A Digital Traceability Method and System for Marine Waste

Abstract

The present invention relates to the technical field of marine garbage monitoring. The present invention discloses a digital traceability method and system for marine waste. The method includes first dividing a target sea area according to a preset rule to obtain central influence data corresponding to a central detection area and edge influence data corresponding to an edge area, determining a reflection influence degree value based on the central influence data and the edge influence data, then correcting an initial detection image according to the reflection influence degree value and a color correction algorithm to obtain a target detection image, and finally inputting the target detection image into a pre-constructed garbage classification model to obtain the garbage category, so as to effectively identify the reflection effect of the sea surface, reduce the detection error, improve the recognition accuracy of garbage targets, then perform path prediction through a garbage distribution map and real-time environmental data, and perform traceability processing in combination with a preset path inversion algorithm to accurately track the source of the garbage.

Description

Technical Field

[0001] The present invention relates to the technical field of marine garbage monitoring, and more specifically, to a digital tracing method and system for marine waste. Background Art

[0002] Plastic waste is the main component of marine debris. Since most plastic waste is floating, it can be monitored by obtaining sea level images. With the continuous maturity of target detection technology, it has become feasible to use equipment with machine vision to realize automatic monitoring of garbage. The automatic detection of marine debris by machines can generate a large amount of data and store it in the background database, providing important support for subsequent garbage recycling work. Although existing technologies have made certain progress in marine debris detection, there are still some problems that need to be solved urgently.

[0003] For example, the patent application with publication number CN117789047A provides an unsupervised marine debris identification and monitoring method based on remote sensing images, which improves the accuracy of the sea surface plastic debris identification network in identifying and detecting sea surface plastic debris. The patent application with publication number CN117994499A provides a marine debris detection method based on deep learning and electronic equipment and storage media, which supplements the target sample synthetic image obtained through virtual rendering processing into the marine data set, enriching the sample data, which is conducive to compensating for the loss of model accuracy caused by insufficient sample data.

[0004] Although the above-mentioned technical solutions provide relevant technical solutions for marine debris detection, none of them take into account the reflection effect of the sea surface that affects the accuracy of sea surface debris monitoring. Due to the undulations of the sea surface, changes in the angle of incidence of the sun and the influence of suspended particulate matter, the sea surface reflection phenomenon is relatively complex, which may cause highlights, artifacts or reduced contrast in the image, thereby reducing the detection system's accuracy in identifying garbage targets. This detection error not only affects the accurate identification of garbage, but also accumulates errors in the tracing process, resulting in large deviations in the source tracking and movement path prediction of marine garbage.

[0005] In view of this, the present invention proposes a digital traceability method and system for marine waste to solve the above problems. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a digital traceability method and system for marine waste.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] First, a digital traceability method for marine waste, comprising:

[0009] Divide the target sea area according to preset rules to obtain a central detection area and at least one marginal area, obtain the central influence data corresponding to the central detection area and the marginal influence data corresponding to the marginal area, and determine the reflection influence degree value according to the central influence data and the marginal influence data;

[0010] Obtain the initial detection image of the central detection area, and correct the initial detection image according to the reflection influence degree value and the color correction algorithm to obtain the target detection image;

[0011] Input the target detection image into a pre-constructed garbage classification model to obtain the garbage category, determine the garbage distribution map of the target sea area according to the garbage category, obtain the real-time environmental data, perform path prediction according to the real-time environmental data and the garbage distribution map to obtain the garbage flow path, and perform traceability processing on the garbage flow path according to the preset path inversion algorithm to obtain the garbage source area.

[0012] Further, the method for dividing the target sea area according to preset rules includes:

[0013] Take the monitoring range of the monitoring equipment in the target sea area as the central detection area, divide the other areas in the target sea area in an equal difference spacing manner to obtain M marginal areas, where the other areas are the areas in the target sea area except the central detection area, and M is an integer greater than 0.

[0014] Further, the method for determining the reflection influence degree value according to the central influence data and the marginal influence data includes:

[0015] Calculate the central influence degree value according to the central influence data, then calculate the marginal influence degree value according to the marginal influence data, and calculate the reflection influence degree value based on the central influence degree value and the marginal influence degree value.

[0016] Further, the method for correcting the initial detection image according to the reflection influence degree value and the color correction algorithm to obtain the target detection image includes:

[0017] Obtain the chromaticity average deviation value corresponding to the initial detection image based on the color correction algorithm, and input the initial detection image, the chromaticity average deviation value and the reflection influence degree value into a pre-constructed image correction model to obtain the target detection image.

[0018] Further, the method for obtaining the chromaticity average deviation value corresponding to the initial detection image based on the color correction algorithm includes:

[0019] Obtain the chromaticity deviation data in the initial detection image based on the color correction algorithm, and calculate the chromaticity average deviation value according to the chromaticity deviation data. The chromaticity deviation data includes the a chromaticity deviation mean, the b chromaticity deviation mean, the total number of pixels, the number of a chromaticity histogram groups, and the number of b chromaticity histogram groups.

[0020] Further, the method for constructing an image correction model includes:

[0021] Using a preset deep neural network as the basic model, the input layer in the deep neural network receives historical initial detection images, historical chromaticity average deviation values, and historical reflection influence values, the output layer in the deep neural network outputs historical target detection images, and when training the deep neural network, the cross-entropy loss function is selected as the loss function, and the loss function is minimized by the gradient descent method to update the weight parameters of the deep neural network. Through iterative training, an image correction model is obtained.

[0022] Further, the method for determining the garbage distribution map of the target sea area according to the garbage category includes:

[0023] Obtain the initial distribution map of the target sea area, and mark the initial distribution map according to the garbage category to obtain the garbage distribution map.

[0024] Further, the method for predicting the path based on the real-time environmental data and the garbage distribution map to obtain the garbage flow path includes:

[0025] Input the garbage distribution map into a preset hydrodynamic model to obtain a path probability distribution, and perform Monte Carlo simulation based on the path probability distribution and the real-time environmental data to obtain S garbage flow paths.

[0026] In a second aspect, a digital traceability system for marine waste, which is used to implement the above-mentioned digital traceability method for marine waste, includes:

[0027] A processing module: used to divide the target sea area according to a preset rule to obtain a central detection area and at least one edge area, obtain the central influence data corresponding to the central detection area, and the edge influence data corresponding to the edge area, and determine the reflection influence value according to the central influence data and the edge influence data;

[0028] A correction module: used to obtain the initial detection image of the central detection area, and correct the initial detection image according to the reflection influence value and the color correction algorithm to obtain the target detection image;

[0029] A classification module: used to input the target detection image into a pre-constructed garbage classification model to obtain the garbage category, determine the garbage distribution map of the target sea area according to the garbage category, obtain the real-time environmental data, predict the path based on the real-time environmental data and the garbage distribution map to obtain the garbage flow path, and perform traceability processing on the garbage flow path according to a preset path inversion algorithm to obtain the garbage source area.

[0030] Compared with the prior art, the beneficial effects of the present invention are:

[0031] The present invention first divides the target sea area according to preset rules, obtains the central influence data corresponding to the central detection area and the edge influence data corresponding to the edge area, determines the reflection influence degree value based on the central influence data and the edge influence data, then corrects the initial detection image according to the reflection influence degree value and the color correction algorithm to obtain the target detection image, and finally inputs the target detection image into a pre-constructed garbage classification model to obtain the garbage category. Then, by dividing the target sea area, obtaining the central influence data and the edge influence data, and calculating the reflection influence degree value, the present invention effectively identifies the reflection effect of the sea level and reduces the detection error. Then, by using the color correction algorithm to correct the initial detection image to obtain the target detection image, the recognition accuracy of the garbage target is improved. Finally, through the garbage distribution map and real-time environmental data for path prediction, and combined with the preset path inversion algorithm for traceability processing, the source of the garbage is accurately traced. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flowchart of a digital traceability method for marine waste in the present invention;

[0033] Figure 2 It is a schematic structural diagram of a digital traceability system for marine waste in the present invention;

[0034] Figure 3 It is a schematic diagram of the target sea area divided according to preset rules in the present invention.

[0035] Reference Signs:

[0036] 10, buoy; 20, central detection area; 30, first edge area; 40, second edge area. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] To make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the following further describes the present invention in detail with reference to specific embodiments and the accompanying drawings. In the following detailed description, many specific details are set forth to provide a thorough understanding of the described exemplary embodiments. However, it is obvious to those skilled in the art that some or all of these specific details may not be necessary to practice the described embodiments. In other exemplary embodiments, well-known structures are not described in detail to avoid unnecessarily obscuring the concepts of the present disclosure. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. At the same time, the various aspects described in the embodiments can be combined arbitrarily without conflict.

[0038] Embodiment 1

[0039] Please refer to Figure 1As shown in the figure, this embodiment discloses a digital traceability method for marine waste, including:

[0040] S10: Divide the target sea area according to a preset rule to obtain a central detection area and at least one edge area, obtain the central influence data corresponding to the central detection area and the edge influence data corresponding to the edge area, and determine the reflection influence degree value according to the central influence data and the edge influence data;

[0041] In this embodiment, the method for dividing the target sea area according to a preset rule includes:

[0042] Take the monitoring range of the monitoring device in the target sea area as the central detection area, divide the other areas in the target sea area in an arithmetic progression spacing manner to obtain M edge areas, where the other areas are the areas in the target sea area except the central detection area, and M is an integer greater than 0.

[0043] It should be noted that the above-mentioned monitoring device refers to a device that monitors the garbage on the sea surface by obtaining sea surface images. The monitoring device can be a buoy in the target sea area or a drone above the target sea area. It can be understood that when the monitoring device needs to take sea surface images, it is fixed. At this time, the monitoring device can be used as the origin, and the monitoring range of the monitoring device can be used as the central detection area.

[0044] Then, taking the monitoring device as a buoy for exemplary illustration, as Figure 3 shown, Figure 3 shows the buoy 10, the central detection area 20, the first edge area 30 and the second edge area 40, as well as the ring width rd corresponding to the first edge area 30 and the ring width RD corresponding to the second edge area 40. Then, from Figure 3 it can be seen that the central detection area 20 is circular, while the first edge area 30 and the second edge area 40 are both circular rings, and the ring width RD is twice the ring width rd. It can be understood that since this embodiment is divided in an arithmetic progression spacing manner, each time an edge area is extended outward, the corresponding ring width will be extended to twice the original.

[0045] It should be added that the central detection area 20 in this embodiment can also be of other shapes. Similarly, it can also be not divided in an arithmetic progression spacing manner. This embodiment does not limit this.

[0046] In this embodiment, the purpose of dividing the target sea area is that due to the influence of factors such as waves on the sea surface, the incident angle of the sun, and suspended particulate matter, sea surface reflection occurs. Moreover, the closer to the monitoring range the sea surface is, the greater the impact of sea surface reflection on garbage monitoring. Therefore, through area division, the impact of reflection on the monitoring effect can be better analyzed and adjusted, thereby improving the accuracy of garbage monitoring.

[0047] Among the above, the central influence data includes but is not limited to real-time wind speed, real-time incident angle, real-time particulate matter concentration value, and real-time water temperature. Taking real-time wind speed and real-time water temperature as examples, the greater the real-time wind speed, the easier it is to generate large waves on the sea surface, and the greater the wind speed, the rougher the sea surface, and the more obvious the scattering of reflected light. And temperature changes will cause changes in the refractive index of water, thereby affecting the intensity and properties of reflected light. Therefore, the smaller the real-time water temperature, the usually greater the reflectivity of the sea surface. The real-time incident angle mentioned above refers to the angle between the incident light and the normal line of the sea surface (i.e., the imaginary line perpendicular to the sea surface). The real-time incident angle can be obtained through an optical sensor, and this embodiment will not elaborate on this too much.

[0048] Similarly, the edge influence data includes but is not limited to real-time wind speed, real-time incident angle, real-time particulate matter concentration value, and real-time water temperature. In this embodiment, the central influence data and the edge influence data only differ in numerical size, and the types of data can be the same.

[0049] The method for determining the reflection influence degree value based on the central influence data and the edge influence data includes:

[0050] Calculate the central influence degree value according to the central influence data, then calculate the edge influence degree value according to the edge influence data, and calculate the reflection influence degree value based on the central influence degree value and the edge influence degree value.

[0051] The method for calculating the central influence degree value according to the central influence data includes:

[0052] CIV = ;

[0053] In the formula, CIV is the central influence degree value, is the real-time water temperature, is the real-time wind speed, is the real-time incident angle, is the real-time particulate matter concentration value, is the standard particulate matter concentration value, is the arccotangent function, is the arccosine function, is the arctangent function, is the exponential function with base e, is the logarithmic function with base 2, is a constant greater than 0, and e is the natural constant.

[0054] In this embodiment, taking the real-time wind speed and the real-time particulate matter concentration value as an example, when the real-time wind speed is greater, the sea surface is rougher, and the scattering of reflected light is more obvious. When the real-time particulate matter concentration value is lower than the standard particulate matter concentration value, the light scattering and absorption in the water decrease, and more light is directly reflected back to the sea surface, resulting in an increase in the reflection intensity. When the real-time particulate matter concentration value is higher than the standard particulate matter concentration value, the multiple reflections of scattered light on the water surface may increase the reflection intensity in certain directions, and this phenomenon will also lead to an increase in the reflection intensity. Therefore, from the above content, it can be seen that the greater the central influence degree value, the greater the reflection intensity of the corresponding central detection area.

[0055] It can be understood that the method for calculating the edge influence degree value is the same as the method for calculating the central influence degree value, and no further elaboration will be made in this embodiment.

[0056] The method for calculating the reflection influence degree value based on the central influence degree value and the edge influence degree value includes:

[0057] RIV = ;

[0058] In the formula, RIV is the reflection influence degree value, is the edge influence degree value, is the number of edge regions, is the weight factor corresponding to the central detection area, is the th weight factor corresponding to the edge region.

[0059] In this embodiment, it should be particularly noted that since the reflection has a more significant impact on garbage monitoring closer to the sea surface within the monitoring range, when setting the corresponding weight factors, the same principle is followed, that is, the closer to the sea surface within the monitoring range, the greater the corresponding weight factor. For example, an arithmetic sequence can be used to set these weight factors. In addition, the above reflection influence degree value is used to characterize the impact of the sea surface reflection of the target sea area on the monitoring effect. When the reflection influence degree value is greater, the interference of the reflection on the sea surface garbage monitoring is also greater.

[0060] The purpose of doing this is to more accurately evaluate and adjust the impact of reflection on garbage monitoring, thereby improving the accuracy and reliability of monitoring. By reasonably setting the weight factors and calculating the reflection influence degree value, potential problems in marine garbage monitoring can be better identified and managed.

[0061] S20: Obtain the initial detection image of the central detection area, and correct the initial detection image according to the reflection influence degree value and the color correction algorithm to obtain the target detection image;

[0062] It should be noted that the initial detection image refers to the garbage image within the central detection area. The garbage can be plastic garbage, which usually floats on the sea surface due to its own characteristics. The color correction algorithm refers to the technology used to adjust and optimize the image color to ensure that the color performance of the image is closer to the real scene or achieves a specific visual effect. Color correction is very important in various image processing tasks. Sea surface garbage monitoring is usually carried out in a complex natural environment, including different weather conditions, time periods (such as sunrise, sunset), and water body reflection, etc. These factors will cause color distortion of the image. Through color correction, the white balance, contrast, and brightness of the image can be adjusted, making the color of the garbage closer to its real color under standard lighting conditions. The corrected image not only has a better visual effect but also can improve the classification accuracy in subsequent processing steps, especially in machine learning algorithms.

[0063] The method for correcting the initial detection image according to the reflection influence degree value and the color correction algorithm to obtain the target detection image includes:

[0064] Obtain the chromaticity average deviation value corresponding to the initial detection image based on the color correction algorithm, and input the initial detection image, the chromaticity average deviation value, and the reflection influence degree value into a pre-constructed image correction model to obtain the target detection image.

[0065] The method for obtaining the chromaticity average deviation value corresponding to the initial detection image based on the color correction algorithm includes:

[0066] Obtain the chromaticity deviation data in the initial detection image based on the color correction algorithm, and calculate the chromaticity average deviation value according to the chromaticity deviation data. The chromaticity deviation data includes the a chromaticity deviation mean, the b chromaticity deviation mean, the total number of pixels, the number of a chromaticity histogram groups, and the number of b chromaticity histogram groups.

[0067] The method for calculating the chromaticity average deviation value according to the chromaticity deviation data includes:

[0068] MCD = ;

[0069] In the formula, MCD is the chromaticity average deviation value, is the image chromaticity value index, is the a chromaticity deviation mean of the initial detection image, is the b chromaticity deviation mean of the initial detection image, is the total number of pixels of the initial detection image, is the number of a chromaticity histogram groups, is the number of b chromaticity histogram groups.

[0070] It should be noted that in the CIELab color space, the a chromaticity and b chromaticity are used to represent the color attributes of colors. The CIELab color space is more in line with the way of color perception than RGB, and can perform color correction and analysis more accurately. The a chromaticity represents the position of the color on the green-to-red axis, with negative values tending towards green and positive values tending towards red. The b chromaticity represents the position of the color on the blue-to-yellow axis, with negative values tending towards blue and positive values tending towards yellow. In this embodiment, before processing the initial detection image through the color correction algorithm, the initial detection image is first converted from the RGB color space to the CIELab color space, and the histogram distribution of the image based on the a and b chromaticities is analyzed, so as to obtain the , , , and . It can be understood that refers to the a chromaticity histogram frequency at the h chromaticity value, that is, in the CIELab color space, the number of pixel points with the a chromaticity channel value equal to h. Similarly.

[0071] It should be added that the above chromaticity average deviation value characterizes the color deviation degree of each chromaticity channel in the initial detection image. Specifically, the chromaticity average deviation value reflects the average deviation of the color of the initial detection image relative to the standard chromaticity mean value (usually 128, the neutral gray value commonly used in the CIELab color space). It can be understood that in garbage classification, the colors of some garbage may be very close, and fine color differences are needed to distinguish them. However, when the MCD value is large, these fine differences may be masked by the chromaticity deviation, making it difficult to correctly distinguish garbage with similar colors. For example, a light blue plastic bottle and white foam material may become indistinguishable under high MCD conditions, thus affecting the classification accuracy.

[0072] The method for constructing an image correction model includes:

[0073] Using a preset deep neural network as the basic model, the input layer in the deep neural network receives the historical initial detection image, the historical chromaticity average deviation value, and the historical reflection influence degree value, and the output layer in the deep neural network outputs the historical target detection image. When training the deep neural network, the cross-entropy loss function is selected as the loss function, and the loss function is minimized through the gradient descent method to update the weight parameters of the deep neural network. Through iterative training, an image correction model is obtained.

[0074] It can be understood that in this embodiment, the initial detection image is optimized by the image correction model to obtain the target detection image. The advantage of this is that it can significantly improve the accuracy and reliability of garbage monitoring and classification, enhance the adaptability and real-time performance of the system, simplify the processing flow, and ultimately improve the effectiveness and efficiency of the entire marine garbage monitoring system. These advantages make the image correction model an indispensable part of the marine garbage monitoring process.

[0075] S30: Input the target detection image into the pre-constructed garbage classification model to obtain the garbage category. Determine the garbage distribution map of the target sea area according to the garbage category, obtain the real-time environmental data, perform path prediction based on the real-time environmental data and the garbage distribution map to obtain the garbage flow path, and perform traceability processing on the garbage flow path according to the preset path inversion algorithm to obtain the garbage source area;

[0076] The construction method of the garbage classification model includes:

[0077] Obtain Q groups of training data, where Q is a positive integer greater than 1. The training data includes historical target detection images and historical garbage categories. Use the historical target detection images and historical garbage categories as the sample set, divide the sample set into a training set and a test set, construct a classifier, use the historical target detection images in the training set as input data, use the historical garbage categories in the training set as output data, train the classifier to obtain an initial classifier, and use the test set to test the initial classifier. Output the classifier that meets the preset accuracy as the garbage classification model. The classifier is preferably a Naive Bayes model or a Support Vector Machine model;

[0078] In this embodiment, the garbage categories include but are not limited to soft plastic garbage, hard plastic garbage, and foam plastic garbage. Specifically, they can be at least 4 types of typical marine leakage plastic garbage such as fishing nets, foam, plastic bottles, and plastic bags.

[0079] The method for determining the garbage distribution map of the target sea area according to the garbage category includes:

[0080] Obtain the initial distribution map of the target sea area, and mark the initial distribution map according to the garbage category to obtain the garbage distribution map.

[0081] It should be noted that the initial distribution map refers to a preliminary data map based on the garbage types and locations in the target sea area. This map is not the most accurate garbage distribution, but a graphical representation that includes garbage types and preliminary locations. After the precise classification in Embodiment 1, the precise garbage categories are obtained, and then the initial distribution map is marked according to the garbage categories, which provides high-precision data support for garbage flow path prediction and source traceability, and further enhances the reliability and accuracy of garbage source tracking.

[0082] In this embodiment, the real-time environmental data includes but is not limited to real-time wind speed, sea current speed value, wind direction angle value, and tidal rise value. The tidal rise value can be the change in the tide level compared to the reference water level at a specific time point, and is usually used to measure the ebb and flow amplitude of the tide, that is, the height change of the water level.

[0083] The method for predicting the path based on the real-time environmental data and the garbage distribution map includes:

[0084] Input the garbage distribution map into a preset hydrodynamic model to obtain a path probability distribution, and perform Monte Carlo simulation based on the path probability distribution and the real-time environmental data to obtain S garbage flow paths.

[0085] It can be understood that the hydrodynamic model can be the Lagrangian particle tracking model in the prior art. Taking the Lagrangian particle tracking model as an example, the Lagrangian model is a tracking model based on material points. In the water flow, each piece of garbage is regarded as a particle, and its trajectory along with the fluid flow is tracked. These particles are affected by external environmental factors such as sea currents, wind speeds, and tides.

[0086] It should be added that Monte Carlo simulation is a numerical calculation method based on random sampling, which solves problems through a large number of random trials and is applicable to dealing with complex systems with uncertainties. Since the flow of garbage in this embodiment is uncertain, in this embodiment, traditional models may not be able to fully consider the complex interaction relationships of all environmental variables, resulting in inaccurate path prediction. Monte Carlo simulation generates multiple flow paths through multiple random trials, provides a path probability distribution, and thus finds the garbage flow path under various possible situations.

[0087] Similarly, the above path inversion algorithm can be methods such as the shortest path algorithm, reverse simulation algorithm, or particle filter algorithm in the prior art. These algorithms can effectively deduce the source area of the garbage from its final position. In this embodiment, the path inversion algorithm can be the reverse simulation algorithm. By using the reverse simulation algorithm, the flow trajectory of the garbage is deduced backward to simulate the source of the garbage flow. This process is similar to inferring the starting point given the end point, and can effectively track the source of the garbage, especially in a complex flow field.

[0088] In this embodiment, by combining accurate calibration of garbage categories, hydrodynamic models, and path inversion algorithms, the system can accurately trace the source area of garbage. In a complex marine environment, the flow of garbage is affected by various factors. By reverse-simulating the garbage flow path, the system can deduce the source from the final position of the garbage. Especially in areas with complex ocean currents and large tidal changes, the accuracy of source tracking has been greatly improved. Moreover, through regional division, calculation of reflection influence degree, and color correction, the image processing process is optimized, and the accuracy of monitoring images is improved. Through path prediction and source tracing, the distribution and flow trend of garbage are further converted into actionable data to achieve real-time tracking of the garbage source. This combination not only improves the detection accuracy of marine garbage but also provides more targeted strategies for subsequent garbage management and treatment, thus effectively improving the overall efficiency of garbage classification and management.

[0089] In this embodiment, the target sea area is first divided according to preset rules to obtain the central influence data corresponding to the central detection area and the edge influence data corresponding to the edge area. The reflection influence degree value is determined based on the central influence data and the edge influence data. Then, the initial detection image is corrected according to the reflection influence degree value and the color correction algorithm to obtain the target detection image. Finally, the target detection image is input into a pre-constructed garbage classification model to obtain the garbage category. Therefore, in this embodiment, by dividing the target sea area, obtaining the central influence data and the edge influence data, and calculating the reflection influence degree value, the reflection effect of the sea surface is effectively identified, and the detection error is reduced. Then, the initial detection image is corrected using the color correction algorithm to obtain the target detection image, improving the recognition accuracy of garbage targets. Finally, path prediction is performed through the garbage distribution map and real-time environmental data, and source tracing is carried out in combination with the preset path inversion algorithm to accurately track the source of the garbage.

[0090] Embodiment 2

[0091] As Figure 2 shown, on the basis of Embodiment 1, this embodiment provides a digital traceability system for marine waste, including:

[0092] A processing module: used to divide the target sea area according to preset rules to obtain a central detection area and at least one edge area, obtain the central influence data corresponding to the central detection area and the edge influence data corresponding to the edge area, and determine the reflection influence degree value based on the central influence data and the edge influence data;

[0093] In this embodiment, the method for dividing the target sea area according to preset rules includes:

[0094] Take the monitoring range of the monitoring device in the target sea area as the central detection area, divide the other areas in the target sea area in an arithmetic progression interval manner to obtain M marginal areas, where the other areas are the areas in the target sea area except the central detection area, and M is an integer greater than 0.

[0095] It should be noted that the above-mentioned monitoring device refers to a device that monitors garbage on the sea surface by obtaining sea surface images. The monitoring device can be a buoy in the target sea area or a drone above the target sea area. It can be understood that when the monitoring device needs to take sea surface images, it is fixed. At this time, the monitoring device can be used as the origin, and the monitoring range of the monitoring device can be used as the central detection area.

[0096] Then, taking the monitoring device as a buoy for exemplary illustration, as Figure 3 shown, Figure 3 shows the buoy 10, the central detection area 20, the first marginal area 30 and the second marginal area 40, as well as the ring width rd corresponding to the first marginal area 30 and the ring width RD corresponding to the second marginal area 40. Then from Figure 3 it can be seen that the central detection area 20 is circular, while the first marginal area 30 and the second marginal area 40 are both circular rings, and the ring width RD is twice the ring width rd. It can be understood that since in this embodiment, the division is carried out in an arithmetic progression interval manner, therefore, for each outward expansion of a marginal area, the corresponding ring width will be expanded to twice the original.

[0097] In the above, the central influence data includes but is not limited to the real-time wind speed, real-time incident angle, real-time particulate matter concentration value and real-time water temperature. Taking the real-time wind speed and real-time water temperature as examples, the greater the real-time wind speed, the easier it is to generate large waves on the sea surface, and the greater the wind speed, the rougher the sea surface, and the more obvious the scattering of reflected light. And the change in temperature will cause the refractive index of water to change, thereby affecting the intensity and properties of the reflected light. Therefore, the smaller the real-time water temperature, the usually greater the reflectivity of the sea surface. The above-mentioned real-time incident angle refers to the angle between the incident light and the normal line of the sea surface (i.e., the imaginary line perpendicular to the sea surface). The real-time incident angle can be obtained through an optical sensor, and this embodiment will not elaborate on this too much.

[0098] Similarly, the marginal influence data includes but is not limited to the real-time wind speed, real-time incident angle, real-time particulate matter concentration value and real-time water temperature. In this embodiment, the central influence data and the marginal influence data are only different in numerical values, and the types of data can be the same.

[0099] The method for determining the reflection influence degree value according to the central influence data and the marginal influence data includes:

[0100] Calculate the central influence degree value based on the central influence data, then calculate the marginal influence degree value based on the marginal influence data, and calculate the reflection influence degree value based on the central influence degree value and the marginal influence degree value.

[0101] Calibration module: used to obtain the initial detection image of the central detection area, and calibrate the initial detection image according to the reflection influence degree value and the color calibration algorithm to obtain the target detection image;

[0102] It should be noted that the initial detection image refers to the garbage image in the central detection area. The garbage can be plastic garbage. Due to its own characteristics, plastic garbage usually floats on the sea surface. The color calibration algorithm refers to the technology used to adjust and optimize the image color to ensure that the color performance of the image is closer to the real scene or achieves a specific visual effect. Color calibration is very important in various image processing tasks. Sea surface garbage monitoring is usually carried out in a complex natural environment, including different weather conditions, time periods (such as sunrise, sunset), and water body reflection, etc. These factors will cause color distortion of the image. Through color calibration, the white balance, contrast, and brightness of the image can be adjusted, so that the color of the garbage is closer to its real color under standard lighting conditions. The calibrated image not only has a better visual effect, but also can improve the classification accuracy in subsequent processing steps, especially in machine learning algorithms.

[0103] The method of calibrating the initial detection image according to the reflection influence degree value and the color calibration algorithm to obtain the target detection image includes:

[0104] Obtain the chromaticity average deviation value corresponding to the initial detection image based on the color calibration algorithm, and input the initial detection image, the chromaticity average deviation value, and the reflection influence degree value into the pre-constructed image calibration model to obtain the target detection image.

[0105] The construction method of the image calibration model includes:

[0106] Obtain a sample data set, which includes historical initial detection images, historical chromaticity average deviation values, historical reflection influence degree values, and historical target detection images. Divide the sample data set into a sample training set and a sample test set, construct a regression network, use the historical initial detection images, historical chromaticity average deviation values, and historical reflection influence degree values in the sample training set as the input data of the regression network, use the historical target detection images in the sample training set as the output data of the regression network, train the regression network to obtain an initial regression network for predicting real-time target detection images, use the sample test set to test the initial regression network, and output the initial regression network that meets the requirement of being less than the preset error value as the image calibration model. The initial regression network is preferably a deep neural network model.

[0107] It can be understood that in this embodiment, the initial detection image is optimized by the image correction model to obtain the target detection image. The advantage of this is that it can significantly improve the accuracy and reliability of garbage monitoring and classification, enhance the adaptability and real-time performance of the system, simplify the processing flow, and ultimately improve the effectiveness and efficiency of the entire marine garbage monitoring system. These advantages make the image correction model an indispensable part of the marine garbage monitoring process.

[0108] Classification module: It is used to input the target detection image into the pre-constructed garbage classification model to obtain the garbage category, determine the garbage distribution map of the target sea area according to the garbage category, obtain real-time environmental data, perform path prediction based on the real-time environmental data and the garbage distribution map to get the garbage flow path, and perform traceability processing on the garbage flow path according to the preset path inversion algorithm to obtain the garbage source area;

[0109] The construction method of the garbage classification model includes:

[0110] Obtain Q groups of training data, where Q is a positive integer greater than 1. The training data includes historical target detection images and historical garbage categories. Use the historical target detection images and historical garbage categories as the sample set, divide the sample set into a training set and a test set, construct a classifier, use the historical target detection images in the training set as input data, use the historical garbage categories in the training set as output data, train the classifier to obtain an initial classifier, use the test set to test the initial classifier, and output the classifier that meets the preset accuracy as the garbage classification model. The classifier is preferably a Naive Bayes model or a Support Vector Machine model;

[0111] In this embodiment, the garbage categories include but are not limited to soft plastic garbage, hard plastic garbage, and foam plastic garbage. Specifically, they can be at least 4 types of typical marine leakage plastic garbage such as fishing nets, foams, plastic bottles, and plastic bags.

[0112] The detailed description elaborated above with reference to the drawings describes examples and does not represent all examples that can be implemented or fall within the scope of the claims. The terms "example" and "exemplary" when used in this specification mean "serving as an example, instance, or illustration" and do not mean "superior to or better than other examples".

[0113] The reference to "one embodiment" or "an embodiment" throughout this specification means that the specific features, structures, or characteristics described in connection with that embodiment are included in at least one embodiment of the present invention. Therefore, the use of these phrases may refer to more than just one embodiment. Additionally, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0114] It should also be noted that these embodiments may be described as processes depicted as flowcharts, flow diagrams, structural diagrams, or block diagrams. Although a flowchart may describe the operations as a sequential process, many of these operations can be performed in parallel or concurrently. Additionally, the order of these operations may be rearranged.

Claims

1. A digital traceability method for marine waste, characterized in that, Including: Dividing the target sea area according to preset rules to obtain a central detection area and at least one edge area, acquiring central influence data corresponding to the central detection area and edge influence data corresponding to the edge area, and determining a reflection influence degree value according to the central influence data and the edge influence data; The central influence data includes real-time wind speed, real-time incident angle, real-time particulate matter concentration value, and real-time water temperature; The edge influence data includes real-time wind speed, real-time incident angle, real-time particulate matter concentration value, and real-time water temperature; Acquiring an initial detection image of the central detection area, and correcting the initial detection image according to the reflection influence degree value and a color correction algorithm to obtain a target detection image; Inputting the target detection image into a pre-constructed garbage classification model to obtain garbage categories, determining a garbage distribution map of the target sea area according to the garbage categories, acquiring real-time environmental data, predicting a path according to the real-time environmental data and the garbage distribution map to obtain a garbage flow path, and performing a traceability process on the garbage flow path according to a preset path inversion algorithm to obtain a garbage source area.

2. The digital traceability method for marine waste according to claim 1, wherein, The method for dividing the target sea area according to preset rules includes: Taking the monitoring range of the monitoring device in the target sea area as the central detection area, dividing the other areas in the target sea area in an equal difference spacing manner to obtain M edge areas, where the other areas are the areas in the target sea area except the central detection area, and M is an integer greater than 0.

3. A digital traceability method for marine waste according to claim 2, characterized in that, The method for determining the reflection influence degree value according to the central influence data and the edge influence data includes: Calculating a central influence degree value according to the central influence data, then calculating an edge influence degree value according to the edge influence data, and calculating the reflection influence degree value based on the central influence degree value and the edge influence degree value; The method for calculating the central influence degree value according to the central influence data includes: CIV= ; In the formula, CIV is the central influence degree value, is the real-time water temperature, is the real-time wind speed, is the real-time incident angle, is the real-time particulate matter concentration value, is the standard particulate matter concentration value, is the arccotangent function, is the arccosine function, is the arctangent function, is the exponential function with base e, is the logarithmic function with base 2, is a constant greater than 0, and e is the natural constant; The method for calculating the edge influence degree value is the same as the method for calculating the central influence degree value; The method for calculating the reflection influence degree value based on the central influence degree value and the edge influence degree value includes: RIV= ; wherein, RIV is the reflection influence value, is the edge influence value, is the number of edge regions, is the weight factor corresponding to the central detection region, is the weight factor corresponding to the 4. A digital traceability method for marine waste according to claim 3, characterized in that The method for correcting the initial detection image according to the reflection influence degree value and the color correction algorithm to obtain the target detection image includes: Obtaining a chromaticity average deviation value corresponding to the initial detection image based on the color correction algorithm, inputting the initial detection image, the chromaticity average deviation value, and the reflection influence degree value into a pre-constructed image correction model to obtain the target detection image; The method for obtaining the chromaticity average deviation value corresponding to the initial detection image based on the color correction algorithm includes: Obtaining chromaticity deviation data in the initial detection image based on the color correction algorithm, and calculating the chromaticity average deviation value according to the chromaticity deviation data, where the chromaticity deviation data includes an a chromaticity deviation mean value, a b chromaticity deviation mean value, the total number of pixels, the number of a chromaticity histogram groups, and the number of b chromaticity histogram groups; The method for calculating the chromaticity average deviation value according to the chromaticity deviation data includes: MCD= ; where MCD is the chromaticity average deviation value, is the index of the image chromaticity value, is the deviation of the a chromaticity of the initial detection image from the mean value, is the deviation of the b chromaticity of the initial detection image from the mean value, is the total number of pixels of the initial detection image, is the number of groups of the a chromaticity histogram, is the number of groups of the b chromaticity histogram.

5. A digital traceability method for marine waste according to claim 4, characterized in that, The method for constructing the image correction model includes: Taking a preset deep neural network as the basic model, the input layer in the deep neural network receives the historical initial detection image, the historical chromaticity average deviation value, and the historical reflection influence degree value, and the output layer in the deep neural network outputs the historical target detection image. When training the deep neural network, the cross-entropy loss function is selected as the loss function, and the loss function is minimized by the gradient descent method to update the weight parameters of the deep neural network. Through iterative training, an image correction model is obtained.

6. The digital traceability method for marine waste according to claim 5, characterized in that, The method for determining the garbage distribution map of the target sea area according to the garbage category includes: Obtaining the initial distribution map of the target sea area, and marking the initial distribution map according to the garbage category to obtain the garbage distribution map.

7. The digital traceability method for marine waste according to claim 6, wherein, The method for predicting the path based on the real-time environmental data and the garbage distribution map to obtain the garbage flow path includes: Inputting the garbage distribution map into a preset hydrodynamic model to obtain a path probability distribution, and performing Monte Carlo simulation based on the path probability distribution and the real-time environmental data to obtain S garbage flow paths.

8. A digital traceability system for marine waste, which is used to implement a digital traceability method for marine waste described in any one of claims 1-7, characterized in that, Including: Processing module: used to divide the target sea area according to preset rules to obtain a central detection area and at least one edge area, obtain the central influence data corresponding to the central detection area, and the edge influence data corresponding to the edge area, and determine the reflection influence degree value according to the central influence data and the edge influence data; Correction module: used to obtain the initial detection image of the central detection area, and correct the initial detection image according to the reflection influence degree value and the color correction algorithm to obtain the target detection image; Classification module: used to input the target detection image into a pre-constructed garbage classification model to obtain the garbage category, determine the garbage distribution map of the target sea area according to the garbage category, obtain the real-time environmental data, predict the path according to the real-time environmental data and the garbage distribution map to obtain the garbage flow path, and perform source tracing processing on the garbage flow path according to the preset path inversion algorithm to obtain the garbage source area.

Images