target region and the surrounding seawater is calculated through the thermal infrared image. Plastic waste usually presents an abnormal temperature zone due to the difference in heat capacity. Third, the visible light contour and the thermal infrared temperature difference abnormal region are spatially superimposed, and the region with a coincidence degree higher than a threshold is selected as the candidate target of plastic waste. Finally, a contour-temperature correlation database is established to provide a feature matching basis for subsequent trajectory prediction.

[0063] 102、segment the contour boundary of plastic waste according to the contour feature, and obtain the position change of the contour boundary in the continuous image sequence to generate a trajectory prediction path of plastic waste moving on the sea surface along the ocean current;

[0064] In this step, the contour boundary refers to the geometric feature of the edge of the shape of an object. The position change refers to the change amount of the position of an object over time. The trajectory prediction path refers to the future moving route calculated according to the motion law. The ocean current movement refers to the phenomenon of large-scale directional flow of seawater.

[0065] In the embodiments of the present application, first, based on the contour feature extracted in step 101, the optical flow method is used to track the pixel displacement of plastic waste in the continuous image sequence, and the motion vector between adjacent frames is calculated. Second, combined with the ocean current velocity vector and wind direction data, the future displacement trend of plastic waste is predicted through a Kalman filter to generate multiple possible trajectory prediction paths. Third, the confidence of the prediction path is evaluated: when the displacement direction of consecutive frames is consistent with the direction of the ocean current, the confidence weight of the path is improved. Finally, the optimal prediction path is encoded as the boundary of a geofence to provide a spatial reference for the layout of a collection device.

[0066] 103. Deploying buoy-type collection devices in the sea area covered by the trajectory prediction path, the collection devices convert the kinetic energy of sea wave fluctuation into mechanical energy to drive the rotation of the propeller, so that the propeller generates surface water flow converging to the direction of the storage bin of the collection device, and the plastic garbage is guided to move along the direction of the trajectory prediction path by the surface water flow to form a dynamic accumulation area;

[0067] In this step, the buoy-type collection device refers to the garbage collection equipment floating on the sea surface. Kinetic energy conversion refers to the conversion process of energy form. Mechanical energy refers to the energy form of the mechanical movement of an object. Propeller rotation refers to the mechanical movement of generating thrust by rotating blades. Surface water flow refers to the flow phenomenon of the surface layer of seawater. Dynamic accumulation area refers to a temporary area formed by the accumulation of garbage along the water flow.

[0068] In the embodiments of the present application, first, buoy-type collection devices are deployed in the sea area covered by the trajectory prediction path, and a hinged wave energy conversion mechanism is installed at the bottom to convert the kinetic energy of sea wave fluctuation into rotational power of the propeller shaft through a crank linkage. Secondly, the centrifugal water flow is generated when the propeller rotates at high speed, and the guide plate inside the device is designed to make the surface seawater converge to the direction of the storage bin, forming a continuous surface drainage effect. Then, the drainage intensity is controlled by adjusting the rotation speed of the propeller to ensure that the plastic garbage migrates along the predicted path to the center of the dynamic accumulation area. Finally, the density of the accumulation area is monitored in real time, and the storage bin closing instruction is triggered when the preset threshold is reached.

[0069] 104. In the dynamic accumulation area, the difference in reflection characteristics of the plastic garbage under visible light is analyzed in the continuous image sequence, and the difficult-to-decompose plastic garbage with the preset degradation attribute is identified according to the difference in reflection characteristics and marked as a supervision target;

[0070] In this step, the difference in reflection characteristics refers to the difference in reflection of light by different material surfaces. Degradation attribute refers to the decomposition characteristics of a material in a natural environment. Difficult-to-decompose plastic garbage refers to plastic waste that is difficult to be naturally degraded. Supervision target refers to a specific garbage object that needs to be treated.

[0071] In the embodiments of the present application, first, multispectral reflectance analysis is performed on the visible light images in the dynamic accumulation area, the difference in reflection characteristics of the plastic garbage surface to different wavelengths of visible light is extracted, and a reflectance fingerprint library is established. Secondly, the measured reflection characteristics are matched with the preset degradation attribute database (such as the spectral characteristics of polyethylene and polypropylene) to identify the types of plastic that are difficult to be naturally degraded. Then, the interference of fragments is removed by morphological filtering, and electronic tags are added to the difficult-to-decompose plastic garbage that has been successfully matched, and the coordinate information is uploaded to the supervision platform. Finally, a target distribution heat map is generated, and the high-density difficult-to-decompose plastic area is marked as a priority interception zone.

[0072] 105. Based on the temperature distribution difference of the supervision target in the dynamic aggregation area, the temperature distribution difference is captured by the monitoring camera distributed on the coastline, the offset of the supervision target is calculated, and the path correction instruction is generated according to the offset to correct the moving track of the collection device, so that the collection device preferentially intercepts the difficult-to-decompose plastic garbage with the corrected moving track, and a target supervision closed loop is completed based on real-time data updating of the continuous image sequence.

[0073] In this step, the offset refers to the distance of the object position deviating from the expected trajectory. The path correction instruction refers to the control signal for adjusting the motion route of the device. The moving track refers to the motion path of the device in space. The target supervision closed loop refers to a complete monitoring and processing flow system.

[0074] In the embodiments of the present application, first, the real-time temperature distribution difference between the supervision target and the surrounding environment is calculated according to the thermal infrared image, and the position offset of the target is detected by a target tracking algorithm. Second, the offset is input into a path planning model to dynamically correct the moving track of the collection device: when the target deviates to the left of the predicted path, the control device turns right to reduce the interception distance. Then, based on the real-time data updating of the continuous image sequence, the PID control algorithm is used to adjust the propeller speed and the angle of the deflector to optimize the guiding efficiency of the surface water flow. Finally, when the supervision target enters the capture range of the storage bin, the closing mechanism is triggered to complete a single interception task, and the target supervision closed loop is formed through image feedback verification.

[0075] In summary, steps 101 to 105 realize intelligent closed-loop supervision of marine plastic garbage. As shown in Figure 2 by fusing visible light and thermal infrared image data, the contour features and temperature distribution difference of plastic garbage are accurately identified, the collection device is dynamically arranged in combination with the current trajectory prediction, and the plastic garbage is guided by the surface water flow. Further, by filtering the difficult-to-decompose plastic garbage as the supervision target based on the reflection characteristic difference, and dynamically correcting the collection path based on the real-time temperature distribution difference, a complete supervision closed loop from identification, tracking, aggregation to preferential interception is formed. This technology effectively improves the automation level of marine plastic garbage supervision, solves the problems of inaccurate target identification and low interception efficiency in traditional methods, and enhances the targeted processing capability of difficult-to-decompose plastics through dynamic path correction.

[0076] In some embodiments, as described in step 105, based on the temperature distribution difference of the supervision target in the dynamic aggregation area, the offset of the supervision target is calculated, and the path correction instruction is generated according to the offset to correct the moving track of the collection device, so that the collection device preferentially intercepts the difficult-to-decompose plastic garbage with the corrected moving track, and a target supervision closed loop is completed based on real-time data updating of the continuous image sequence, including:

[0077] 201、acquire the position coordinates of the supervision target in the thermal infrared image where the temperature distribution difference is greater than a preset value, and acquire the current position coordinates of the collection device, according to the temperature distribution difference of the supervision target in the dynamic aggregation area;

[0078] In step 201, the temperature distribution difference refers to the temperature variation characteristics of different regions of the supervision target in the thermal infrared image. The preset value refers to the preset temperature difference threshold for identifying significant temperature variation regions. The position coordinates refer to the precise positioning data of the supervision target in two-dimensional or three-dimensional space. The current position coordinates refer to the real-time spatial position information of the collection device.

[0079] In the embodiments of the present application, first, in the thermal infrared image of the dynamic aggregation area, the pixel regions with temperature distribution difference exceeding the preset value are filtered out by threshold segmentation algorithm, and the center point coordinates of these regions are extracted as the real-time position of the supervision target. Secondly, the current latitude and longitude coordinates of the collection device are acquired through the GPS module or wireless positioning beacon, and are converted into a plane coordinate system matching the thermal infrared image. Then, the two types of coordinates are input into the spatial alignment module to ensure that the position data of the supervision target and the collection device are in the same reference system. Finally, the coordinate difference matrix of the target position and the device position is recorded, providing basic data for the deviation calculation.

[0080] 202、perform spatial comparison between the position coordinates of the supervision target and the current position coordinates of the collection device to calculate the moving direction deviation angle and the deviation distance of the supervision target relative to the collection device;

[0081] In step 202, spatial comparison refers to the process of comparing and analyzing the position data of the supervision target and the collection device. The moving direction deviation angle refers to the deflection angle of the supervision target relative to the motion direction of the collection device. The deviation distance refers to the shortest spatial interval distance between the supervision target and the collection device.

[0082] In the embodiments of the present application, first, the position coordinates of the supervision target and the current position coordinates of the collection device are input into the vector analysis model to calculate the straight-line distance between the two points as the deviation distance. Secondly, the direction angle of the vector is calculated to determine the orientation of the supervision target relative to the collection device, and the difference between the current heading angle of the device is the moving direction deviation angle. Then, the deviation angle is quadrant corrected: negative value when the target is on the left side of the device, positive value when the target is on the right side of the device. Finally, the deviation parameter table containing the deviation angle, distance and relative position quadrant is generated for path correction.

[0083] 203、calculate the lateral offset and longitudinal offset of the supervision target in the moving direction of the collection device based on the moving direction deviation angle and the deviation distance, generate a path correction instruction combining the superposition relationship of the lateral offset and the longitudinal offset, and the path correction instruction contains amplitude parameters and direction parameters;

[0084] In step 203, the lateral offset refers to the deviation of the monitored target perpendicular to the direction of motion of the collection device. The longitudinal offset refers to the deviation of the monitored target along the direction of motion of the collection device. The superposition relationship refers to the interaction and combined influence between the lateral and longitudinal offsets. The amplitude parameter is a quantitative indicator of the adjustment amount in the path correction instruction. The direction parameter is the direction of the adjustment in the path correction instruction.

[0085] In an embodiment of the present application, first, based on the deviation angle of the moving direction, the deviation distance is decomposed into a lateral offset perpendicular to the device heading and a longitudinal offset along the heading. Secondly, the correction direction is determined according to the positive or negative value of the lateral offset (left or right deviation), and the acceleration or deceleration instruction is set according to the size of the longitudinal offset (leading or lagging). Then, a weight distribution strategy is adopted: the lateral offset weight is set as the priority, and the longitudinal offset is used as an auxiliary parameter to generate a path correction instruction containing an amplitude parameter (speed adjustment ratio) and a direction parameter (left / right turn, acceleration / deceleration). Finally, the instruction is encoded into a control signal format and transmitted to the power system of the collection device.

[0086] 204. Adjust the rotational speed ratio of the propellers of the collection device according to the amplitude parameter and the direction parameter of the path correction instruction, so that the propellers in the direction corresponding to the lateral offset generate reverse compensatory thrust, and the propeller thrust output in the direction corresponding to the longitudinal offset matches a preset ratio, so as to correct the movement trajectory of the collection device;

[0087] In step 204, the rotational speed ratio refers to the ratio of the rotational speeds of the different propellers in the collection device. The reverse compensating thrust refers to the reverse force applied in the corresponding direction to offset lateral offset. The thrust output refers to the amount of propulsive force generated by the propellers. The preset ratio refers to the preset thrust distribution scheme for each propeller.

[0088] In the embodiment of the present application, first, the amplitude parameter and direction parameter in the path correction instruction are parsed, and the correction requirement corresponding to the lateral offset is converted into an adjustment strategy for the ratio of the left and right propeller speeds: when the lateral offset shows that the supervision target is located on the left side of the collection device, the right propeller speed is increased and the left propeller speed is reduced to generate a reverse compensation thrust to offset the offset; when the offset is on the right, the operation is reversed. Secondly, according to the size of the longitudinal offset, the basic speed of the two propellers is adjusted according to a preset ratio - if the supervision target is far ahead of the device's heading, the speed of the two propellers is simultaneously increased to enhance the longitudinal thrust; if the target is close, the speed is reduced to prevent overshoot. Then, the speed difference between the left and right propellers is controlled by a dynamic balancing algorithm to ensure that the compensation amount of the lateral thrust accurately matches the preset ratio of the longitudinal thrust. Finally, the adjusted propeller power output is applied to the heading control system of the collection device, gradually reducing the actual position deviation between the supervision target and the device, and completing the real-time correction of the moving trajectory.

[0089] 205. The collecting device is used to preferentially intercept the difficult-to-decompose plastic waste with the corrected moving trajectory, and the moving trajectory of the collecting device is continuously adjusted based on the real-time data of the continuous image sequence to complete the closed-loop interception of the difficult-to-decompose plastic waste.

[0090] In step 205, the corrected trajectory refers to the collection device's motion path after being adjusted using the path correction instructions. Closed-loop interception refers to the complete process from identification to capture through continuous monitoring and adjustment. Real-time data refers to the instantly updated monitoring information in the continuous image sequence. Continuous adjustment refers to the continuous optimization of the collection device's motion trajectory based on the latest data.

[0091] In an embodiment of the present application, first, the collection device is controlled to approach the supervision target according to the corrected movement trajectory, and the visible light image is used to identify whether the target has entered the capture area in front of the device. Secondly, when the target enters the capture range, the diversion baffle is activated to enhance the convergence intensity of the surface water flow, guiding the difficult-to-decompose plastic waste to move toward the entrance of the storage bin. Then, the target position is tracked in real time based on the continuous image sequence. If an offset is detected, steps 201-204 are repeated for dynamic fine-tuning. Finally, when the target completely enters the storage bin, the door is closed, the interception success signal is uploaded, and the supervision list is updated, forming a complete process of closed-loop interception.

[0092] Here's a specific example:

[0093] In the offshore floating garbage management scenario, an intelligent monitoring system implements dynamic interception through the linkage of a thermal unmanned aerial vehicle and a water surface cleaning ship. The system identifies the dynamic aggregation area of difficult-to-decompose plastic garbage affected by the convergence of ocean currents based on thermal infrared images (step 201), extracts the patch-like position coordinates of the surface water temperature distribution difference exceeding the preset threshold, and synchronously obtains the real-time current position coordinates provided by the Beidou positioning module of the cleaning ship. Through three-dimensional space coordinate system conversion (step 202), the garbage aggregation area polygon vertex coordinates and the cleaning ship coordinates are compared in space, and the moving direction deviation angle and deviation distance of the floating object under the combined action of the offshore wind and the rising tide are calculated. The system decomposes the lateral ocean current offset and the longitudinal wind force offset according to the deviation angle (step 203), generates path correction instructions containing differential control parameters of the left and right propellers of the catamaran, and the direction parameter in the instructions corresponds to the direction of the northwest-southeast composite force. According to the amplitude parameter of the correction instruction (step 204), the left propeller speed of the cleaning ship is increased to generate a southeast counteracting thrust, and the right propeller output is controlled to match the preset anti-wind and wave parameters, so that the ship body is zigzag-shaped and cuts into the edge of the garbage patch. The corrected cleaning ship sails along the tangent direction of the convergence zone (step 205), preferentially intercepts the high-density plastic suspended layer, and dynamically adjusts the heading angle based on the continuous thermal image sequence returned by the unmanned aerial vehicle in real time, finally forms a spiral closing interception trajectory, makes the capture completeness rate of patch-shaped pollutants more than twice that of the traditional grid cruising mode, and realizes precise cooperative control of thermodynamic characteristics and fluid dynamics.

[0094] In summary, steps 201 to 205 realize precise dynamic correction of the moving trajectory of the collection device. The relative position of the monitoring target and the collection device is located by the temperature distribution difference in the thermal infrared image, the deviation angle and distance are calculated and decomposed into lateral and longitudinal components, and the correction instruction containing the amplitude and direction parameters is generated. Based on the proportional adjustment of the propeller speed, the counteracting thrust and the longitudinal thrust are matched, so that the collection device can respond to the target offset change in real time and preferentially intercept difficult-to-decompose plastic garbage. This technology breaks through the limitations of traditional fixed path interception, significantly improves the interception accuracy through real-time data-driven closed-loop control, ensures that the collection device always dynamically tracks the target with the optimal path, and greatly improves the recycling efficiency of difficult-to-decompose plastics.

[0095] In some embodiments, based on the moving direction deviation angle and deviation distance, the lateral offset and longitudinal offset of the monitoring target in the moving direction of the collection device are calculated in step 203, and the path correction instruction is generated based on the superposition relationship of the lateral offset and longitudinal offset. The path correction instruction contains amplitude parameters and direction parameters, including:

[0096] 301、based on the numerical relationship between the moving direction deviation angle and the deviation distance, the deviation distance is decomposed into a transverse offset and a longitudinal offset of the monitoring target in the moving direction of the collection device, the transverse offset is the component of the deviation distance perpendicular to the moving direction of the collection device, and the longitudinal offset is the same direction component of the deviation distance in the moving direction of the collection device;

[0097] In step 301, the moving direction deviation angle refers to the included angle between the motion direction of the monitoring target and the motion direction of the collection device. The deviation distance refers to the spatial interval distance between the monitoring target and the collection device. The transverse offset refers to the component of the deviation distance perpendicular to the moving direction of the collection device. The longitudinal offset refers to the same direction component of the deviation distance in the moving direction of the collection device. The numerical relationship refers to the quantitative corresponding relationship between the parameters.

[0098] In the embodiments of the present application, first, based on the numerical value of the moving direction deviation angle, the deviation distance is decomposed into two components by vector decomposition technique: the transverse offset perpendicular to the moving direction of the collection device, and the longitudinal offset along the moving direction. Second, by establishing a coordinate system with the current heading of the collection device as the reference axis, the positive and negative values of the transverse offset (left deviation is negative and right deviation is positive) and the relative length of the longitudinal offset are calculated using the trigonometric function relationship. Then, according to the decomposition result, the offset component distribution map is generated, and the direction attribute of the transverse offset and the distance attribute of the longitudinal offset are marked. Finally, the decomposed offset data is packaged into a structured parameter set for subsequent correction instruction generation.

[0099] 302、generate a direction correction instruction according to the relative relationship between the transverse offset and the longitudinal offset, and generate an amplitude correction instruction of a path correction instruction according to the overall offset degree of the transverse offset and the longitudinal offset;

[0100] In step 302, the direction correction instruction refers to the instruction parameter for adjusting the motion direction of the collection device. The amplitude correction instruction refers to the instruction parameter for adjusting the motion intensity of the collection device. The relative relationship refers to the proportional relationship between the transverse and longitudinal offsets. The overall offset degree refers to the overall size of the comprehensive offset. The path correction instruction refers to the composite instruction for correcting the motion trajectory of the collection device.

[0101] In the embodiments of the present application, first, the direction attribute of the lateral offset is analyzed: if it is a negative value (left deviation), a "turn right" direction correction instruction is generated; if it is a positive value (right deviation), a "turn left" instruction is generated. Second, according to the absolute value of the longitudinal offset, the strength level of the amplitude correction instruction is set - the larger the offset, the higher the amplitude adjustment ratio. Third, the priority rule is introduced: when the lateral offset and the longitudinal offset exist at the same time, the direction correction of the lateral offset is responded first, and the amplitude correction of the longitudinal offset is superimposed at the same time. Finally, the direction correction instruction (left / right turn) and the amplitude correction instruction (strength level) are packaged as independent control parameter groups to ensure the logical independence of the two.

[0102] 303, fuse the direction correction instruction and the amplitude correction instruction to obtain a path correction instruction containing an amplitude parameter and a direction parameter.

[0103] In step 303, the amplitude parameter refers to the quantitative index of the adjustment strength in the path correction instruction. The direction parameter refers to the indication parameter of the adjustment direction in the path correction instruction. Fusion refers to the process of integrating multiple instruction parameters into a unified instruction.

[0104] In the embodiments of the present application, first, the direction correction instruction and the amplitude correction instruction input parameters are fused into a module, the direction parameter (left / right turn) is converted into a propeller differential control signal through an instruction encoder, and the amplitude parameter (strength level) is mapped into a rotation speed adjustment ratio. Second, the weight superposition rule is designed: the direction correction occupies the dominant weight, and the amplitude correction is used as an auxiliary parameter, and the two are proportionally synthesized into the final control quantity. Third, the rationality of the instruction is verified through the verification logic, for example, the left turn instruction cannot conflict with the right deviation amplitude parameter. Finally, the path correction instruction containing the complete amplitude parameter and the direction parameter is output, and is converted into an executable instruction format of the power system of the collection device.

[0105] The following is a specific example:

[0106] In the scenario of plastic pollution control in mangrove coastline, an intelligent system implements dynamic interception through the linkage of thermal unmanned aerial vehicle and catamaran collection ship. The system identifies the dynamic aggregation area of plastic garbage affected by backflow at the entrance of tidal creek based on thermal infrared image, and obtains the patch-like position coordinates formed by the temperature distribution difference (step 301). The system synchronously receives the real-time coordinates provided by the collection ship positioning system. Through spatial vector decomposition, the deviation distance of the garbage patch center point relative to the collection ship is decomposed into lateral deviation (north-east tidal component perpendicular to the ship heading) and longitudinal deviation (tidal flow pushing component along the heading). According to the condition that the lateral deviation exceeds the channel safety threshold, the direction correction instruction is generated (step 302), and at the same time, the amplitude correction instruction is generated based on the lag amplitude of the longitudinal deviation, requiring to increase the power level of the main propeller. The deflection angle parameter of the direction correction instruction and the thrust gain parameter of the amplitude correction instruction are fused (step 303) to generate a composite path correction instruction containing differential control and power adjustment of the left and right propellers of the catamaran, which controls the right propeller to increase the speed to generate southwest compensating thrust, and increases the left propeller output by a preset proportion, so that the collection ship approaches the garbage patch core area in a curved cutting trajectory, and finally forms a closed loop interception path around the pollution source, which improves the capture completeness rate of fragmented plastics compared with the traditional straight line cruising mode, and realizes the precise matching of fluid dynamics deviation compensation and equipment control parameters.

[0107] In summary, steps 301 to 303 realize the decomposition of deviation and the fine generation of correction instructions. By decomposing the deviation distance into lateral and longitudinal components, the spatial relationship of the target deviation is determined, and the direction and amplitude correction instructions are generated according to the relative deviation degree. This technology solves the coupling problem of target deviation calculation in complex marine environment, realizes the precise quantization of deviation correction through component decomposition, and makes the path correction instruction consider direction correction and movement amplitude adjustment at the same time. This hierarchical processing mechanism effectively improves the scientificity and operability of the correction instruction, provides a reliable basis for the precise control of propeller thrust distribution, and ensures the stability and response speed of the collection device movement trajectory correction.

[0108] In some embodiments, according to the amplitude parameter and direction parameter of the path correction instruction, the speed ratio of the propeller of the collection device is adjusted to generate a reverse compensating thrust in the direction corresponding to the lateral deviation, and the propeller thrust output in the direction corresponding to the longitudinal deviation is matched with a preset proportion, so as to correct the movement trajectory of the collection device (step 204), including:

[0109] 401. Determine the thrust intensity of the propeller according to the amplitude parameter of the path correction instruction, the thrust intensity corresponding to the numerical value of the amplitude parameter;

[0110] In step 401, the thrust intensity refers to the propulsion force generated by the propeller. The numerical value corresponds to the proportional matching relationship between parameters. Determination refers to the process of obtaining the target parameter through calculation.

[0111] In the embodiments of the present application, first, the amplitude parameter in the path correction instruction is analyzed and mapped to the grade division of the thrust intensity of the propeller: the larger the amplitude parameter, the higher the corresponding thrust intensity grade. Second, the amplitude parameter is converted into the target speed value of the propeller motor by table lookup method, establishing a linear correspondence between the parameter value and the speed. Then, according to the real-time speed of the current propeller, the proportional control algorithm is used to gradually adjust to the target speed, ensuring smooth transition of the thrust intensity. Finally, the thrust intensity calibration result is fed back to the control center to verify the consistency of the actual output and the instruction requirement.

[0112] 402. Adjust the thrust direction of the propeller on the side corresponding to the lateral offset according to the direction parameter of the path correction instruction to the reverse compensation thrust, the reverse compensation thrust being opposite to the deviation angle of the moving direction to offset the influence of the lateral offset on the moving trajectory;

[0113] In step 402, the thrust direction refers to the action direction of the thrust generated by the propeller. The reverse compensation thrust refers to the reverse force used to offset the deviation. Offset refers to the process of eliminating the influence by reverse action. Influence refers to the interference of the deviation on the motion trajectory.

[0114] In the embodiments of the present application, first, the lateral deviation direction is identified according to the direction parameter of the path correction instruction: if the direction parameter indicates left deviation, the right propeller is positioned as the side corresponding to the lateral offset. Second, the reverse compensation thrust opposite to the deviation angle of the moving direction is generated by reversing the thrust direction of the propeller on this side or increasing the reverse speed. Then, the effect of the compensation thrust is monitored in real time, and if the lateral offset has not been significantly reduced, the reverse thrust intensity is increased in proportion. Finally, the trajectory change data after thrust adjustment is recorded to provide a reference for longitudinal correction.

[0115] 403. Adjust the thrust intensity of the propeller in the direction corresponding to the longitudinal offset according to the preset proportion, so that the thrust output of the propeller in the direction corresponding to the longitudinal offset matches the amplitude parameter, to correct the longitudinal offset;

[0116] In step 403, the preset proportion refers to the pre-set thrust distribution scheme. Matching refers to the state of achieving coordination between parameters. Correction refers to the process of eliminating deviation by adjustment.

[0117] In the embodiments of the present application, first, based on the amplitude parameter of the path correction instruction, the reference value of the thrust in the direction corresponding to the longitudinal offset is determined, and the reference value is distributed to the front and rear propellers according to a preset proportion. Second, if the longitudinal offset shows that the target is far in front of the device, the thrust intensity of the double propellers is simultaneously increased to accelerate the approach; if the target has approached, the thrust intensity is reduced to prevent overshoot. Then, the differential control algorithm is used to dynamically adjust the change rate of the thrust output, so as to ensure the stability of the longitudinal correction process. Finally, the longitudinal thrust adjustment result is superimposed with the lateral compensation thrust to form a combined thrust vector acting on the device heading.

[0118] 404, repeat the adjustment process of the propeller thrust direction and the propeller thrust intensity to correct the moving track of the collection device.

[0119] In step 404, the adjustment process is repeated. The adjustment process refers to the operation process of optimizing the thrust parameters. The correction is to make the track tend to be ideal by continuous adjustment.

[0120] In the embodiments of the present application, first, the actual moving track data of the collection device is continuously collected after initial adjustment, and the residual deviation from the expected path is calculated. Second, if the residual deviation exceeds the threshold, the adjustment process of the propeller thrust direction and the propeller thrust intensity is retriggered: the path correction instruction is updated according to the latest deviation value, and steps 401 to 403 are iteratively executed. Then, the incremental adjustment strategy is used to gradually reduce the deviation, so as to avoid track oscillation caused by large correction. Finally, when the consistency of the actual track and the expected path reaches the preset standard, the adjustment process is terminated and the current thrust parameters are locked, and the final correction of the moving track is completed.

[0121] The following is a specific example:

[0122] In the scene of dynamic interception of plastic pollution in the bay, an intelligent system cooperates with a three-body collection ship through a thermal unmanned aerial vehicle. After the system receives the amplitude parameter of the path correction instruction (step 401), the main propeller thrust intensity is increased according to the parameter level, so that the basic speed of the ship body adapts to the strength of the ocean current. According to the northeast deviation trend indicated by the direction parameter in the instruction (step 402), the left auxiliary propeller generates a southwest counteracting thrust to accurately offset the traction of the tidal lateral offset on the heading. At the same time, the right main propeller thrust intensity is adjusted to the set level according to the preset longitudinal anti-flow proportion (step 403), so that the propulsion force of the ship head direction and the longitudinal offset of the ebb tide form a dynamic balance. The system continuously receives the updated displacement data of the plastic garbage belt from the unmanned aerial vehicle (step 404), and the propeller thrust direction and intensity adjustment mechanism is triggered in a loop, so that the three-body ship maintains a zigzag correction track in a strong tidal current environment, and finally forms a gradually converging interception loop along the edge of the garbage accumulation belt. Compared with the traditional single correction mode, the interception efficiency is improved, and closed-loop path tracking in a fluid interference environment is realized.

[0123] In summary, steps 401 to 404 achieve dynamic collaborative control of propeller thrust. The thrust intensity is determined by the amplitude parameter, the direction parameter adjusts the reverse compensation thrust, and the preset proportional matching of the longitudinal thrust forms a multi-dimensional thrust distribution strategy. This technology breaks through the limitations of traditional single thrust adjustment, and through the synergistic effect of lateral reverse compensation and longitudinal thrust optimization, significantly improves the flexibility and accuracy of trajectory correction of the collection device. The dynamic adjustment mechanism of propeller thrust can quickly offset the influence of lateral deviation, while maintaining the longitudinal propulsion efficiency, ensuring that the collection device can still track the target trajectory stably in complex sea conditions, greatly enhancing the sustained interception capability of difficult-to-decompose plastic waste.

[0124] In some embodiments, as described in step 102, the contour boundary of the plastic waste is segmented according to the contour features, and the position change of the contour boundary in the continuous image sequence is obtained to generate a trajectory prediction path of the plastic waste moving on the sea surface with the ocean current, including:

[0125] 501. Segmenting the contour boundary of the plastic waste according to the shape closure and edge continuity of the contour features;

[0126] In step 501, the contour feature refers to the shape feature of the plastic waste in the image. Shape closure refers to the characteristic of whether the contour boundary forms a complete closed loop. Edge continuity refers to the degree of continuity of the contour edge. Segmentation refers to the process of extracting a specific region from the image. The contour boundary refers to the geometric features of the plastic waste shape edge.

[0127] In the embodiments of the present application, first, the visible light image is preprocessed, the morphological processing technology is used to enhance the shape closure of the contour features, and the broken contour line segments are connected through the edge continuity verification algorithm. Second, the edge tracking is performed on the region that meets the closure condition, the complete contour boundary is marked, and the internal holes are eliminated through the region filling algorithm. Then, the segmentation result is post-processed, the noise interference region with too small area is removed, and the closed contour meeting the size characteristics of the plastic waste is retained. Finally, the contour boundary coordinates are converted into vector polygon data to provide structured input for displacement tracking.

[0128] 502. Tracking the position change of the same contour boundary in the continuous image sequence, and obtaining the movement direction and displacement of the contour boundary according to the position change;

[0129] In step 502, tracking refers to the process of following the target motion in continuous images. Position change refers to the difference in spatial coordinates of the target at different time points. Position change refers to the amount of change of the target position over time. Movement direction refers to the azimuth angle of the target trajectory. Displacement refers to the distance value of the target movement.

[0130] In the embodiments of the present application, first, in the continuous image sequence, the corner points or texture features of the same contour boundary are tracked by a feature point matching algorithm, and the position offset in the adjacent frame is recorded. Second, the average displacement vector of all feature points is calculated to determine the overall movement direction of the contour boundary, and the actual displacement amount is obtained by pixel distance conversion. Then, the abnormal displacement data caused by wave shaking is filtered out, and the effective displacement record consistent with the current of the trend is retained. Finally, a trajectory tracking dataset containing time stamp, direction angle and displacement length is generated.

[0131] 503, obtaining a current of the movement trend, the current of the movement trend being used to correct the movement direction of the contour boundary by a preset correction ratio to generate a predicted movement direction of the contour boundary;

[0132] In step 503, the current of the movement trend refers to the main direction of the large-scale flow of seawater. The preset correction ratio refers to the pre-set adjustment coefficient. The correction generation refers to the process of adjusting according to the reference data. The predicted movement direction refers to the expected movement direction considering the influence of the current.

[0133] In the embodiments of the present application, first, the current of the movement trend data of the current sea area is obtained from the ocean monitoring platform, including the main direction of the flow velocity and the periodic fluctuation range. Second, the angle between the original movement direction of the contour boundary and the direction of the current trend is calculated, and the direction angle is adjusted by a preset correction ratio: if the current trend is east and the contour movement is north, the movement direction is corrected east by a certain ratio. Then, a predicted movement direction considering the measured displacement and the influence of the current is generated by a weighted fusion algorithm. Finally, the corrected direction parameter and the confidence evaluation result are output.

[0134] 504, generating a trajectory prediction path of the plastic garbage under the action of the current according to the predicted movement direction, the trajectory prediction path covering a sea area range determined by the displacement amount of the contour boundary.

[0135] In step 504, the action of the current refers to the influence of the flow of seawater on the floating object. The trajectory prediction path refers to the expected route of the future movement of the plastic garbage. The sea area range refers to the spatial area covered by the prediction path. The determination refers to the process of calculating the target parameter.

[0136] In the embodiments of the present application, first, based on the predicted movement direction and the cumulative displacement amount, the extrapolation method is used to calculate the possible movement range of the plastic garbage in the future period, and the trajectory prediction path covering the sea area is generated. Second, the confidence bandwidth of the path is set according to the standard deviation of the displacement amount: the greater the fluctuation of the displacement amount, the wider the coverage range of the path. Then, the prediction path is encoded into a geofence polygon, and the high-risk aggregation area is labeled. Finally, the path and the covered area are rendered by a visualization engine and superimposed on the electronic chart for reference of interception operation.

[0137] The following is a specific example:

[0138] In the scenario of plastic pollution monitoring in coral reef waters, an intelligent system cooperates with a hyperspectral unmanned aerial vehicle and an array of ocean buoys to implement dynamic trajectory prediction. Based on multispectral images (step 501), the system segments the outline boundary with a closed polygonal structure according to the spectral reflectance difference and edge continuity features of the plastic garbage ring-shaped accumulation zone. Through continuous image sequence analysis (step 502), the periodic position changes of the outline boundary caused by surface ocean currents are tracked, and the overall movement trend and relative displacement of the outline boundary to the southeast are accurately analyzed. Combined with the movement trend of the ocean currents provided by the regional ocean dynamics model (step 503), a preset fluid correction coefficient is used to compensate for the deflection angle of the original movement direction, and the adjusted northeastward predicted movement direction under the influence of the monsoon circulation is generated. Based on the corrected predicted movement direction (step 504), the trajectory prediction path of the plastic garbage affected by the vortex flow is simulated, and the coverage sea area range is determined by the elliptical diffusion area formed by the displacement of the outline boundary, so that the cleaning fleet can be arranged in advance on the predicted path to set up an interception barrier in the wind direction, which can improve the operation efficiency by two times compared with the passive patrol mode, and realize the spatiotemporal precise matching of pollution diffusion trend and cleaning resources.

[0139] In summary, steps 501 to 505 realize high-precision prediction of the motion trajectory of plastic garbage. By segmenting the target boundary based on the closed and continuous features of the outline, combining displacement tracking in continuous image sequences and ocean current trend correction, and generating a trajectory prediction path covering the sea area range, the technology solves the problem of insufficient consideration of environmental factors in traditional trajectory prediction, significantly improves the accuracy of trajectory prediction by integrating target displacement data and ocean current dynamic correction. The sea area range defined based on the predicted path provides a scientific basis for the arrangement of collection devices, ensures the coordination of surface flow guidance direction and ocean current action, lays a data foundation for the construction of subsequent dynamic accumulation areas, and greatly improves the plastic garbage accumulation efficiency.

[0140] In some embodiments, as described in step 103, a buoy-type collection device is arranged in the sea area range covered by the trajectory prediction path, the collection device causes the propeller to generate a surface flow converging towards the storage bin of the collection device, and the plastic garbage is guided to move along the direction of the trajectory prediction path by the surface flow to form a dynamic accumulation area, including:

[0141] 601、In the sea area range covered by the trajectory prediction path, a buoy-type collection device is arranged, the storage bin of the collection device is oriented towards the extension direction of the trajectory prediction path, and the propeller of the collection device is started to rotate towards the storage bin to generate a surface flow converging towards the storage bin;

[0142] In step 601, the buoy-type collection device refers to the garbage collection equipment floating on the sea surface. The storage bin refers to the container in the collection device for storing garbage. The extension direction refers to the spatial trend of the predicted path. The propeller rotation refers to the mechanical movement of generating thrust by rotating the blade. The surface water flow refers to the flow generated by the propeller affecting the surface layer of seawater.

[0143] In the embodiments of the present application, first, in the sea area covered by the trajectory prediction path, the layout point of the buoy-type collection device is determined according to the extension direction of the predicted path, so as to ensure that the opening of the storage bin of the device is directly opposite to the extension direction of the path. Second, the propeller driving system of the collection device is started, and the propeller is controlled to rotate at a preset rotating speed to the direction of the storage bin. The centrifugal force generated by the rotation of the blade forms the surface water flow converging to the bin. Then, the convergence strength and coverage range of the surface water flow are verified by the water flow sensor, so as to ensure that the initial water flow can effectively guide the plastic garbage to the device. Finally, the operation parameters of the propeller are locked and the continuous operation mode is started, so as to provide the basic power conditions for subsequent path alignment adjustment.

[0144] 602, adjust the rotating direction and thrust distribution of the propeller, so that the flow direction of the surface water flow is consistent with the extension direction of the trajectory prediction path;

[0145] In step 602, the rotating direction refers to the clockwise or counterclockwise direction of the propeller rotation. The thrust distribution refers to the distribution of the propelling force generated by different propellers. The flow direction refers to the main direction of the flow of the surface layer of seawater. Consistent refers to the state of coordination and unity of different parameters.

[0146] In the embodiments of the present application, first, based on the extension direction of the trajectory prediction path, the rotating direction of the propeller is adjusted: if the path direction is slightly left, the rotating speed of the left propeller is slightly increased to correct the flow direction; if the path direction is slightly right, the rotating speed of the right propeller is slightly increased to correct the flow direction. Second, the thrust distribution of the left and right propellers is dynamically adjusted by the thrust distribution algorithm, so that the overall flow direction of the surface water flow is completely consistent with the path direction. Then, the coverage range of the adjusted water flow is predicted by using the water flow trajectory simulation software. If it is found that the local area flow direction deviates, the propeller rotating speed difference is further optimized to eliminate the deviation. Finally, the dynamic matching between the water flow direction and the path is maintained through the real-time feedback mechanism, so as to ensure that the convergence path of the plastic garbage does not deviate from the predicted trajectory.

[0147] 603, guide the plastic garbage in the sea area covered by the trajectory prediction path to gather along the extension direction of the trajectory prediction path to form a dynamic aggregation area through the surface water flow.

[0148] In step 603, guiding refers to the process of changing the motion state of an object by external force. Gathering refers to the phenomenon of making dispersed objects converge to the center. Dynamic aggregation area refers to the temporary area formed by the aggregation of garbage with water flow. Forming refers to the process of generating a target state under certain conditions.

[0149] In the embodiments of the present application, first, the adjusted surface water flow forms a directional flow effect in the sea area covered by the trajectory prediction path, guiding the dispersed plastic waste to move along the path extension direction to the central area. Second, the plastic waste is gradually gathered by the water flow convergence effect, forming a high-density dynamic aggregation area, and the expansion range of the aggregation area is monitored in real time through visible light images. Then, according to the monitoring results, the rotation speed and thrust distribution of the propeller are dynamically optimized: when the density of the aggregation area is insufficient, the water flow intensity is increased, and when the density is saturated, the water flow intensity is reduced to prevent the waste from overflowing. Finally, when the aggregation area stably covers the preset range, the storage bin closing instruction is triggered to complete a single aggregation task and prepare for the next cycle.

[0150] The following is a specific example:

[0151] In the scenario of mangrove tidal channel plastic pollution control, an intelligent system cooperates with a buoy array to implement dynamic interception through a prediction model. The system arranges the buoy collector devices along the northeast predicted path at equal intervals within the sea area covered by the trajectory prediction path (step 601), with the funnel-shaped storage bin opening facing the trajectory extension direction, and the bottom propeller of the starting device rotates in the opposite direction to generate surface water flow converging into the storage bin. According to the ocean current monitoring data, the rotation direction of the propeller is adjusted in real time (step 602) to make the laminar direction of the surface water flow accurately match the extension direction of the trajectory prediction path adjusted by the monsoon, and to optimize the thrust distribution of the propeller to offset the lateral tidal interference. Through the continuous guidance of the directional surface water flow (step 603), the fragmented plastic waste scattered in the mangrove waterway is gathered along the predicted path to the buoy array, forming a strip-shaped dynamic aggregation area, which makes the collection efficiency of the collector device three times higher than the random arrangement mode, realizing the closed-loop cooperative control of the predicted path and water power guidance.

[0152] In summary, steps 601 to 603 achieve dynamic and efficient aggregation of plastic waste. By arranging the buoy collector devices and directing the propeller, a surface converging water flow consistent with the trajectory prediction path is generated, guiding the plastic waste to move along the predetermined direction to form an aggregation area. This technology breaks through the efficiency bottleneck of the traditional passive collection mode, realizes the directional aggregation of waste through active water flow guidance, and solves the problem of dispersed plastic waste collection. The precise regulation of the rotation direction and thrust distribution of the propeller ensures the high coordination of the surface water flow and the predicted path, significantly improving the formation speed and scale of the dynamic aggregation area, creating favorable conditions for subsequent target sorting and preferential interception.

[0153] In some embodiments, in step 104, in the dynamic aggregation area, the difference in the reflection characteristics of the plastic waste under visible light in the continuous image sequence is analyzed, and the plastic waste with the preset degradation property is identified and marked as a supervision target according to the difference in the reflection characteristics, including:

[0154] 701、extracting a reflection intensity value of a surface of the plastic garbage under visible light in the continuous image sequence in the dynamic gathering area, different materials of the plastic garbage causing different reflection intensity values of the surface of the plastic garbage under visible light;

[0155] In step 701, the dynamic gathering area refers to a temporary gathering area of the plastic garbage formed by the water flow. The continuous image sequence refers to a set of images taken in time sequence. The visible light refers to the range of electromagnetic spectrum that can be perceived by the human eye. The reflection intensity value refers to a quantitative index of the reflection ability of the surface of an object to visible light. The different materials refer to the difference in chemical composition of the plastic garbage.

[0156] In the embodiments of the present application, first, in the continuous image sequence of the dynamic gathering area, the pixel-level reflection spectrum data of the plastic garbage under the visible light band is extracted by using the multispectral analysis technology, and the influence of the difference in light intensity on the reflection intensity value is eliminated by normalization processing. Second, image segmentation is performed on each plastic garbage target, the boundary range of a single target is delineated by using the region growing algorithm, and the average reflection intensity value is calculated. Then, a reflection intensity distribution histogram is established, and the reflection peak interval of different materials of plastic (such as PET and PVC) is labeled. Finally, an attribute list containing the spatial position and the reflection intensity value is generated, which provides basic data for reflection characteristic comparison.

[0157] 702、comparing the reflection intensity values of different plastic garbage to obtain the reflection characteristic difference of different plastic garbage;

[0158] In step 702, comparison refers to the process of comparing and analyzing different data. The reflection characteristic difference refers to the distinguishing features of different plastic garbage in the reflection intensity. The reflection characteristic refers to the reflection performance characteristics of the surface of an object to light.

[0159] In the embodiments of the present application, first, the reflection intensity values of different plastic garbage are extracted from the attribute list generated in step 701, and the targets with similar reflection values are classified into the same material group by using clustering analysis algorithm. Second, the mean and variance of the reflection intensity of each group are calculated to quantify the difference between the groups and form a reflection characteristic difference matrix. Then, the reflection characteristic band with the highest contribution to material differentiation is identified by principal component analysis. Finally, a decision tree model of the reflection characteristic difference is constructed to divide the reflection threshold boundary of different material types.

[0160] 703、according to the reflection characteristic difference, comparing the reflection intensity value with the reflection characteristic corresponding to the preset degradation attribute, the reflection characteristic being the reflection intensity value, and screening out the plastic garbage with the reflection intensity value matching the degradation attribute of the difficult-to-decompose plastic;

[0161] In step 703, the preset degradation attribute refers to a pre-set material decomposition characteristic standard. The reflectance characteristic refers to the reflectance intensity pattern exhibited by a specific material. Screening refers to the process of selecting specific objects based on criteria. Refractory plastics refer to plastic materials that are difficult to degrade in the natural environment. Matching refers to the process of determining whether the object's characteristics match the standard.

[0162] In an embodiment of the present application, first, a predefined degradation property database is loaded, which contains standard reflection characteristics of difficult-to-decompose plastics such as polyethylene and polypropylene (such as the reflection intensity threshold of a specific band). Secondly, the reflection characteristic difference data obtained in step 702 is compared with the database item by item: if the reflection intensity of a target is within the characteristic interval of difficult-to-decompose plastics, it is marked as a candidate target. Then, the matching degree between the measured reflection value and the standard feature is calculated through a similarity scoring algorithm (such as cosine similarity), and the targets with scores higher than the threshold are screened. Finally, a list of coordinates of successfully matched difficult-to-decompose plastic garbage is output, and its degradation property classification label is marked.

[0163] 704. The matched plastic waste is regarded as difficult-to-decompose plastic waste, and the difficult-to-decompose plastic waste is marked as a regulatory target.

[0164] In step 704, the "regulatory target" refers to a specific waste object that requires focused treatment. "Tag" refers to adding identification information to identify the target object. "Refractory plastic waste" refers to plastic waste that meets the criteria for being difficult to decompose. "Tack" refers to the operation of classifying an object into a specific category.

[0165] In an embodiment of the present application, first, the difficult-to-decompose plastic waste targets screened out in step 703 are spatially superimposed on the electronic nautical chart, and the real-time distribution area of ​​the regulatory target is delineated by geo-fencing technology. Secondly, a visual mark (such as a red border) is added to each target in the visible light image, and the target tracking list of the regulatory platform is updated synchronously. Then, the regulatory priority is set according to the matching degree between the target size and the reflective characteristics: targets with high matching degree and large size are listed as first-level regulatory objects. Finally, the location, attributes and priority data of the regulatory target are pushed to the control system of the collection device, and the driving device performs a directional interception task.

[0166] Here's a specific example:

[0167] In the scenario of plastic pollution governance in coral reef areas, an intelligent system cooperates with high-resolution drones and spectral analysis buoys to implement target sorting. Based on continuous aerial images (step 701), the system extracts the reflection intensity value of floating plastic in the visible light range in the dynamically aggregated area of the reef disc edge. The mirror reflection of polycarbonate fragments due to smooth surface produces a significant difference with the diffuse reflection characteristics of weathered polyethylene. Through multi-dimensional spectral comparison (step 702), a difference spectrum of reflection characteristics of different materials in the visible light band is constructed, and the unique honeycomb reflection characteristics of polyurethane foam are identified. The collected reflection intensity value is matched with the pre-set ocean environment century degradation material database (step 703) to screen out difficult-to-decompose plastic degradation attribute targets with ultraviolet resistance. The fluoropolymer float ball is precisely locked due to its stable infrared reflection curve. The system labels the matched plastic garbage as a high-priority supervision target (step 704), and guides the mechanical arm of the catamaran to implement directional salvage through real-time coordinate mapping, so as to improve the targetedness of microplastic capture compared with the broad-spectrum capture mode, and realize intelligent correlation decision of optical characteristics and material durability.

[0168] In summary, steps 701 to 704 realize accurate identification and labeling of difficult-to-decompose plastic garbage. By extracting the reflection intensity difference of plastic garbage under visible light and comparing it with the pre-set degradation attribute characteristics, difficult-to-decompose targets are screened out as supervision objects. This technology overcomes the low efficiency problem of traditional manual sorting, and establishes an automatic identification mechanism through correlation analysis of optical reflection characteristics and material degradation attributes. The differential screening based on reflection intensity value ensures the accuracy of target labeling, provides reliable data support for subsequent priority interception, significantly improves the supervision targetedness of difficult-to-decompose plastic garbage, and effectively avoids the long-term retention of highly polluted plastic in the marine environment.

[0169] Figure 3 A structure diagram of an intelligent target supervision system based on data analysis is provided for the embodiments of the present application, as shown in Figure 3 The system comprises:

[0170] A capture module is configured to capture a continuous image sequence containing visible light and thermal infrared bands through a coastline-distributed monitoring camera, and identify the contour features of plastic garbage under visible light and the temperature distribution difference in the thermal infrared image in the continuous image sequence.

[0171] A segmentation module is configured to segment the contour boundary of the plastic garbage according to the contour features, and obtain the position change of the contour boundary in the continuous image sequence to generate a trajectory prediction path of the plastic garbage moving on the sea surface with ocean currents.

[0172] The deployment module is configured to deploy a buoy-type collection device in the sea area covered by the trajectory prediction path, the collection device causing a propeller to generate a surface water flow converging towards the storage bin of the collection device, and the surface water flow guiding the plastic garbage to move along the direction of the trajectory prediction path to form a dynamic aggregation area.

[0173] The analysis module is configured to analyze the difference in the reflection characteristics of the plastic garbage under visible light in the continuous image sequence in the dynamic aggregation area, and identify and mark the non-degradable plastic garbage with the preset degradation attribute as a supervision target according to the difference in the reflection characteristics.

[0174] The calculation module is configured to calculate the offset of the supervision target based on the temperature distribution difference of the supervision target in the dynamic aggregation area, and generate a path correction instruction to correct the moving trajectory of the collection device, so that the collection device preferentially intercepts the non-degradable plastic garbage along the corrected moving trajectory, and completes the target supervision closed loop based on the real-time data of the continuous image sequence.

[0175] Figure 3 The intelligent target supervision system based on data analysis can perform Figure 1 The intelligent target supervision method based on data analysis has the same implementation principle and technical effects as the above-mentioned embodiments, and will not be described in detail. The specific operation of each module and unit of the intelligent target supervision system based on data analysis in the above-mentioned embodiments has been described in detail in the embodiments related to the method, and will not be described in detail here.

[0176] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An intelligent target supervision method based on data analysis, characterized in that: include: Using surveillance cameras distributed along the coastline to capture a continuous sequence of images containing visible light and thermal infrared bands, and identifying the difference in the contour features of plastic waste in the continuous image sequence under visible light and the temperature distribution in the thermal infrared image; Segmenting the contour boundaries of the plastic waste based on the contour features, and obtaining position changes of the contour boundaries in a continuous image sequence to generate a predicted trajectory path of the plastic waste moving on the sea surface with the ocean current; Deploy a buoy-type collection device within the sea area covered by the predicted trajectory path. The propeller of the collection device generates a surface water flow that converges toward the storage bin of the collection device. The surface water flow guides the plastic waste to move along the direction of the predicted trajectory path to form a dynamic accumulation area. Within the dynamic aggregation area, analyzing the differences in the surface reflectance characteristics of the plastic waste under visible light in the continuous image sequence, and identifying difficult-to-degrade plastic waste with preset degradation properties based on the differences in the reflectance characteristics and marking it as a regulatory target; Based on the temperature distribution difference of the supervision target in the dynamic aggregation area, the offset of the supervision target is calculated, and a path correction instruction is generated according to the offset to correct the moving trajectory of the collection device, so that the collection device can preferentially intercept the difficult-to-decompose plastic waste with the corrected moving trajectory, and complete the target supervision closed loop based on the real-time data update of the continuous image sequence.

2. The method according to claim 1, characterized in that Based on the temperature distribution difference of the supervision target in the dynamic aggregation area, the offset of the supervision target is calculated, and a path correction instruction is generated according to the offset to correct the movement trajectory of the collection device, so that the collection device preferentially intercepts the difficult-to-decompose plastic waste with the corrected movement trajectory, and the target supervision closed loop is completed based on the real-time data update of the continuous image sequence, including: According to the temperature distribution difference of the surveillance target in the dynamic gathering area, the position coordinates of the surveillance target where the temperature distribution difference in the thermal infrared image is greater than a preset value are obtained, and the current position coordinates of the collection device are obtained; Comparing the position coordinates of the surveillance target with the current position coordinates of the collection device to calculate the deviation angle and deviation distance of the surveillance target relative to the movement direction of the collection device; Calculating the lateral and longitudinal offsets of the surveillance target in the moving direction of the collection device based on the deviation angle and the deviation distance of the moving direction, and generating a path correction instruction based on the superposition relationship between the lateral and longitudinal offsets, wherein the path correction instruction includes an amplitude parameter and a direction parameter; According to the amplitude parameter and direction parameter of the path correction instruction, the speed ratio of the propeller of the collection device is adjusted so that the propeller in the direction corresponding to the lateral offset generates reverse compensatory thrust and the propeller thrust output in the direction corresponding to the longitudinal offset matches a preset ratio, thereby correcting the movement trajectory of the collection device; The collecting device is used to preferentially intercept the difficult-to-decompose plastic waste with a corrected moving trajectory, and the moving trajectory of the collecting device is continuously adjusted based on the real-time data of the continuous image sequence to complete the closed-loop interception of the difficult-to-decompose plastic waste.

3. The method according to claim 2, characterized in that Based on the deviation angle and deviation distance of the moving direction, the lateral offset and longitudinal offset of the supervised target in the moving direction of the collection device are calculated, and a path correction instruction is generated based on the superposition relationship between the lateral offset and the longitudinal offset. The path correction instruction includes an amplitude parameter and a direction parameter, including: Based on the numerical relationship between the deviation angle of the moving direction and the deviation distance, the deviation distance is decomposed into the lateral offset and longitudinal offset of the surveillance target in the moving direction of the collection device. The lateral offset is the component of the deviation distance in the direction perpendicular to the moving direction of the collection device, and the longitudinal offset is the component of the deviation distance in the same direction as the moving direction of the collection device. generating a direction correction instruction according to the relative relationship between the lateral offset and the longitudinal offset, and generating an amplitude correction instruction of the path correction instruction according to the overall offset degree of the lateral offset and the longitudinal offset; The direction correction instruction and the amplitude correction instruction are combined to obtain a path correction instruction including an amplitude parameter and a direction parameter.

4. The method according to claim 2, characterized in that According to the amplitude parameter and direction parameter of the path correction instruction, the speed ratio of the propeller of the collection device is adjusted so that the propeller in the direction corresponding to the lateral offset generates reverse compensatory thrust and the propeller thrust output in the direction corresponding to the longitudinal offset matches a preset ratio to correct the movement trajectory of the collection device, including: determining a thrust intensity of the propeller according to an amplitude parameter of the path correction instruction, wherein the thrust intensity corresponds to a value of the amplitude parameter; Adjusting the propeller thrust direction on the side corresponding to the lateral offset to a reverse compensation thrust according to the direction parameter of the path correction instruction. The reverse compensation thrust is opposite to the deviation angle of the moving direction to offset the influence of the lateral offset on the moving trajectory; Adjusting the propeller thrust intensity in the direction corresponding to the longitudinal offset according to a preset ratio so that the propeller thrust output in the direction corresponding to the longitudinal offset matches the amplitude parameter to correct the longitudinal offset; The process of adjusting the propeller thrust direction and the propeller thrust intensity is repeated to correct the movement trajectory of the collection device.

5. The method according to claim 1, wherein Segmenting the contour boundary of the plastic waste according to the contour features and obtaining the position change of the contour boundary in the continuous image sequence to generate a predicted trajectory path of the plastic waste moving on the sea surface with the ocean current, including: Segmenting the contour boundary of the plastic waste according to the shape closure and edge continuity of the contour feature; Tracking the position change of the same contour boundary in the continuous image sequence, and obtaining the moving direction and displacement of the contour boundary according to the position change; Obtaining an ocean current movement trend, wherein the ocean current movement trend is used to correct the contour boundary movement direction according to a preset correction ratio to generate a predicted movement direction of the contour boundary; According to the predicted moving direction, a predicted trajectory path of the plastic waste under the influence of ocean currents is generated, and the sea area covered by the predicted trajectory path is determined by the displacement of the contour boundary.

6. The method according to claim 1, wherein A buoy-type collection device is deployed within the sea area covered by the predicted trajectory path. The propeller of the collection device generates a surface water flow that converges toward the storage bin of the collection device. The surface water flow guides the plastic waste to move along the direction of the predicted trajectory path to form a dynamic aggregation area, including: Deploying a buoy-type collection device within the sea area covered by the predicted trajectory path, with the storage bin of the collection device facing the extension direction of the predicted trajectory path, and starting the propeller of the collection device to rotate toward the storage bin to generate a surface water flow converging toward the storage bin; Adjusting the rotation direction and thrust distribution of the propeller so that the flow direction of the surface water flow is consistent with the extension direction of the trajectory prediction path; The surface water flow guides the plastic waste within the sea area covered by the trajectory prediction path to gather along the extension direction of the trajectory prediction path to form a dynamic gathering area.

7. The method according to claim 1, wherein Within the dynamic aggregation area, the differences in the surface reflection characteristics of the plastic waste under visible light in the continuous image sequence are analyzed, and based on the differences in the reflection characteristics, difficult-to-degrade plastic waste with preset degradation properties is identified and marked as a regulatory target, including: Extracting the reflection intensity values ​​of the surface of the plastic waste to visible light in the continuous image sequence within the dynamic aggregation area, where the reflection intensity values ​​of the surface of the plastic waste to visible light vary due to the different materials of the plastic waste; Compare the reflection intensity values ​​of different plastic waste to obtain the differences in reflection characteristics of different plastic waste; According to the difference in the reflection characteristics, the reflection intensity value is compared with the reflection characteristics corresponding to the preset degradation attribute. The reflection characteristics are the reflection intensity value, and the plastic waste whose reflection intensity value matches the degradation attribute of the difficult-to-decompose plastic is screened out; The matched plastic waste is regarded as difficult-to-decompose plastic waste, and the difficult-to-decompose plastic waste is marked as a regulatory target.

8. An intelligent target supervision system based on data analysis, characterized in that: include: A capture module is configured to capture a continuous image sequence containing visible light and thermal infrared bands using surveillance cameras distributed along the coastline, and identify the difference between the contour features of plastic waste in the continuous image sequence under visible light and the temperature distribution in the thermal infrared image; a segmentation module, configured to segment the contour boundaries of the plastic waste based on the contour features and obtain position changes of the contour boundaries in a continuous image sequence to generate a predicted trajectory path of the plastic waste moving on the sea surface with the ocean current; A deployment module is configured to deploy buoy-type collection devices within the sea area covered by the predicted trajectory path, wherein the propellers of the collection devices generate surface water currents that converge toward the storage bins of the collection devices, thereby guiding the plastic waste to move along the predicted trajectory path through the surface water currents to form a dynamic accumulation area; an analysis module configured to analyze, within the dynamic aggregation area, differences in the reflective properties of the plastic waste surfaces under visible light in the continuous image sequence, and identify, based on the differences in the reflective properties, difficult-to-degrade plastic waste with preset degradation properties and mark it as a regulatory target; The calculation module is used to calculate the offset of the supervision target based on the temperature distribution difference of the supervision target in the dynamic aggregation area, and generate a path correction instruction according to the offset to correct the movement trajectory of the collection device, so that the collection device can preferentially intercept the difficult-to-decompose plastic waste with the corrected movement trajectory, and complete the target supervision closed loop based on the real-time data update of the continuous image sequence.

Images