A method and system for analyzing marine oil spill ecotoxicity based on image recognition

Abstract

The present invention belongs to the technical field of ecotoxicity analysis, and discloses a method and system for analyzing the ecotoxicity of marine oil spills based on image recognition. The method comprises the following steps: collecting remote sensing image data of historical marine oil spill accidents, historical ocean current data, and historical marine water quality monitoring data, and using image recognition algorithms to construct a marine oil spill area prediction model; obtaining a number of petroleum ecotoxicity measurement indicators, and performing data screening; constructing a marine oil spill ecotoxicity analysis training sample set; using a deep learning algorithm to construct a marine oil spill ecotoxicity analysis model; collecting real-time marine oil spill accident remote sensing image data, real-time ocean current data, and real-time marine water quality monitoring data, and predicting the marine oil spill area; performing data screening according to a number of petroleum ecotoxicity measurement indicators; and performing marine oil spill ecotoxicity analysis. The present invention solves the problems of long time consumption, high cost, and difficulty in real-time monitoring in the prior art.

Description

Technical Field

[0001] The invention belongs to the technical field of ecotoxicity analysis, and in particular relates to a method and system for analyzing the ecotoxicity of marine oil spills based on image recognition. Background Art

[0002] In recent years, with the rapid development of the economy, the petrochemical industry has grown stronger and the risk of marine oil spills has also increased. Marine oil spills cause serious damage to the ecological environment, and ecotoxicity analysis is an important means to assess the impact of oil spills. At present, marine oil spill ecotoxicity analysis mainly relies on laboratory testing and on-site investigations, which are time-consuming, costly, and difficult to monitor in real time. Therefore, it is of great significance to seek a fast, efficient, and real-time marine oil spill ecotoxicity analysis method. Summary of the invention

[0003] In order to solve the problems of long time consumption, high cost and difficulty in real-time monitoring in the prior art, the present invention aims to provide a method and system for analyzing the ecotoxicity of marine oil spills based on image recognition.

[0004] The technical solution adopted by the present invention is:

[0005] A method for analyzing the ecotoxicity of marine oil spills based on image recognition comprises the following steps:

[0006] Collecting the historical remote sensing image data of marine oil spill accidents in several sea areas, as well as the historical ocean current data and historical marine water quality monitoring data of the corresponding sea areas, and constructing a marine oil spill area prediction model based on the remote sensing image data of several historical marine oil spill accidents using image recognition algorithms to obtain the corresponding historical marine oil spill area prediction data;

[0007] According to the petroleum ecotoxicity test data, a number of petroleum ecotoxicity measurement indicators are obtained, and according to the petroleum ecotoxicity measurement indicators, a number of historical marine water quality monitoring data are screened to construct a number of corresponding historical petroleum ecotoxicity impact data;

[0008] Integrate historical marine oil spill area prediction data, historical ocean current data, and historical petroleum ecotoxicity impact data in the same sea area to construct a marine oil spill ecotoxicity analysis training sample set;

[0009] Based on the marine oil spill ecotoxicity analysis training sample set, a marine oil spill ecotoxicity analysis model was constructed using a deep learning algorithm;

[0010] Collect real-time remote sensing image data of marine oil spill accidents, real-time ocean current data and real-time marine water quality monitoring data of the current sea area, and input the real-time remote sensing image data of marine oil spill accidents into the marine oil spill area prediction model to predict the marine oil spill area and obtain the corresponding real-time marine oil spill area prediction data;

[0011] According to several petroleum ecotoxicity measurement indicators, data screening is performed on several petroleum ecotoxicity measurement indicators to obtain real-time petroleum ecotoxicity impact data;

[0012] The real-time marine oil spill area prediction data, real-time ocean current data and real-time petroleum ecotoxicity impact data of the current sea area are input into the marine oil spill ecotoxicity analysis model to conduct marine oil spill ecotoxicity analysis and obtain the corresponding marine oil spill ecotoxicity analysis results.

[0013] Furthermore, the historical remote sensing image data of marine oil spill accidents in several sea areas, and the historical ocean current data and historical marine water quality monitoring data of the corresponding sea areas are collected, and according to the several historical remote sensing image data of marine oil spill accidents, an image recognition algorithm is used to construct a marine oil spill area prediction model to obtain the corresponding several historical marine oil spill area prediction data, including the following steps:

[0014] Collect historical marine oil spill accident remote sensing image data of several sea areas, as well as historical ocean current data and historical marine water quality monitoring data of the corresponding sea areas;

[0015] Preprocessing several historical marine oil spill accident remote sensing image data to obtain a preprocessed historical marine oil spill accident remote sensing image dataset;

[0016] Based on the pre-processed historical marine oil spill accident remote sensing image dataset, an image semantic segmentation sub-model is constructed, and the corresponding historical marine area segmentation image data are obtained;

[0017] According to several historical ocean area segmentation image data, a marine oil spill target recognition sub-model is constructed, and several corresponding historical marine oil spill target area image data are obtained;

[0018] Based on several historical marine oil spill target area image data, a marine oil spill area prediction sub-model is constructed, and several corresponding historical marine oil spill area prediction data are obtained;

[0019] The image semantic segmentation sub-model, the marine oil spill target recognition sub-model and the marine oil spill area prediction sub-model are integrated to obtain the marine oil spill area prediction model.

[0020] Furthermore, the preprocessing includes cropping, translation, Gaussian denoising, grayscale processing, normalization processing and labeling in sequence.

[0021] Furthermore, the image semantic segmentation sub-model is constructed based on the GSCNN algorithm, and the image semantic segmentation sub-model includes a first input layer, a rule flow module, a shape flow module, a fusion module and a first output layer, the first input layer is connected to the rule flow module and the shape flow module respectively, the rule flow module and the shape flow module are both connected to the fusion module, and the fusion module is connected to the first output layer;

[0022] The marine oil spill target recognition sub-model is constructed based on the YOLOv8 algorithm, and the marine oil spill target recognition sub-model includes a second input layer, a Backbone module, a Neck module, a Head module, a Loss module and a second output layer connected in sequence;

[0023] The marine oil spill area prediction sub-model is based on the LFMCNN algorithm, and the marine oil spill area prediction sub-model includes a third input layer, a first convolution module, a downsampling module, a second convolution module, a third convolution module, a fully connected layer and a third output layer which are connected in sequence.

[0024] Furthermore, according to the petroleum ecotoxicity test data, a number of petroleum ecotoxicity measurement indicators are obtained, and according to the number of petroleum ecotoxicity measurement indicators, a number of historical marine water quality monitoring data are screened to construct a number of corresponding historical petroleum ecotoxicity impact data, including the following steps:

[0025] Based on the PCA method, the petroleum ecotoxicity test data are converted into a petroleum ecotoxicity test data matrix;

[0026] Obtain the transformation matrix of the petroleum ecotoxicity test data matrix;

[0027] According to the petroleum ecotoxicity test data matrix and the corresponding conversion matrix, a principal component matrix is ​​obtained; the principal component matrix includes several candidate principal components;

[0028] According to the cumulative variance contribution rate of all candidate principal components, if it exceeds 85%, the corresponding candidate principal components are used as principal components to obtain several principal component indices related to petroleum ecotoxicity, and the several principal component indices are used as several petroleum ecotoxicity measurement indices;

[0029] According to several petroleum ecotoxicity measurement indicators, several historical marine water quality monitoring data were screened and several corresponding historical petroleum ecotoxicity impact data were constructed.

[0030] Furthermore, based on the marine oil spill ecotoxicity analysis training sample set, a marine oil spill ecotoxicity analysis model is constructed using a deep learning algorithm, including the following steps:

[0031] The marine oil spill ecotoxicity analysis training sample set is divided into a model training sample set and a model testing sample set in a ratio of 7:3;

[0032] Based on the model training sample set, the DBN-IFWA algorithm is used to perform optimization training and construct the initial marine oil spill ecotoxicity analysis model;

[0033] Input the model test sample set into the initial marine oil spill ecotoxicity analysis model, conduct marine oil spill ecotoxicity analysis, and obtain the model prediction accuracy;

[0034] If the model prediction accuracy is greater than the preset model prediction accuracy threshold, the optimal marine oil spill ecotoxicity analysis model is output, otherwise, the optimization training continues.

[0035] Furthermore, the marine oil spill ecotoxicity analysis model includes a marine oil spill ecotoxicity analysis module constructed based on a DBN algorithm and a model optimization module constructed based on an IFWA optimization algorithm, which are connected in sequence.

[0036] Furthermore, based on the model training sample set, the DBN-IFWA algorithm is used to perform optimization training to construct an initial marine oil spill ecotoxicity analysis model, including the following steps:

[0037] The initial network parameters of the DBN network are used as the optimization target of the IFWA optimization algorithm, and the algorithm parameters of the IFWA optimization algorithm are initialized; the algorithm parameters include the IFWA population parameters and the maximum number of iterations;

[0038] The optimization target is the position of the IFWA individual, and the fitness function of the IFWA individual is set;

[0039] According to the IFWA population parameters, the IFWA population is initialized using the Tent-Logistic-Cosine chaotic mapping sequence to obtain the initialized IFWA population;

[0040] According to the fitness function, the fitness value of each IFWA individual in the initialized IFWA population is obtained;

[0041] Get the explosion radius and number of sparks of each IFWA individual in the initialized IFWA population;

[0042] According to the explosion radius and the number of sparks of each initial IFWA individual in the initialized IFWA population, fireworks are exploded to obtain an updated IFWA population;

[0043] Use the Gaussian mutation algorithm to perform Gaussian mutation on the initialized IFWA population to generate a Gaussian mutated IFWA population;

[0044] Use the dynamic reverse learning algorithm to perform dynamic reverse learning on the initialized IFWA population to obtain the reverse IFWA population;

[0045] Obtain the fitness values ​​of the IFWA individuals of the updated IFWA population, the Gaussian-mutated IFWA population, and the reverse IFWA population, and take the IFWA individual with the minimum fitness value as the optimal IFWA individual;

[0046] If the number of iterations reaches the threshold or the fitness value of the optimal IFWA individual meets the requirements, the optimal solution corresponding to the current optimal IFWA individual is output to obtain the optimal initial network parameters of the DBN network;

[0047] According to the optimal initial network parameters, the DBN network is set up, and the model training sample set is used for optimization training to construct the initial marine oil spill ecotoxicity analysis model.

[0048] A marine oil spill ecotoxicity analysis system based on image recognition is used to implement a marine oil spill ecotoxicity analysis method. The system includes a prediction model construction unit, an indicator acquisition unit, a data integration unit, an analysis model construction unit, an oil spill area prediction unit, a data screening unit and an ecotoxicity analysis unit connected in sequence;

[0049] The prediction model building unit is used to collect the historical remote sensing image data of marine oil spill accidents in several sea areas, and the historical ocean current data and historical marine water quality monitoring data of the corresponding sea areas, and to build a marine oil spill area prediction model based on the remote sensing image data of several historical marine oil spill accidents by using an image recognition algorithm to obtain the corresponding several historical marine oil spill area prediction data;

[0050] An index acquisition unit is used to obtain a number of petroleum ecotoxicity measurement indicators based on petroleum ecotoxicity test data, and to screen a number of historical marine water quality monitoring data based on the number of petroleum ecotoxicity measurement indicators to construct a number of corresponding historical petroleum ecotoxicity impact data;

[0051] The data integration unit is used to integrate the historical marine oil spill area prediction data, historical ocean current data and historical petroleum ecotoxicity impact data of the same sea area to construct a training sample set for marine oil spill ecotoxicity analysis;

[0052] An analysis model building unit, used to build a marine oil spill ecotoxicity analysis model using a deep learning algorithm based on a marine oil spill ecotoxicity analysis training sample set;

[0053] The oil spill area prediction unit is used to collect real-time remote sensing image data of marine oil spill accidents, real-time ocean current data and real-time marine water quality monitoring data of the current sea area, and input the real-time remote sensing image data of marine oil spill accidents into the marine oil spill area prediction model to perform marine oil spill area prediction and obtain corresponding real-time marine oil spill area prediction data;

[0054] A data screening unit, for screening data of a number of petroleum ecotoxicity measurement indicators according to a number of petroleum ecotoxicity measurement indicators, and obtaining real-time petroleum ecotoxicity impact data;

[0055] The ecotoxicity analysis unit is used to input the real-time marine oil spill area prediction data, real-time ocean current data and real-time petroleum ecotoxicity impact data of the current sea area into the marine oil spill ecotoxicity analysis model, conduct marine oil spill ecotoxicity analysis, and obtain the corresponding marine oil spill ecotoxicity analysis results.

[0056] The beneficial effects of the present invention are:

[0057] The present invention provides a marine oil spill ecotoxicity analysis method and system based on image recognition. The method and system are combined with remote sensing image data of marine oil spill accidents, use image recognition algorithms, build a marine oil spill area prediction model, obtain the region and area of ​​marine oil spill accidents, and improve real-time and practicality by real-time monitoring of specific on-site conditions of marine oil spill accidents. The scope of marine oil spills can be grasped at the first time, providing data support for subsequent ecotoxicity analysis. According to petroleum ecotoxicity test data, a number of petroleum ecotoxicity measurement indicators are obtained, and qualitative and quantitative analysis of petroleum ecotoxicity is performed, thereby improving the accuracy of marine oil spill ecotoxicity analysis. A marine oil spill ecotoxicity analysis model is built using a deep learning algorithm, and automatic analysis is performed according to real-time marine oil spill area prediction data, real-time ocean current data, and real-time petroleum ecotoxicity impact data of the current sea area, thereby avoiding reliance on laboratory testing and on-site investigations, reducing material, human and time cost investment, and improving the value of marine oil spill ecotoxicity analysis in actual accidents.

[0058] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flowchart of the marine oil spill ecotoxicity analysis method based on image recognition in the present invention.

[0060] Figure 2 It is a structural block diagram of the marine oil spill ecotoxicity analysis system based on image recognition in the present invention. DETAILED DESCRIPTION

[0061] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments.

[0062] Embodiment 1:

[0063] like Figure 1 As shown, this embodiment provides a marine oil spill ecotoxicity analysis method based on image recognition, comprising the following steps:

[0064] Collecting the historical remote sensing image data of marine oil spill accidents in several sea areas, and the historical ocean current data and historical marine water quality monitoring data of the corresponding sea areas, and constructing a marine oil spill area prediction model based on the remote sensing image data of several historical marine oil spill accidents, using image recognition algorithm, and obtaining the corresponding several historical marine oil spill area prediction data, including the following steps:

[0065] Collect historical marine oil spill accident remote sensing image data of several sea areas, as well as historical ocean current data and historical marine water quality monitoring data of the corresponding sea areas;

[0066] Preprocessing several historical marine oil spill accident remote sensing image data to obtain a preprocessed historical marine oil spill accident remote sensing image dataset; the preprocessing includes cropping, translation, Gaussian denoising, grayscale processing, normalization processing and label adding in sequence;

[0067] Based on the pre-processed historical marine oil spill accident remote sensing image dataset, an image semantic segmentation sub-model is constructed, and the corresponding historical marine area segmentation image data are obtained;

[0068] The image semantic segmentation sub-model is constructed based on the Gated-Shape Convolutional Neural Networks (GSCNN) algorithm, and the image semantic segmentation sub-model includes a first input layer, a rule flow module, a shape flow module, a fusion module and a first output layer. The first input layer is connected to the rule flow module and the shape flow module respectively, the rule flow module and the shape flow module are both connected to the fusion module, and the fusion module is connected to the first output layer.

[0069] The rule flow module of the image semantic segmentation sub-model adopts the classic convolutional neural network structure for processing semantic region information. The shape flow module consists of multiple gated convolutional layers. The low-level feature map in the rule flow module is used to process the region boundary information. The outputs of the two modules are sent to a fusion module. In the fusion module, the outputs of the rule flow module and the shape flow module are combined using the Atrous Special Pyramid Pooling (ASPP) structure. ASPP helps preserve multi-scale context information. Finally, the fusion module generates object semantic regions with defined boundaries to obtain accurately segmented semantic segmentation image data.

[0070] The remote sensing image data of marine oil spill accidents can be accurately divided into building areas, marine areas, oil production areas, riverside areas, and island areas, and the focus areas can be placed on marine areas and oil production areas where marine oil spill accidents occur, thus improving the accuracy and efficiency of subsequent model predictions;

[0071] According to several historical ocean area segmentation image data, a marine oil spill target recognition sub-model is constructed, and several corresponding historical marine oil spill target area image data are obtained;

[0072] The marine oil spill target recognition sub-model is constructed based on the YOLOv8 algorithm, and the marine oil spill target recognition sub-model includes a second input layer, a Backbone module, a Neck module, a Head module, a Loss module and a second output layer connected in sequence;

[0073] The Backbone module refers to the cspdarknet-53 network, which is a cspdarknet structure; the Neck module mainly includes the Path Aggregation Network (PANet) structure, and the PANet part is upsampling fusion first and then downsampling fusion; the Head module separates classification and detection, using a decoupled head; the Loss module uses positive and negative matching of samples instead of IoU matching in the loss function; it has advantages in speed, accuracy, adaptability, compatibility and easy understanding, making it a deep learning model with broad application prospects in the field of computer vision;

[0074] Based on several historical marine oil spill target area image data, a marine oil spill area prediction sub-model is constructed, and several corresponding historical marine oil spill area prediction data are obtained;

[0075] The marine oil spill area prediction sub-model is based on a low level feature fusion multi kernel convolutional neural network (LFMCNN) algorithm, and the marine oil spill area prediction sub-model includes a third input layer, a first convolution module, a downsampling module, a second convolution module, a third convolution module, a fully connected layer and a third output layer connected in sequence;

[0076] The model can obtain information of different scales of the input image, and then downsample the original image and fuse it with the feature map obtained after the convolution operation of the first convolution module to obtain more space, color, contour and other information. After the two convolution modules are used for feature extraction and final prediction, the accuracy and efficiency of area prediction are greatly improved.

[0077] Integrate the image semantic segmentation sub-model, the marine oil spill target recognition sub-model and the marine oil spill area prediction sub-model to obtain the marine oil spill area prediction model;

[0078] According to the petroleum ecotoxicity test data, several petroleum ecotoxicity measurement indicators are obtained, and according to the several petroleum ecotoxicity measurement indicators, several historical marine water quality monitoring data are screened to construct several corresponding historical petroleum ecotoxicity impact data, including the following steps:

[0079] Based on the principal component analysis (PCA), the petroleum ecotoxicity test data is converted into the petroleum ecotoxicity test data matrix X = [x1, x2, ... x i ,...,x n ] T , where x i is the row vector of the i-th petroleum ecotoxicity test data, i is the data indicator, and n is the total number of data; the historical user behavior data matrix is ​​composed of several petroleum ecotoxicity test data row vectors {x1, x2, ...x i ,...,x n}constitute;

[0080] The conversion matrix of the petroleum ecotoxicity test data matrix is ​​obtained as follows:

[0081]

[0082] Where D is the covariance matrix of the principal component matrix Y; Y is the principal component matrix; P is the transformation matrix; E is the unit eigenvector matrix;

[0083] According to the petroleum ecotoxicity test data matrix X and the transformation matrix P, the principal component matrix H = [h1,h2,...h l ,...,h L ], the principal component matrix includes several candidate principal components, among which h l is the lth candidate principal component column vector, l is the data dimension indicator, L is the total number of data dimensions, and the formula is:

[0084] H=PX

[0085] Where H is the principal component matrix; P is the conversion matrix; X is the petroleum ecotoxicity test data matrix;

[0086] According to the cumulative contribution rate of the variance of all candidate principal components, if it exceeds 85%, the corresponding K≤L candidate principal components are used as principal components to obtain several principal component indicators related to petroleum ecotoxicity, and several principal component indicators are used as several petroleum ecotoxicity measurement indicators. The formula is:

[0087]

[0088] In the formula, G is the cumulative contribution rate of variance; lThe lth candidate principal component h l The variance of ; l is the indicator of data dimension; L is the total number of data dimensions; K is the total number of principal components;

[0089] According to several petroleum ecotoxicity measurement indicators, several historical marine water quality monitoring data are screened to construct several corresponding historical petroleum ecotoxicity impact data;

[0090] Integrate historical marine oil spill area prediction data, historical ocean current data, and historical petroleum ecotoxicity impact data in the same sea area to construct a marine oil spill ecotoxicity analysis training sample set;

[0091] By calculating the cumulative contribution rate of variance of data of different dimensions in the petroleum ecotoxicity test data, the potential relationship between data of different dimensions and the predicted value of user characteristics is reflected, and data dimension reduction is performed based on this relationship. Under the premise of retaining key data features, the high-latitude petroleum ecotoxicity test data is mapped to low-latitude, which reduces the data volume and improves the data representation ability.

[0092] Based on the marine oil spill ecotoxicity analysis training sample set, a marine oil spill ecotoxicity analysis model is constructed using a deep learning algorithm, including the following steps:

[0093] The marine oil spill ecotoxicity analysis training sample set is divided into a model training sample set and a model testing sample set in a ratio of 7:3;

[0094] According to the model training sample set, the Deep Belief Network (DBN)-Improved Fireworks Optimization Algorithm (IFWA) algorithm is used for optimization training to build an initial marine oil spill ecotoxicity analysis model;

[0095] The marine oil spill ecotoxicity analysis model includes a marine oil spill ecotoxicity analysis module based on the DBN algorithm and a model optimization module based on the IFWA optimization algorithm, which are connected in sequence;

[0096] The steps include:

[0097] The initial network parameters of the DBN network are used as the optimization target of the IFWA optimization algorithm, and the algorithm parameters of the IFWA optimization algorithm are initialized; the algorithm parameters include the IFWA population parameters and the maximum number of iterations;

[0098] The optimization target is the position of the IFWA individual, and the fitness function of the IFWA individual is set as follows:

[0099]

[0100] In the formula, f(q l' ) is the initial IFWA individual q l' The fitness value; MSE is the prediction mean square error function; Y' l' , Y l' are the predicted value and the true value; L is the total number of samples;

[0101] According to the IFWA population parameters, the Tent-Logistic-Cosine chaotic mapping sequence is used to initialize the IFWA population, and the initialized IFWA population is obtained. The formula is:

[0102]

[0103] In the formula, q' i" The initial position of the IFWA individuals generated by the Tent-Logistic-Cosine chaotic mapping sequence; q' i is the initial position of the randomly generated IFWA individual; r is a preset parameter of 0-1; i" is the IFWA individual indicator;

[0104] According to the fitness function, the fitness value of each IFWA individual in the initialized IFWA population is obtained;

[0105] Get the explosion radius and number of sparks of each IFWA individual in the initialized IFWA population. The formula is:

[0106]

[0107] In the formula, S l' is the initial IFWA individual q l' The number of sparks; M' is a constant; f max is the maximum fitness value in the initialized IFWA population; f(q l' ) is the initial IFWA individual q l' The fitness value of ; τ is an infinitesimal constant;

[0108]

[0109] In the formula, R l' is the initial IFWA individual q l' Explosion radius; f is the explosion radius adjustment constant; min is the minimum fitness value in the initialized IFWA population;

[0110] According to the explosion radius and number of sparks of each initial IFWA individual in the initialized IFWA population, fireworks are exploded to obtain an updated IFWA population. The formula is:

[0111] q' l' =q l' +S l' ×rand(-1,1)

[0112] In the formula, q' l' is the updated IFWA individual; rand(-1,1) is a random number between -1 and 1;

[0113] Use the Gaussian mutation algorithm to perform Gaussian mutation on the initialized IFWA population to generate a Gaussian mutated IFWA population. The formula is:

[0114] q" l' =q l’ +S l' ×G(1,1)

[0115] In the formula, q" l' is an IFWA individual with Gaussian variation; G(1,1) is a random number with Gaussian distribution whose mean and variance are both 1;

[0116] Use the dynamic reverse learning algorithm to perform dynamic reverse learning on the initialized IFWA population to obtain the reverse IFWA population. The formula is:

[0117] q"' l' =γ(L max +L min )-q l'

[0118] In the formula, q"' l' is the reverse IFWA individual; γ is the decreasing inertia coefficient; L max , L min are the maximum and minimum values ​​of the vector space respectively;

[0119] Obtain the fitness values ​​of the IFWA individuals of the updated IFWA population, the Gaussian-mutated IFWA population, and the reverse IFWA population, and take the IFWA individual with the minimum fitness value as the optimal IFWA individual;

[0120] If the number of iterations reaches the threshold or the fitness value of the optimal IFWA individual meets the requirements, the optimal solution corresponding to the current optimal IFWA individual is output to obtain the optimal initial network parameters of the DBN network;

[0121] According to the optimal initial network parameters, the DBN network is set up, and the model training sample set is used for optimization training to build the initial marine oil spill ecotoxicity analysis model;

[0122] Input the model test sample set into the initial marine oil spill ecotoxicity analysis model, conduct marine oil spill ecotoxicity analysis, and obtain the model prediction accuracy;

[0123] If the model prediction accuracy is greater than the preset model prediction accuracy threshold, the optimal marine oil spill ecotoxicity analysis model is output, otherwise, the optimization training continues;

[0124] Collect real-time remote sensing image data of marine oil spill accidents, real-time ocean current data and real-time marine water quality monitoring data of the current sea area, and input the real-time remote sensing image data of marine oil spill accidents into the marine oil spill area prediction model to predict the marine oil spill area and obtain the corresponding real-time marine oil spill area prediction data;

[0125] According to several petroleum ecotoxicity measurement indicators, data screening is performed on several petroleum ecotoxicity measurement indicators to obtain real-time petroleum ecotoxicity impact data;

[0126] The real-time marine oil spill area prediction data, real-time ocean current data and real-time petroleum ecotoxicity impact data of the current sea area are input into the marine oil spill ecotoxicity analysis model to conduct marine oil spill ecotoxicity analysis and obtain the corresponding marine oil spill ecotoxicity analysis results.

[0127] Embodiment 2:

[0128] like Figure 2 As shown, this embodiment provides a marine oil spill ecotoxicity analysis system based on image recognition, which is used to implement a marine oil spill ecotoxicity analysis method. The system includes a prediction model construction unit, an indicator acquisition unit, a data integration unit, an analysis model construction unit, an oil spill area prediction unit, a data screening unit, and an ecotoxicity analysis unit connected in sequence;

[0129] The prediction model building unit is used to collect the historical remote sensing image data of marine oil spill accidents in several sea areas, and the historical ocean current data and historical marine water quality monitoring data of the corresponding sea areas, and to build a marine oil spill area prediction model based on the remote sensing image data of several historical marine oil spill accidents by using an image recognition algorithm to obtain the corresponding several historical marine oil spill area prediction data;

[0130] An index acquisition unit is used to obtain a number of petroleum ecotoxicity measurement indicators based on petroleum ecotoxicity test data, and to screen a number of historical marine water quality monitoring data based on the number of petroleum ecotoxicity measurement indicators to construct a number of corresponding historical petroleum ecotoxicity impact data;

[0131] The data integration unit is used to integrate the historical marine oil spill area prediction data, historical ocean current data and historical petroleum ecotoxicity impact data of the same sea area to construct a training sample set for marine oil spill ecotoxicity analysis;

[0132] An analysis model building unit, used to build a marine oil spill ecotoxicity analysis model using a deep learning algorithm based on a marine oil spill ecotoxicity analysis training sample set;

[0133] The oil spill area prediction unit is used to collect real-time remote sensing image data of marine oil spill accidents, real-time ocean current data and real-time marine water quality monitoring data of the current sea area, and input the real-time remote sensing image data of marine oil spill accidents into the marine oil spill area prediction model to perform marine oil spill area prediction and obtain corresponding real-time marine oil spill area prediction data;

[0134] A data screening unit, for screening data of a number of petroleum ecotoxicity measurement indicators according to a number of petroleum ecotoxicity measurement indicators, and obtaining real-time petroleum ecotoxicity impact data;

[0135] The ecotoxicity analysis unit is used to input the real-time marine oil spill area prediction data, real-time ocean current data and real-time petroleum ecotoxicity impact data of the current sea area into the marine oil spill ecotoxicity analysis model, conduct marine oil spill ecotoxicity analysis, and obtain the corresponding marine oil spill ecotoxicity analysis results.

[0136] The present invention provides a marine oil spill ecotoxicity analysis method and system based on image recognition. The method and system are combined with remote sensing image data of marine oil spill accidents, use image recognition algorithms, build a marine oil spill area prediction model, obtain the region and area of ​​marine oil spill accidents, and improve real-time and practicality by real-time monitoring of specific on-site conditions of marine oil spill accidents. The scope of marine oil spills can be grasped at the first time, providing data support for subsequent ecotoxicity analysis. According to petroleum ecotoxicity test data, a number of petroleum ecotoxicity measurement indicators are obtained, and qualitative and quantitative analysis of petroleum ecotoxicity is performed, thereby improving the accuracy of marine oil spill ecotoxicity analysis. A marine oil spill ecotoxicity analysis model is built using a deep learning algorithm, and automatic analysis is performed according to real-time marine oil spill area prediction data, real-time ocean current data, and real-time petroleum ecotoxicity impact data of the current sea area, thereby avoiding reliance on laboratory testing and on-site investigations, reducing material, human and time cost investment, and improving the value of marine oil spill ecotoxicity analysis in actual accidents.

[0137] The present invention is not limited to the above optional implementations, and anyone can derive other various forms of products under the enlightenment of the present invention. The above specific implementations should not be understood as limiting the scope of protection of the present invention. The scope of protection of the present invention should be based on the definition in the claims, and the description can be used to interpret the claims.

Claims

1. A method for analyzing the ecotoxicity of marine oil spills based on image recognition, characterized in that: The steps include: Collecting the historical remote sensing image data of marine oil spill accidents in several sea areas, as well as the historical ocean current data and historical marine water quality monitoring data of the corresponding sea areas, and constructing a marine oil spill area prediction model based on the remote sensing image data of several historical marine oil spill accidents using image recognition algorithms to obtain the corresponding historical marine oil spill area prediction data; According to the petroleum ecotoxicity test data, a number of petroleum ecotoxicity measurement indicators are obtained, and according to the petroleum ecotoxicity measurement indicators, a number of historical marine water quality monitoring data are screened to construct a number of corresponding historical petroleum ecotoxicity impact data; Integrate historical marine oil spill area prediction data, historical ocean current data, and historical petroleum ecotoxicity impact data in the same sea area to construct a marine oil spill ecotoxicity analysis training sample set; Based on the marine oil spill ecotoxicity analysis training sample set, a marine oil spill ecotoxicity analysis model is constructed using a deep learning algorithm, including the following steps: The marine oil spill ecotoxicity analysis training sample set is divided into a model training sample set and a model testing sample set in a ratio of 7:3; Based on the model training sample set, the DBN-IFWA algorithm is used to perform optimization training and construct the initial marine oil spill ecotoxicity analysis model; The marine oil spill ecotoxicity analysis model comprises a marine oil spill ecotoxicity analysis module constructed based on a DBN algorithm and a model optimization module constructed based on an IFWA optimization algorithm, which are connected in sequence; According to the model training sample set, the DBN-IFWA algorithm is used to perform optimization training to build an initial marine oil spill ecotoxicity analysis model, including the following steps: The initial network parameters of the DBN network are used as the optimization target of the IFWA optimization algorithm, and the algorithm parameters of the IFWA optimization algorithm are initialized; the algorithm parameters include the IFWA population parameters and the maximum number of iterations; The optimization target is the position of the IFWA individual, and the fitness function of the IFWA individual is set; According to the IFWA population parameters, the Tent-Logistic-Cosine chaotic mapping sequence is used to initialize the IFWA population to obtain the initialized IFWA population; According to the fitness function, the fitness value of each IFWA individual in the initialized IFWA population is obtained; Get the explosion radius and number of sparks of each IFWA individual in the initialized IFWA population; According to the explosion radius and the number of sparks of each initial IFWA individual in the initialized IFWA population, fireworks are exploded to obtain an updated IFWA population; Use the Gaussian mutation algorithm to perform Gaussian mutation on the initialized IFWA population to generate a Gaussian mutated IFWA population; Use the dynamic reverse learning algorithm to perform dynamic reverse learning on the initialized IFWA population to obtain the reverse IFWA population; Obtain the fitness values ​​of the IFWA individuals of the updated IFWA population, the Gaussian-mutated IFWA population, and the reverse IFWA population, and take the IFWA individual with the minimum fitness value as the optimal IFWA individual; If the number of iterations reaches the threshold or the fitness value of the optimal IFWA individual meets the requirements, the optimal solution corresponding to the current optimal IFWA individual is output to obtain the optimal initial network parameters of the DBN network; According to the optimal initial network parameters, the DBN network is set up, and the model training sample set is used for optimization training to build the initial marine oil spill ecotoxicity analysis model; Input the model test sample set into the initial marine oil spill ecotoxicity analysis model, conduct marine oil spill ecotoxicity analysis, and obtain the model prediction accuracy; If the model prediction accuracy is greater than the preset model prediction accuracy threshold, the optimal marine oil spill ecotoxicity analysis model is output, otherwise, the optimization training continues; Collect real-time remote sensing image data of marine oil spill accidents, real-time ocean current data and real-time marine water quality monitoring data of the current sea area, and input the real-time remote sensing image data of marine oil spill accidents into the marine oil spill area prediction model to predict the marine oil spill area and obtain the corresponding real-time marine oil spill area prediction data; According to several petroleum ecotoxicity measurement indicators, data screening is performed on several petroleum ecotoxicity measurement indicators to obtain real-time petroleum ecotoxicity impact data; The real-time marine oil spill area prediction data, real-time ocean current data and real-time petroleum ecotoxicity impact data of the current sea area are input into the marine oil spill ecotoxicity analysis model to conduct marine oil spill ecotoxicity analysis and obtain the corresponding marine oil spill ecotoxicity analysis results.

2. A method for analyzing marine oil spill ecotoxicity based on image recognition according to claim 1, characterized in that: Collecting the historical remote sensing image data of marine oil spill accidents in several sea areas, and the historical ocean current data and historical marine water quality monitoring data of the corresponding sea areas, and constructing a marine oil spill area prediction model based on the remote sensing image data of several historical marine oil spill accidents, using image recognition algorithm, and obtaining the corresponding several historical marine oil spill area prediction data, including the following steps: Collect historical marine oil spill accident remote sensing image data of several sea areas, as well as historical ocean current data and historical marine water quality monitoring data of the corresponding sea areas; Preprocessing several historical marine oil spill accident remote sensing image data to obtain a preprocessed historical marine oil spill accident remote sensing image dataset; Based on the pre-processed historical marine oil spill accident remote sensing image dataset, an image semantic segmentation sub-model is constructed, and the corresponding historical marine area segmentation image data are obtained; According to several historical ocean area segmentation image data, a marine oil spill target recognition sub-model is constructed, and several corresponding historical marine oil spill target area image data are obtained; Based on several historical marine oil spill target area image data, a marine oil spill area prediction sub-model is constructed, and several corresponding historical marine oil spill area prediction data are obtained; The image semantic segmentation sub-model, the marine oil spill target recognition sub-model and the marine oil spill area prediction sub-model are integrated to obtain the marine oil spill area prediction model.

3. A method for analyzing marine oil spill ecotoxicity based on image recognition according to claim 2, characterized in that: The preprocessing includes cropping, translation, Gaussian denoising, grayscale processing, normalization processing and label adding performed in sequence.

4. A method for analyzing marine oil spill ecotoxicity based on image recognition according to claim 2, characterized in that: The image semantic segmentation sub-model is constructed based on the GSCNN algorithm, and the image semantic segmentation sub-model includes a first input layer, a rule flow module, a shape flow module, a fusion module and a first output layer, the first input layer is connected to the rule flow module and the shape flow module respectively, the rule flow module and the shape flow module are both connected to the fusion module, and the fusion module is connected to the first output layer; The marine oil spill target recognition sub-model is constructed based on the YOLOv8 algorithm, and the marine oil spill target recognition sub-model includes a second input layer, a Backbone module, a Neck module, a Head module, a Loss module and a second output layer connected in sequence; The marine oil spill area prediction sub-model is based on the LFMCNN algorithm, and the marine oil spill area prediction sub-model includes a third input layer, a first convolution module, a downsampling module, a second convolution module, a third convolution module, a fully connected layer and a third output layer connected in sequence.

5. The method for analyzing marine oil spill ecotoxicity based on image recognition according to claim 1, characterized in that: According to the petroleum ecotoxicity test data, several petroleum ecotoxicity measurement indicators are obtained, and according to the several petroleum ecotoxicity measurement indicators, several historical marine water quality monitoring data are screened to construct several corresponding historical petroleum ecotoxicity impact data, including the following steps: Based on the PCA method, the petroleum ecotoxicity test data are converted into a petroleum ecotoxicity test data matrix; Obtain the transformation matrix of the petroleum ecotoxicity test data matrix; According to the petroleum ecotoxicity test data matrix and the corresponding conversion matrix, a principal component matrix is ​​obtained; the principal component matrix includes a plurality of candidate principal components; According to the cumulative variance contribution rate of all candidate principal components, if it exceeds 85%, the corresponding candidate principal components are used as principal components to obtain several principal component indices related to petroleum ecotoxicity, and the several principal component indices are used as several petroleum ecotoxicity measurement indices; According to several petroleum ecotoxicity measurement indicators, several historical marine water quality monitoring data were screened and several corresponding historical petroleum ecotoxicity impact data were constructed.

6. A marine oil spill ecotoxicity analysis system based on image recognition, used to implement the marine oil spill ecotoxicity analysis method as described in any one of claims 1 to 5, characterized in that: The system comprises a prediction model building unit, an indicator acquisition unit, a data integration unit, an analysis model building unit, an oil spill area prediction unit, a data screening unit and an ecotoxicity analysis unit which are connected in sequence; The prediction model building unit is used to collect the historical remote sensing image data of marine oil spill accidents in several sea areas, and the historical ocean current data and historical marine water quality monitoring data of the corresponding sea areas, and to build a marine oil spill area prediction model based on the remote sensing image data of several historical marine oil spill accidents by using an image recognition algorithm to obtain the corresponding several historical marine oil spill area prediction data; An index acquisition unit is used to obtain a number of petroleum ecotoxicity measurement indicators based on petroleum ecotoxicity test data, and to screen a number of historical marine water quality monitoring data based on the number of petroleum ecotoxicity measurement indicators to construct a number of corresponding historical petroleum ecotoxicity impact data; The data integration unit is used to integrate the historical marine oil spill area prediction data, historical ocean current data and historical petroleum ecotoxicity impact data of the same sea area to construct a training sample set for marine oil spill ecotoxicity analysis; An analysis model building unit, used to build a marine oil spill ecotoxicity analysis model using a deep learning algorithm based on a marine oil spill ecotoxicity analysis training sample set; The oil spill area prediction unit is used to collect real-time remote sensing image data of marine oil spill accidents, real-time ocean current data and real-time marine water quality monitoring data of the current sea area, and input the real-time remote sensing image data of marine oil spill accidents into the marine oil spill area prediction model to perform marine oil spill area prediction and obtain corresponding real-time marine oil spill area prediction data; A data screening unit, for screening data of a number of petroleum ecotoxicity measurement indicators according to a number of petroleum ecotoxicity measurement indicators, to obtain real-time petroleum ecotoxicity impact data; The ecotoxicity analysis unit is used to input the real-time marine oil spill area prediction data, real-time ocean current data and real-time petroleum ecotoxicity impact data of the current sea area into the marine oil spill ecotoxicity analysis model, conduct marine oil spill ecotoxicity analysis, and obtain the corresponding marine oil spill ecotoxicity analysis results.

Images