Water body micro-plastic detection method and system based on unmanned aerial vehicle optics and Raman spectrum technology

Abstract

The invention belongs to the field of water body pollution monitoring, and discloses a water body micro-plastic monitoring method and system based on unmanned aerial vehicle optics and Raman spectrum technology, a multi-mode water body micro-plastic sensor based on optical imaging and a hyperspectral Raman laser radar is constructed, and the constructed sensor is used for collecting water body micro-plastic data; establishing a solar flare correction model and an underwater light attenuation compensation model to pre-process the collected water body micro-plastic data; performing water body micro-plastic detection on the preprocessed data by using a multi-mode underwater micro-plastic semantic segmentation model; a spatial and temporal distribution model of micro-plastic pollution in the urban water body is constructed, and migration and accumulation, main sources, future distribution trend and the like of micro-plastics in the water body are comprehensively analyzed. According to the invention, the problems of incapability of large-scale real-time detection, large interference of a water body environment and the like in a traditional water body micro-plastic detection method can be effectively solved, and the capability and efficiency of water body micro-plastic detection by using the unmanned aerial vehicle are greatly improved; reliable data support and technical guarantee are provided for water body pollution monitoring, ecological environment assessment and micro-plastic treatment, and large-scale water body monitoring and environment protection work are promoted to be carried out.

Description

Technical Field

[0001] The present invention belongs to the field of water pollution monitoring, and particularly relates to a method and system for detecting microplastics in water based on unmanned aerial vehicle optical and Raman spectroscopy technologies. Background Art

[0002] Against the backdrop of the current global environmental issues attracting much attention, water microplastic pollution has become a new and severe challenge. The extensive application of plastics in modern society, from daily necessities to industrial packaging, has led to a large amount of plastic waste entering the natural environment due to their massive production and random disposal. Urban water bodies, including rivers, lakes, and artificial water systems within cities, have become the main pollution areas for microplastics. Their sources are extensive and complex, generally including two categories: primary microplastics and secondary microplastics. Among them, primary microplastics mainly originate from tiny plastic particles directly discharged or leaked during industrial production processes; secondary microplastics are formed by the decomposition of larger plastic wastes (such as plastic bottles, bags, fishing nets, plastic packaging materials, etc.) in the natural environment through long-term photolysis, mechanical abrasion, or biodegradation. The distribution of microplastics in water shows complex spatiotemporal characteristics. Spatially, they not only float on the water surface but also suspend at different depths in the water body and even deposit on the bottom, forming a three-dimensional pollution from the water surface to the bottom. In the time dimension, their distribution is affected by various factors such as water flow, tides, and seasonal changes, showing dynamic characteristics. This complex distribution state makes microplastics ubiquitous in the environment. Currently, microplastic pollution has been detected in various media such as air, soil, lakes, rivers, and oceans, and even in food and drinking water. Additionally, microplastics are prone to enriching microorganisms, heavy metals, and organic pollutants, which will cause serious harm to human health as they are transmitted and enriched along the food chain. Since most cities are built along water bodies and the depths of rivers and lakes are relatively shallow, developing a large-scale and real-time detection technology to achieve the detection, identification, and quantification of plastics and microplastics in the water environment, and at the same time trace the source, is of irreplaceable key significance for maintaining the stability of the ecosystem and the sustainable development of humanity.

[0003] The existing methods for detecting microplastics in water bodies can be mainly divided into two categories: physical characterization and chemical characterization. In physical characterization, the morphology and physical characteristics of microplastics are often evaluated by means of microscopy (visual) analysis, scanning electron microscopy, etc.; in chemical characterization, techniques such as spectral analysis, thermal analysis, and mass spectrometry are used to determine the polymer type and quantitative composition of microplastics. Among them, microscopy analysis is one of the most commonly used detection methods in physical characterization. However, it highly relies on manual operation, its detection limit is usually only applicable to microplastics with a size larger than 500 μm, and it is highly subjective, easily affected by the experience and judgment of the detection personnel, resulting in misjudgment and missed judgment, and it is also impossible to obtain the chemical composition information of microplastics; in chemical characterization, sample preparation in energy-dispersive X-ray spectroscopy is complex and time-consuming, and each particle under the electron microscope needs to be analyzed one by one, with low work efficiency; infrared spectroscopy analysis is also a detection method based on chemical characterization, and it has strict requirements for samples, such as requiring the samples to be dry, thin in thickness, and fixed on a transparent base. To sum up, the existing methods for detecting microplastics in water bodies have problems such as low detection efficiency, strong dependence on manual labor, complex sample preparation, and strict requirements for samples, making it difficult to achieve efficient and real-time on-site detection.

[0004] Through the above analysis, the problems and defects existing in the existing detection methods are as follows:

[0005] (1) Interference from complex environments in outdoor water body detection. Most of the existing microplastic detection technologies are carried out in a laboratory environment, but more external environmental challenges will be faced in the actual detection of microplastics in urban water bodies. Specifically, solar flares and the attenuation of light by water have an adverse impact on the detection of microplastics in outdoor water bodies. The high-intensity specular reflection formed by solar flares on the water surface seriously affects the identification and analysis of microplastics, greatly reducing the accuracy and stability of detection. And the attenuation effect of water on light causes the intensity of light to continuously weaken when penetrating the water body. Whether it is visible light for imaging or specific wavelength light for spectral analysis, their propagation distance and effect are severely limited. These complex outdoor environmental factors lead to a reduction in the water depth and range that the detection equipment can detect and a decrease in the optical imaging quality, greatly increasing the difficulty and uncertainty of monitoring microplastic pollution in large outdoor water areas.

[0006] (2) Limitations of relying on manual sampling. Most of the existing water body microplastic detection methods rely on manual sampling followed by further analysis, and there are still many limitations in terms of sensitivity, efficiency, and adaptability. The manual sampling process is extremely cumbersome, requiring testers to collect water samples at different locations and depths. This not only consumes a large amount of manpower and material resources but also has low efficiency. The collected water samples are then taken back to the laboratory for processing, which usually involves complex steps such as filtration, microscopic observation, and chemical analysis. The entire process takes a long time, often taking several days to several weeks from sampling to obtaining results, while the distribution of microplastics is often dynamic. In addition, limited by manpower and material resources, manual sampling can only be carried out at limited points, unable to comprehensively cover large areas of water bodies, and it is difficult to meet the requirements for remote, real-time, and large-scale monitoring of microplastic pollution in urban water bodies. Summary of the Invention

[0007] Aiming at the problems and defects existing in the existing water body microplastic detection technology, the present invention provides a water body microplastic detection method and system based on unmanned aerial vehicle (UAV) optical and Raman spectroscopy technologies, specifically involving a multi-modal water body microplastic data collection, detection, and analysis method and system with an optical imaging and Raman spectroscopy lidar carried on a UAV platform.

[0008] The present invention is implemented as follows. A water body microplastic detection method based on UAV optical and Raman spectroscopy technologies includes: using a multi-modal water body microplastic sensor built based on optical imaging and hyperspectral Raman lidar to collect water body microplastic data; establishing a solar flare correction model and an underwater light attenuation compensation model to preprocess the collected water body microplastic data; using a multi-modal underwater microplastic semantic segmentation model to detect water body microplastics in the preprocessed data; and constructing a spatio-temporal distribution model to comprehensively analyze the distribution trend of water body microplastics, etc.

[0009] Further, the water body microplastic detection method based on UAV optical and Raman spectroscopy technologies includes the following steps:

[0010] Step 1, construct a multi-modal water body microplastic sensor based on optical imaging and hyperspectral Raman lidar, and use the constructed sensor to collect water body microplastic data;

[0011] Step 2, establish a solar flare correction model and an underwater light attenuation compensation model to preprocess the collected water body microplastic data;

[0012] Step 3, use a multi-modal underwater microplastic semantic segmentation model to detect water body microplastics in the preprocessed data;

[0013] Step 4, construct a spatio-temporal distribution model of microplastic pollution in urban water bodies to comprehensively analyze the migration and accumulation, main sources, and future distribution trends of water body microplastics, etc.

[0014] Furthermore, the specific mathematical model expression of the solar flare correction model in step two is as follows:

[0015]

[0016] where L i is the radiance of band i before correction, and L i ′ is the radiance of band i after correction; α ij is the correction coefficient, L j is the radiance of the NIR band, is the average radiance of the NIR band. Furthermore, α ij can be determined by the following formula:

[0017]

[0018] where ρ ij is the covariance between band i and band j (near-infrared band), which is used to measure the linear correlation degree between the radiance sample data of the two bands, and ρ jj is the covariance of the NIR band itself; L i,n and L j,n in formula (3) are the nth radiance sample values of band i and band j respectively, and N is the number of samples. The first part on the right side of formula (3) calculates the average value of the product of the corresponding sample values of the two bands, and the second part on the right side is the product of the average values of the two band samples. When the change trends of the radiances of the two bands are similar, that is, the effects of solar flares on these two bands are similar, the value of ρ ij will be larger; if the change trends are completely uncorrelated, ρ ij is close to 0; if the trends are opposite, ρ ij is negative. Furthermore, the contribution of solar flares to the radiance is determined by the following formula:

[0019] L sg = c(λ)p(η xs , η ys ) (4)

[0020]

[0021] where c(λ) depends on the wavelength and includes the Fresnel reflectivity of the surface and the spectral intensity of the incident sunlight; p(η xs , η ys ) is the probability density function based on the surface slope and is related to the relative positions of the sun and the sensor. In the actual water surface environment, solar flares are generated when sunlight is reflected by the water surface and enters the sensor's field of view. The water surface is not completely flat but has waves and undulations of various scales, which will cause the normal direction of each point on the water surface to change continuously; the parameters η xs , ηys Obtained from Equation (5), n x , n y , n z are the components of the surface normal vector n, obtained from Equation (6), k s points to the sun, k o points to the sensor. According to this model, during the flight detection of the UAV, each band can be corrected. First, the radiance of the required band is obtained through the sensor, then the contribution of solar flare to the radiance is calculated, and then the correlation degree of the radiance between band i and NIR band j is obtained by calculating the covariance, and then the correction coefficient α ij is determined. Finally, the corrected radiance is calculated through the correction model Equation (1). Based on the calculation results, the parameters of the active and passive optical imaging sensors are dynamically adjusted, so as to obtain a clearer water body image and reduce the interference of solar flare on the detection of microplastics;

[0022] Furthermore, in step two, the underwater light attenuation compensation model adopts the signal light enhancement technology, and its specific mathematical model expression is as follows:

[0023] I(x,y) = D(x,y) + B(x,y) = J(x,y)·t(x,y) + A ∞ (1 - t(x,y)) (7) where D(x,y) represents direct transmission, and B(x,y) represents backscattering. I(x,y) is the captured image, J(x,y) is the image without scattering, representing the image formed when the light is not affected by water body scattering under ideal conditions. A ∞ is the uniform background light, which is mainly formed by the weak luminescence of the water body itself, the multiple scattering of environmental light in the water body, and the inherent noise of the sensor and other factors. t(x,y) is the transmittance, which describes the attenuation of light during propagation and is usually related to the depth of field (DOP) and the attenuation coefficient. It can be expressed as the following formula:

[0024] t(x,y) = e -c(λ)d(x,y) (8)

[0025] where d(x,y) represents DOP, that is, the distance between the target and the camera, and c(λ) represents the total attenuation coefficient. According to Equation (7) and Equation (8), the transmittance t(x,y) and the image without scattering J(x,y) can be calculated. And the beam attenuation involves two basic processes, namely absorption and scattering. Therefore, C(λ) can be expressed as the sum of these two contributions, expressed by the following formula:

[0026] c(λ) = a(λ) + b(λ) (9)

[0027] Among them, λ represents the wavelength of the light beam, and a(λ), b(λ) respectively represent the absorption coefficient and the scattering coefficient. In the above model, the signal light enhancement technology can be used to optimize parameters from the following aspects to reduce the attenuation of light by water: First, it is known that the absorption and scattering coefficients of water in the blue-green spectral region (450 - 570 nm) are relatively small. A light source in this band can be preferentially selected for imaging, so as to reduce the values of a(λ) and b(λ), thereby reducing c(λ) and increasing the transmittance. Components such as a polarizer and a polarization analyzer are carried on the unmanned aerial vehicle, and the circularly polarized light illumination technology is adopted, combined with the polarization imaging algorithm, to better separate the target signal light and the background scattered light, and improve the contrast of the image and the clarity of the spectrum. At the same time, technologies such as multi-frame image averaging and spectral enhancement can be adopted to further improve the signal-to-noise ratio and stability of the signal, and reduce the signal fluctuation and noise interference caused by absorption and scattering. For example, weighted average processing of the spectral data of the same area collected multiple times can effectively reduce the influence of random noise, enhance the characteristic peaks of the spectral signal, and improve the recognition accuracy of microplastics.

[0028] Furthermore, the specific method of the multi-modal underwater microplastic semantic segmentation model in step three is as follows: The U-Net network structure is adopted, and the optical imaging data and the Raman laser data are combined for automatic identification and detection of microplastics. Among them, the optical imaging system extracts features such as the shape, texture, and reflection spectrum of microplastics in the water body through a hyperspectral camera, identifies the microplastic area and gives its precise location. The Raman lidar further determines the category of microplastics by comparing the Raman spectral data of microplastic particles obtained in the water body with the microplastic spectral information database.

[0029] Furthermore, the microplastic spatio-temporal distribution model in step four includes: Combining multi-modal remote sensing data, meteorological data, and watershed hydrological data to analyze the spatio-temporal distribution characteristics of microplastics in the water body. Using historical data and real-time monitoring data to capture the dynamic change process of microplastics in the water body, trace the source of microplastics, and analyze the migration and accumulation laws of microplastics under different seasons, weather conditions, and water flow changes. Using future climate change prediction models and watershed hydrological prediction models to predict the future distribution trend of microplastics, and help analyze its diffusion mode and future pollution risk under different environmental conditions.

[0030] The purpose of the present invention is to provide a water body microplastic detection system based on unmanned aerial vehicle optical and Raman spectroscopy technology applying the described water body microplastic detection method based on unmanned aerial vehicle optical and Raman spectroscopy technology. The water body microplastic detection system based on unmanned aerial vehicle optical and Raman spectroscopy technology includes:

[0031] A multi-modal data acquisition module, which is used to obtain the optical and Raman spectral data of water body microplastics;

[0032] The preprocessing module uses a solar flare correction model and an underwater light attenuation compensation model to improve the impact of strong sunlight reflection on the water surface and the absorption and scattering of water on the underwater imaging quality;

[0033] The microplastic identification module uses a multimodal underwater microplastic semantic segmentation model for automatic identification and detection of microplastics;

[0034] The spatio-temporal distribution modeling module is used to construct a microplastic spatio-temporal distribution model to extract the spatio-temporal distribution characteristics and future distribution trends of microplastics.

[0035] Combined with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by the present invention are as follows:

[0036] First, the present invention constructs an innovative multimodal microplastic pollution detection method and system through an unmanned aerial vehicle (UAV) platform integrating a Raman lidar and an optical imaging system. By integrating technologies such as near-infrared imaging and polarization imaging into the optical imaging system, clear and accurate image information is ensured for microplastic detection. At the same time, in the research and development of Raman lidar, based on the principle of Raman spectroscopy molecular recognition, the accuracy of detecting the chemical composition and particle size distribution of microplastic particles is significantly improved. By combining the UAV optical imaging system and hyperspectral Raman lidar technology, the detection accuracy and precision of microplastics in water can be effectively improved. The present invention uses deep learning algorithms, especially the U-Net network structure, for automatic identification and detection of microplastics, greatly improving the identification efficiency and robustness. Through the fusion of optical and Raman spectroscopy data, the system can accurately classify different types of microplastics, providing comprehensive and reliable technical support for water body microplastic pollution monitoring. Finally, by constructing a multi-dimensional spatio-temporal distribution model of microplastic pollution, the present invention can comprehensively analyze the migration, accumulation process, main sources and future distribution trends of microplastics in water bodies, with good predictive ability. This model can not only reflect the current pollution situation, but also provide a decision-making basis for tracing the source of water body microplastic pollution and pollution prevention and control measures, providing important reference for future water body environmental protection and treatment.

[0037] Second, the water body microplastic detection method and system based on UAV optical and Raman spectroscopy technology of the present invention significantly improve the automation degree and accuracy of microplastic detection, and promote the development of water body pollution monitoring technology. It provides strong support for tracing the sources of microplastic pollution, predicting pollution distribution and pollution control measures in different water areas, and has important social, economic and ecological values.

[0038] Third, as creative auxiliary evidence for the claims of the present invention, it is also reflected in the following important aspects:

[0039] (1) The technical solution of the present invention fills the technical gaps in the domestic and international industries:

[0040] At present, the detection of microplastics in water bodies faces many challenges. Traditional detection methods such as microscopic analysis and some spectral analysis have deficiencies in dealing with complex water body environments and achieving large-scale real-time monitoring. Existing technologies are difficult to achieve large-scale and real-time detection of microplastics while ensuring detection accuracy. The present invention integrates unmanned aerial vehicle (UAV) optical and Raman spectroscopy technologies and combines advanced deep learning methods to provide a new solution for the detection of microplastics in water bodies. Especially for the detection of microplastics in complex environments such as urban water bodies, the flexibility of UAVs can overcome the limitations of previous technologies and fill the gap in the lack of effective and comprehensive detection means in this field.

[0041] (2) Whether the technical solution of the present invention solves the technical problems that people have been eager to solve but have never succeeded in:

[0042] In the context of the increasingly serious pollution of microplastics in water bodies, accurately, quickly, and comprehensively detecting microplastics in water bodies and mastering their pollution status has always been a difficult problem in the field of environmental science. Previous technologies could not meet the requirements of large-scale, real-time, and spatio-temporal dynamic comprehensive analysis at the same time, making it difficult to effectively address the problem of microplastic pollution. The present invention constructs a method and system for detecting microplastics in water bodies, combines multi-modal remote sensing data with advanced deep learning technologies, can not only achieve efficient and accurate identification of microplastics in water bodies, but also comprehensively analyze the migration, accumulation, sources, and future trends of microplastics in water bodies through a spatio-temporal distribution model. Thus, it provides an accurate prediction of the dynamic changes of microplastics in different water body environments, helps scientists and decision-makers timely master the distribution characteristics and change trends of microplastic pollution, and then formulate more targeted and forward-looking pollution control plans to optimize water body environmental protection measures.

[0043] (3) Whether the technical solution of the present invention overcomes technical prejudice:

[0044] Traditional methods for monitoring microplastics in water bodies mostly rely on manual sampling and laboratory analysis. This method is not only costly, but also difficult to cover large areas of water bodies and has poor real-time performance. With the increasing complexity of water pollution problems, especially the increasing severity of microplastic pollution, traditional technologies have limitations in practical applications and are difficult to meet the requirements of efficient, real-time, comprehensive, and accurate monitoring at the same time. The present invention innovatively combines UAV optical imaging technology and hyperspectral Raman lidar technology to propose a new multi-modal method for detecting microplastics in water bodies, which can perform efficient and accurate real-time monitoring in large areas of water bodies, breaking through the limitations of traditional methods that rely on manual labor and single detection technologies. In addition, through a deep learning model for automatic identification and classification of microplastics and combining a spatio-temporal distribution model for dynamic analysis of pollution, it makes up for the deficiencies of previous technologies in the automatic detection and dynamic prediction of microplastic pollution. Description of the Drawings

[0045] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments of the present invention. Obviously, the following described drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0046] Figure 1 It is the architecture diagram of the water body microplastic detection system provided by the embodiments of the present invention;

[0047] Figure 2 It is the schematic diagram of the Raman spectroscopy detection hardware provided by the embodiments of the present invention;

[0048] Figure 3 It is the schematic diagram of the near-infrared spectroscopy assisted detection hardware provided by the embodiments of the present invention;

[0049] Figure 4 It is the schematic diagram of the solar flare correction model principle provided by the embodiments of the present invention;

[0050] Figure 5 It is the schematic diagram of the influence of the sea surface state on the entry of solar flare reflected light into the sensor field of view provided by the embodiments of the present invention;

[0051] Figure 6 It is the schematic diagram of the underwater light attenuation compensation model provided by the embodiments of the present invention; Detailed implementation manners

[0052] In order to make the purpose, technical solutions and advantages of the present invention clearer, the following further details the present invention in conjunction with embodiments. 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.

[0053] Aiming at the problems existing in the prior art, the present invention provides a method and system for detecting water body microplastics based on unmanned aerial vehicle optical and Raman spectroscopy technologies. The following describes the present invention in detail with reference to the drawings.

[0054] As Figure 1 shown, the water body microplastic detection system provided by the embodiments of the present invention includes the following modules:

[0055] S101, a multimodal data acquisition module, for obtaining optical and Raman spectroscopy data of water body microplastics;

[0056] S102, a preprocessing module, using the solar flare correction model and the underwater light attenuation compensation model to improve the influence of the strong reflection of sunlight on the water surface and the absorption and scattering of water on the underwater imaging quality;

[0057] S103, a microplastic identification module, uses a multimodal underwater microplastic semantic segmentation model for automatic identification and detection of microplastics;

[0058] S104, a spatio-temporal distribution modeling module, is used to construct a microplastic spatio-temporal distribution model to extract the spatio-temporal distribution characteristics and future distribution trends of microplastics.

[0059] The present invention first constructs a multimodal water body microplastic sensor based on optical imaging and hyperspectral Raman lidar. This sensor combines a hyperspectral camera carried by an unmanned aerial vehicle and a Raman spectroscopy detection system to achieve efficient collection of water body microplastics. The optical imaging part uses the hyperspectral camera to obtain the shape, texture, and spectral reflection characteristics of surface microplastics in the water body, while the Raman spectroscopy system irradiates the water body with a specific laser band, causing the microplastic particles to generate unique Raman scattering signals, and then comparing them with the microplastic spectral information library to identify different types of microplastics. The unmanned aerial vehicle, as a mobile collection platform, can automatically cruise within different water areas and transmit data in real time, providing high-quality optical and spectral information for subsequent analysis.

[0060] Due to the interference of solar flares on the water surface, which directly affects the accuracy of optical imaging data, the present invention uses a solar flare correction model for data preprocessing. The correction model first calculates the influence of solar flares on each spectral band, determines the radiation rate correction coefficient between different bands through a mathematical formula, and uses the near-infrared band as a reference to eliminate spectral distortion in overexposed areas. The influence of solar flares on the radiation rate is jointly determined by the Fresnel reflectivity, solar spectral intensity, and changes in the surface normal vector. Therefore, the present invention corrects the polarization effect of light through a probability density function and combines the angle relationship between the incident light and the reflected light to calculate the true radiation rate of each band, ensuring that the final imaging data accurately reflects the distribution of microplastics in the water body.

[0061] Due to the serious attenuation of underwater light, which affects the detection accuracy of Raman spectroscopy, the present invention proposes an underwater light attenuation compensation model and uses signal light enhancement technology to improve the resolution of water body microplastic signals. This model is based on the light intensity distribution of direct transmission and backscattering. By calculating the water body transmittance, it compensates for the energy loss caused by light absorption and scattering. The transmittance is determined by various optical parameters, including beam wavelength, absorption coefficient, and scattering coefficient, etc. The present invention establishes a mathematical model of spectral energy attenuation and combines a dehazing algorithm to enhance the contrast of underwater images and optimize the spectral characteristics of microplastic particles, enabling them to maintain high-precision identification capabilities in complex water environments.

[0062] To automatically extract microplastic regions, the present invention proposes a multi-modal underwater microplastic semantic segmentation model based on the U-Net network structure, which combines optical imaging data and Raman laser data to accurately segment the microplastic target regions. The U-Net network adopts an encoder-decoder architecture, extracts the shape, spectrum, and texture features of microplastics at the encoding end, and performs progressive upsampling at the decoding end to achieve pixel-level classification. The hyperspectral camera provides the spatial features of microplastics, while the Raman lidar provides information on the material composition. The combination of the two can effectively distinguish suspended particulate matter from real microplastics in water. This model is trained with a large amount of water microplastic data, increasing the recognition accuracy to 92%, significantly higher than the detection level of traditional methods.

[0063] The present invention further proposes a microplastic spatio-temporal distribution model, which combines multi-modal remote sensing data, meteorological data, and hydrological data to analyze the migration trends and accumulation characteristics of microplastics in water. Through historical data and real-time monitoring data, a dynamic evolution model of microplastic pollution is established to trace its source and evaluate the pollution diffusion path. The present invention uses a basin hydrological simulation and climate change prediction model, combined with parameters such as water flow velocity, wind speed, and rainfall, to infer the diffusion patterns of microplastics under different seasons and meteorological conditions. For example, in areas with fast water flow, microplastics are more likely to spread, while in areas with low flow velocity, they are more likely to deposit. This model can be used to predict future microplastic pollution risks and provide a scientific basis for water environment governance.

[0064] Through the integration of multiple technologies such as unmanned aerial vehicle (UAV) optical imaging, Raman spectroscopy detection, intelligent image segmentation, and spatio-temporal modeling, the present invention realizes the intelligent, high-precision, and large-scale coverage of water microplastic detection. Compared with traditional detection methods, the present invention does not require manual sampling and can achieve non-contact, fast, and real-time microplastic detection, which is applicable to different water environments such as lakes, rivers, and oceans. Due to the combination of multi-modal data fusion and deep learning algorithms, this method can still maintain high detection accuracy in complex water environments and provide decision-making support for microplastic pollution to governments, research institutions, and environmental protection departments. In the future, this technology can be further extended to fields such as marine plastic pollution monitoring and fishery ecological protection, providing more scientific solutions for global water pollution control.

[0065] As Figure 2 shown, the specific implementation process of the Raman spectroscopy detection hardware provided by the embodiment of the present invention is as follows:

[0066] First, the laser emits green light with a wavelength of 532 nm, which has strong penetrability in water. After being expanded and adjusted through a series of lenses and mirrors, it irradiates the target detection object. Then, the photons interact with the molecules of the detection object. According to the Raman scattering principle, the photons will exchange energy with the vibrational and rotational energy levels of the molecules, thereby generating Raman scattered light. Different molecular structures will lead to different energy exchange situations, and then generate Raman scattering spectra with molecular specificity, so as to realize the identification of different types of plastic microplastics. Next, at the receiving end, the Raman scattered light is received by a spectrometer. First, it passes through a filter to filter out stray light and unwanted wavelength components and reaches the lens to collect the scattered light. Then, it passes through a spectrometer with astigmatism compensation and an entrance slit. The role of the slit is to select a specific light region, determine the linear field of view of the system, and only allow light in a specific direction and range to pass through to improve the spectral resolution and reduce the interference of stray light. Finally, the ICCD camera gates and records the optical signal on the nanosecond time scale. Finally, through the grating dispersion of the spectrometer, the light at each point in the field of view is decomposed into spectra of different wavelengths. By analyzing these spectra, different plastic types can be identified.

[0067] As Figure 3 shown, the specific implementation process of the near-infrared spectrum-assisted detection hardware provided by the embodiment of the present invention is as follows:

[0068] The overall structure of the near-infrared light-assisted detection hardware is similar to that of the Raman spectroscopy detection hardware. The difference lies in taking advantage of the fact that different chemical bonds (such as C-H, O-H, N-H, etc.) have different absorption characteristics in the near-infrared region, which can reflect the chemical composition and structural information of the sample, thus providing more comprehensive sample information. First, near-infrared light with a wavelength range of 900 nm - 1700 nm is emitted. This wavelength range needs to be selected and optimized according to the chemical bonds, chemical composition, and structural information contained in the known plastic types so that it can effectively excite the characteristic absorption of the chemical bonds in the sample. Then, after being expanded and adjusted by a series of lenses and mirrors, the near-infrared light irradiates the target detection object. The photons interact with the chemical bonds in the detection object molecules. Based on the different absorption characteristics of the chemical bonds in the near-infrared region, part of the photon energy is absorbed, generating characteristic absorption spectra. These absorption spectra carry the chemical composition and structural information of the sample and can be used as an important basis for identifying and analyzing the sample. Next, at the receiving end, the optical signal is received by a dedicated near-infrared spectrometer. First, stray light and irrelevant wavelength components are removed through a filter, and then the light is focused onto the detector by a focusing lens. The detector converts the optical signal into an electrical signal for preliminary processing. Finally, signal processing algorithms and chemometric methods are used to further analyze and process the electrical signal. Combining with the known near-infrared absorption characteristic database of chemical bonds, the chemical composition and structural information in the sample are identified, which complement and verify the Raman spectroscopy detection results, so as to more accurately identify and analyze plastic microplastics and other possible substance components.

[0069] As Figure 4 shown, the specific implementation of the solar flare correction model provided by the embodiment of the present invention is as follows:

[0070]

[0071] where L i is the radiance of band i before correction, and L i ′ is the radiance of band i after correction; L j is the radiance of the NIR band, is the average radiance of the NIR band; α ij is the correction coefficient, which is determined by Equation (11); in Equation (11), the parameter ρ ij is the covariance between band i and band j (near-infrared band), which is used to measure the linear correlation degree between the radiance sample data of the two bands, and ρ jj is the covariance of the NIR band itself; where ρ ij is obtained by Equation (12), and in Equation (12), the parameters L i,n and L j,nThey respectively represent the nth radiance sample values of band i and band j, and N represents the number of samples. The first part on the right side of Equation (12) calculates the average value of the product of the corresponding sample values of the two bands, and the second part on the right side is the product of the average values of the samples of the two bands.

[0072] Based on this model, the drone carries a hyperspectral sensor and flies above the water surface. The sun's rays irradiate the water surface at a specific angle, generating solar flares. The sensor receives the reflected light from the water surface, which contains the radiance data of different bands affected by solar flares (L i,n and L j,n are respectively the nth radiance sample values of band i and band j). Here, bands i and j are selected as bands that are sensitive to solar flares and are correlated with each other, namely the 532 nm green light band and the near-infrared band (NIR); using the collected radiance sample data of different bands, the correction coefficient α ij is obtained to measure the degree to which the radiance of band i is affected by the radiance of band j. Finally, given the radiance of band i before correction, the radiance of band j, and the average radiance, the radiance of band i is corrected according to the correction model to obtain the corrected radiance L i '. By this step, the influence of solar flares on the radiance of band i is eliminated or reduced, so that the imaging data can more truly reflect the actual situation above and below the water surface. Finally, imaging processing is performed using the corrected radiance data to obtain a more accurate and real image after correction.

[0073] As Figure 5 shown, the influence of the sea surface state on the solar flare reflected light entering the sensor's field of view provided by the embodiment of the present invention is specifically implemented as follows:

[0074] L sg = c(λ)p(η xs , η ys ) (13)

[0075] Among them, L sg is the contribution of solar flares to the radiance. c(λ) is a quantity related to the wavelength, which includes the Fresnel reflectivity of the surface and the spectral intensity of the incident sunlight; and the probability density function p(η xs , η ys ) based on the surface slope is used to describe the influence of the sea surface state on the solar flare reflected light entering the sensor's field of view in the solar flare correction algorithm. In the actual water surface environment, the water surface is not completely flat, but there are waves and undulations of various scales, which cause the normal direction of each point on the water surface to change continuously. And solar flares are generated when the sun's rays are reflected by the water surface and enter the sensor's field of view, which means that p(η xs , η ys) is evaluated at the slope required to specularly reflect sunlight into the sensor's field of view. It reflects the probability that, at specific solar and sensor angles, the portion of the water surface with a specific slope will reflect sunlight into the sensor's field of view. For example, at a certain moment, with the sun at a specific position in the sky and the sensor observing the water surface at a specific angle, the likelihood of different-slope sea surface microfacets reflecting sunlight into the sensor's field of view varies. It is one of the key factors in calculating the contribution L sg of solar flare to the radiance. Since this function is the same for all bands, there is a correlation between solar flare signals in different bands, and this correlation is used in subsequent calculations of the correction coefficient α ij , thereby achieving correction for the impact of solar flare. Among them, the parameter η xs in Equation (13), η ys is obtained through the following Equation (14), while n x , n y , n z are the components of the surface normal vector n, obtained through Equation (15), k s points to the sun, and k o points to the sensor.

[0076]

[0077] Through the above Equations (13), (14), and (15), the influence of the sea surface state on the entry of solar flare reflected light into the sensor's field of view can be obtained. When calculating the inter-band covariance to determine the correction coefficient α ij , p(η xs , η ys ) plays a role in connecting solar flare signals in different bands, enabling solar flare correction for other bands through the near-infrared band.

[0078] As Figure 6 shown, the underwater light attenuation compensation model provided by the embodiment of the present invention is specifically implemented as follows:

[0079] I(x,y) = D(x,y) + B(x,y) = J(x,y)·t(x,y) + A ∞ (1 - t(x,y)) (16)

[0080] where D(x,y) represents direct transmission, and B(x,y) represents backscattering. I(x,y) is the captured image, J(x,y) is the image without scattering, representing the image formed under ideal conditions when light is not affected by water body scattering. A ∞is the uniform background light, which is mainly formed by the combined factors of the weak luminescence of the water body itself, the multiple scattering of ambient light in the water body, and the inherent noise of the sensor. t(x,y) is the transmittance, which describes the attenuation of light during propagation and is usually related to the depth of field (DOP) and the attenuation coefficient. The transmittance in Equation (16) is obtained from the following formula:

[0081] t(x,y) = e -c(λ)d(x,y) (17)

[0082] c(λ) = a(λ) + b(λ) (18)

[0083] where d(x,y) represents the DOP, i.e., the distance between the target and the camera, c(λ) represents the total attenuation coefficient, λ represents the wavelength of the light beam, and a(λ), b(λ) respectively represent the absorption coefficient and the scattering coefficient.

[0084] In the figure, the light rays emitted by the sun represent natural light and shoot towards the water surface in the form of parallel rays. These natural lights will undergo different propagation path changes after reaching the water surface. Below the water surface, there are water impurities and microplastics. After natural light enters the water body, it will interact with these substances, resulting in phenomena such as light scattering and absorption. According to Equation (16), D(x,y) represents the directly transmitted light. It is the part of the light that reaches the sensor directly without being scattered or absorbed during the propagation of the light ray in the water, and its intensity is related to the non-scattered image and the transmittance; the backscattered light B(x,y) is the part of the light that is scattered in all directions after interacting with water impurities, microplastics, etc. in the water, and a part of the scattered light propagates in the reverse direction and finally reaches the sensor. It is related to the uniform background light A ∞ and the transmittance t(x,y), where A ∞ is mainly formed by the combined factors of the weak luminescence of the water body itself, the multiple scattering of ambient light in the water body, and the inherent noise of the sensor. t(x,y) describes the attenuation of light during propagation, and its value determines the ratio of the directly transmitted light to the backscattered light, which has an important impact on the imaging quality; the total attenuation coefficient c(λ) is composed of the absorption coefficient and the scattering coefficient. The attenuation of the light beam involves two basic processes of absorption and scattering, and the value of the coefficient is related to the wavelength of the light beam. The degrees of absorption and scattering are different at different wavelengths. Selecting green light for imaging can reduce c(λ), increase the transmittance, enhance the intensity of the directly transmitted light, and reduce the attenuation of light. Through this model, the attenuation of light can be effectively reduced and the imaging quality can be improved.

[0085] The technical solutions for achieving the objectives of the present invention mainly include the following:

[0086] (1) Design an unmanned aerial vehicle optical and Raman spectroscopy technology platform for monitoring microplastics in water bodies

[0087] The present invention utilizes a drone platform. By carrying an optical imaging system and a Raman lidar, it can monitor different water areas, with mobility and flexibility. The optical imaging system is responsible for obtaining image information of microplastics in water, while the Raman lidar can deeply detect the internal structure of water and the chemical characteristics of microplastics.

[0088] (2) Establish a solar flare correction model to address the impact of solar flares on imaging quality.

[0089] Considering the propagation path of sunlight, based on the position and angle of the sun (solar zenith angle) and the observation angle of the drone (drone observation zenith angle), combined with the optical properties of the atmosphere and water, establish a geometric model of light propagation and reflection. Using the laws of reflection and refraction of light, determine the reflection of sunlight on the water surface and the path of the reflected light entering the sensor, thereby achieving effective correction of solar flares.

[0090] (3) Establish an underwater light attenuation compensation model to address the problem of light attenuation in water affecting imaging quality.

[0091] Establish an underwater light attenuation compensation model. Select a suitable spectral band (such as 532nm green light) as the light source, and configure components such as a polarizer and a polarization analyzer in the optical imaging system of the drone. Adopt circularly polarized light illumination technology, combined with a polarization imaging algorithm, to separate the target signal light and the background scattered light, improving the contrast of the image and the clarity of the spectrum. At the same time, techniques such as multi-frame image averaging and spectral enhancement can be used to further improve the signal-to-noise ratio and stability of the signal, reducing the degradation of imaging quality caused by absorption and scattering.

[0092] This system uses a hyperspectral camera and a Raman lidar to construct a multimodal water microplastic sensor to achieve the acquisition of optical and spectral data of water microplastics. The hyperspectral camera covers the visible to near-infrared spectral range, can capture the shape, texture, and spectral reflection characteristics of microplastic particles in water, and improves the spectral resolution through narrowband filters. The Raman lidar uses a 532nm or 785nm laser band to excite microplastic particles in water, generating unique Raman scattering signals, and combines a microplastic spectral matching algorithm for material identification. The drone carries this sensor and can conduct mobile monitoring in different water areas, obtaining real-time optical imaging and spectral signals of water microplastics, providing high-quality multimodal input for subsequent data processing.

[0093] In the actual water environment, due to the influence of solar flares and underwater light attenuation, the original collected data may have problems of spectral distortion and signal attenuation. Therefore, this system introduces a solar flare correction model. By calculating the radiation rate correction coefficients of each spectral band and using the near-infrared band as a reference, it compensates for the overexposure phenomenon in the flare area and improves the accuracy of optical images. At the same time, an underwater light attenuation compensation model is adopted. Based on the direct transmission and backscattering theories, the water transmittance is calculated, and the dehazing enhancement technology is used to optimize the spectral contrast of underwater microplastics, thereby restoring the true spectral characteristics of microplastics and making the microplastic detection results more accurate.

[0094] The microplastic detection module of this system is based on the U-Net network structure. By combining optical imaging data and Raman spectroscopy data, it realizes pixel-level segmentation and automatic recognition of water microplastics. The U-Net network adopts an encoder-decoder architecture. At the encoding end, it extracts the shape, spectral characteristics, and texture information of microplastics. At the decoding end, it performs step-by-step upsampling to accurately calibrate the microplastic area. Optical imaging data is used to extract spatial features, and Raman spectroscopy data is used for material composition identification. The combination of the two can effectively distinguish suspended particulate matter from microplastic targets. This detection method is trained through deep learning with a large amount of water microplastic data to improve the recognition accuracy and maintain high detection performance in complex water environments.

[0095] This system constructs a spatio-temporal distribution model of water microplastics through the data analysis and prediction module. By combining multi-modal remote sensing data, meteorological data, and watershed hydrological data, it analyzes the migration path, accumulation characteristics, and pollution sources of microplastics. Using historical data and real-time monitoring data, it captures the dynamic changes of water microplastics, and through the watershed hydrological simulation and climate change prediction models, it speculates on the diffusion patterns and future pollution trends of microplastics under different seasons and meteorological conditions. In addition, this module can combine variables such as water flow velocity, wind speed, and rainfall to establish a microplastic migration model, providing pollution control and decision-making support for environmental management departments and optimizing water pollution prevention and control strategies.

[0096] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. Those of ordinary skill in the art can understand that the above devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code is provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuits of programmable hardware devices such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, or can be implemented by a combination of the above hardware circuits and software, such as firmware.

[0097] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall all be covered by the protection scope of the present invention.

Claims

1. A method for detecting microplastics in water based on drone optics and Raman spectroscopy technology, characterized in that: The following steps are involved: (1) Collect water microplastic data using a multimodal water microplastic sensor based on optical imaging and hyperspectral Raman lidar; (2) Preprocessing the collected water microplastic data through the solar flare correction model and underwater light attenuation compensation model; (3) Use a multimodal underwater microplastic semantic segmentation model to detect water microplastics in the preprocessed data; (4) Construct a spatiotemporal distribution model and conduct a comprehensive analysis of the distribution trends of microplastics in water bodies.

2. The method for detecting microplastics in water based on drone optics and Raman spectroscopy technology as claimed in claim 1, characterized in that: In step (1), the multimodal water microplastic sensor includes a hyperspectral camera and a Raman lidar. The hyperspectral camera is used to obtain the shape, texture and spectral reflectance characteristics of microplastics in the water, and the Raman lidar is used to obtain Raman spectral data of microplastics in the water.

3. The method for detecting microplastics in water based on drone optics and Raman spectroscopy technology as claimed in claim 1, characterized in that: In step (2), the solar flare correction model is based on the spectral band emissivity correction coefficient and uses the near-infrared band as a reference to calculate the impact of solar flares on each band to correct the spectral distortion of the overexposed area.

4. The method for detecting microplastics in water based on drone optics and Raman spectroscopy technology as claimed in claim 1, characterized in that: In step (2), the underwater light attenuation compensation model adopts signal light enhancement technology to compensate for the energy loss caused by light absorption and scattering by calculating the water body transmittance, so as to enhance the spectral characteristics of underwater microplastics.

5. The method for detecting microplastics in water based on drone optics and Raman spectroscopy technology as claimed in claim 1, characterized in that: In step (3), the multimodal underwater microplastic semantic segmentation model is based on a U-Net network structure, combined with optical imaging data and Raman spectroscopy data, and uses an encoder-decoder architecture for pixel-level classification to extract the microplastic target area in the water body.

6. The method for detecting microplastics in water based on drone optics and Raman spectroscopy technology as claimed in claim 1, characterized in that: In step (4), the spatiotemporal distribution model combines multimodal remote sensing data, meteorological data and river basin hydrological data to analyze the migration and accumulation patterns of microplastics in water bodies, and uses river basin hydrological simulation and climate change prediction models to predict the future distribution trend of microplastics in water bodies.

7. A water microplastic detection system based on drone optics and Raman spectroscopy technology, characterized in that: include: (1) Multimodal water microplastic sensor, including a hyperspectral camera and a Raman lidar. The hyperspectral camera is used to obtain the shape, texture and spectral reflectance characteristics of microplastics in water, and the Raman lidar is used to obtain Raman spectral data of microplastics in water. (2) A data preprocessing module is used to perform solar flare correction and underwater light attenuation compensation on the collected water microplastic data. A solar flare correction model is used to correct the solar flare effect in the spectral data, and an underwater light attenuation compensation model is used to enhance the spectral characteristics of underwater microplastics. (3) Microplastic detection module, including a multimodal underwater microplastic semantic segmentation model, which is based on the U-Net network structure, combines optical imaging data and Raman spectroscopy data, and uses an encoder-decoder architecture for pixel-level classification to extract microplastic target areas in water bodies; (4) Data analysis and prediction module, which is used to construct a spatiotemporal distribution model of microplastics. It combines multimodal remote sensing data, meteorological data, and river basin hydrological data to analyze the migration and accumulation patterns of microplastics in water bodies, and uses river basin hydrological simulation and climate change prediction models to predict the future distribution trend of microplastics in water bodies.

8. The water microplastic detection system based on drone optics and Raman spectroscopy technology as claimed in claim 7 is characterized in that: The wavelength of the hyperspectral camera covers the visible light to near-infrared spectrum range, and a narrow-band filter is used to improve the spectral resolution of microplastic targets. The Raman lidar uses a 532nm or 785nm laser band and is combined with a spectral matching algorithm to identify different types of microplastics.

9. The water microplastic detection system based on drone optics and Raman spectroscopy technology as claimed in claim 7, characterized in that: The data preprocessing module adopts a spectral offset correction algorithm, uses the near-infrared band to calculate the solar flare correction coefficient of each band, and combines the spectral denoising algorithm to optimize the data quality; the underwater light attenuation compensation calculates the water body transmittance based on direct transmission and backscattering, and optimizes the contrast of microplastic targets through defogging enhancement technology.

10. The water microplastic detection system based on drone optics and Raman spectroscopy technology as claimed in claim 7, characterized in that: The data analysis and prediction module uses a machine learning algorithm to train historical microplastic monitoring data, combines water flow velocity, wind speed, and rainfall parameters to establish a microplastic migration model, and combines watershed hydrological simulation and climate change prediction models to infer the diffusion patterns of microplastics and future pollution trends under different seasons and meteorological conditions.

Images