Ocean benthos detection method based on depth discriminant representation learning under long-tail distribution

Abstract

The invention relates to a deep discriminant characterization learning-based marine benthos detection method under long tail distribution, which comprises the following steps of: constructing a data set based on marine benthos pictures with any one or more of sea urchins, scallops, starfishes and sea cucumbers; dividing the data set into a training set and a test set; constructing an ocean benthos detection model based on depth discriminant representation learning under long-tail distribution; and based on the training set data, training the deep discriminant representation learning-based marine benthos detection model under the long-tail distribution to obtain the trained deep discriminant representation learning-based marine benthos detection model under the long-tail distribution. And inputting test set data into a trained ocean benthos detection model based on deep discriminant representation learning under long tail distribution to realize detection and identification of sea urchins, scallops, starfishes and sea cucumbers in pictures. The diversity of the tail embedding characteristics of the marine benthos is improved, so that the detection precision of the marine benthos under long tail distribution is enhanced.

Description

Technical Field

[0001] The present invention belongs to the field of fully automated products and relates to a method for detecting marine benthic organisms based on deep discriminant representation learning under long-tail distribution. Background Art

[0002] Affected by the economic benefits of marine benthic organisms and the way in which samples are collected, the existing marine benthic organism datasets [1]-[3] It shows different degrees of long-tail distribution phenomenon, where the category with a large number is called the head category, and the category with a small number is called the tail category. The detection network based on deep learning can often fully exploit the features of the head category, but it is difficult to effectively learn the features of the tail category, so that the detector can detect the head category with high precision, but it is difficult to accurately detect the tail category. [4]-[6] .

[0003] Methods to improve the accuracy of marine benthic organism detection under long-tail distribution mainly include resampling [7] , data synthesis[6], reweighting [8] and metric learning [9]-

[11] Etc. Resampling techniques include oversampling and undersampling. Oversampling refers to repeated sampling of the few-sample category to increase the number of samples; undersampling refers to random sampling of the many-sample category to reduce the number of samples. Data synthesis refers to generating images similar to the few-sample category, and reweighting refers to assigning corresponding weights according to the number of samples of each category. Due to their high efficiency and flexibility, data synthesis technology and reweighting strategy have been more widely used than resampling technology.

[0004] In order to achieve high-precision detection of marine benthic organisms under long-tail distribution, the gradient field, divergence and coefficient matrix are calculated in sequence.

[12] Poisson fusion technology is used to expand the categories with fewer samples in the marine benthic biological dataset, thereby improving the detection performance of the few-sample categories. In order to increase the number of samples of sea cucumbers and sea urchins, the literature

[13] A generative adversarial network based on Poisson fusion technology is proposed. In the Poisson fusion stage, the targets cut from the marine benthic dataset are fused with any seabed background to increase the tail class samples. In the learning stage, U-Net is used to generate more realistic underwater images, which effectively improves the detection accuracy of marine benthic organisms under long-tail distribution. By comprehensively considering the color, texture and contrast characteristics of the real underwater environment, the literature

[14] We carefully selected images containing only the few-sample classes, constructed an unpaired training dataset, and used an unsupervised cyclic generative adversarial network to keep the number of samples in each category basically consistent, thereby improving the detection accuracy of the few-sample classes.

[0005] Reweighting is another important way to improve the high-precision detection of marine benthic organisms under long-tail distribution

[15] .literature [8] The classification weights are designed based on the proportion of each category, and the distance weights are constructed by considering the central Euclidean distance between the predicted box and the true box, thereby giving higher weights to categories with a small number of samples and a close center distance. The results show that the combined consideration of classification weights and distance weights can improve the detection accuracy by 2.4%.

[0006] In recent years, a large number of experts and scholars have proposed metric learning strategies to solve the long-tail distribution problem by constructing discriminative embedding features. By minimizing the Euclidean distance between the intra-class embedding features and the corresponding centers, the literature [9] The center loss function is constructed. By introducing the multiplicative angle margin into the Softmax function, the reference

[10] designed the A-Softmax loss to quantitatively control the decision boundary. By applying the normalization technique to the embedded features and weight vectors, the reference

[11] The additive cosine margin was introduced and the Large Margin Cosine Loss (LMCL) technique was developed to help build discriminative features.

[0007] Through the review and analysis of existing work, the following deficiencies still exist in the fuzzy marine benthic organism detection:

[0008] (1) Regarding sample synthesis methods: Poisson fusion technology ignores the overall consistency of the fused image in terms of contrast, tone, and texture, and is overly sensitive to pixels at the boundary between the foreground target and the background image; using generative adversarial networks to achieve data synthesis has problems with training instability and mode collapse; sample synthesis methods such as Cutout and Mixup lack target perception.

[0009] (2) About resampling and reweighting methods: In resampling technology, undersampling of the head class is likely to cause feature information loss, while oversampling of the tail class is likely to cause overfitting problems. The reweighting strategy loses the accuracy of the head class to a certain extent.

[0010] (3) Regarding metric learning: The existing center loss only considers the difference between the center of the embedded feature and samples of the same type, ignoring the inter-class dispersion of the embedded features, that is, it is difficult to simultaneously improve the compactness of intra-class samples and the dispersion of inter-class samples. The LMCL method cannot maintain a consistent linear decision boundary, which reduces the discriminability of the embedded features. Summary of the invention

[0011] In order to solve the above problems, the technical solution adopted by the present invention is: a method for detecting marine benthic organisms based on deep discriminant representation learning under long-tail distribution, comprising the following steps:

[0012] Construct a dataset based on images of any one or more of the following marine benthic organisms: sea urchins, scallops, starfish, and sea cucumbers;

[0013] Divide the dataset into training and testing sets;

[0014] Construct a marine benthic organism detection model based on deep discriminant representation learning under long-tail distribution;

[0015] Based on the training set data, the marine benthic organism detection model based on deep discriminant representation learning under long-tail distribution is trained to obtain the trained marine benthic organism detection model based on deep discriminant representation learning under long-tail distribution;

[0016] The test set data is input into the marine benthic organism detection model based on deep discriminant representation learning under the trained long-tail distribution to achieve the detection and identification of marine benthic organisms such as sea urchins, scallops, starfish and sea cucumbers.

[0017] Furthermore, it is characterized in that: the marine benthic organism detection model based on deep discriminant representation learning under the long-tail distribution includes:

[0018] Backbone network: extracts the informative feature graph contained in the image;

[0019] Classification loss module based on discriminative additional angle margin: Based on the informative feature map extracted by the backbone network, combined with the regional alignment mechanism, the bilinear interpolation method is used to construct the characteristic values ​​of the sampling points of the marine benthic organisms in the area of ​​interest, and the average aggregation strategy is used to obtain the single-interval marine benthic organisms. The embedded features of the marine benthic organisms are constructed with the help of dimensional conversion operations, and the angle between the embedded features and the corresponding category weights is calculated. By increasing the angle margin in the cosine space, the discriminability of the embedded features of the marine benthic organisms is improved;

[0020] Uneven embedding feature distribution loss module: A sliding average strategy is used to update the embedding feature centers of various types of marine benthic organisms, and the angles between the embedded features and their corresponding centers are solved in turn to calculate the differences in the embedding feature distribution between the tail class and other head classes of marine benthic organisms.

[0021] Furthermore, the classification loss module based on the discriminative additional angle margin includes:

[0022] Informative feature map construction submodule: used to construct the informative feature map extracted by the backbone network;

[0023] A submodule for estimating characteristic values ​​of a region of interest: used for estimating characteristic values ​​of a region of interest based on the informative characteristic map transmitted by the informative characteristic map construction submodule;

[0024] Discriminative additional angle margin classification loss submodule: a loss function is designed based on the eigenvalues ​​of the region of interest estimated by the eigenvalue estimation submodule of the region of interest.

[0025] Furthermore, the process of constructing the informative feature map extracted by the backbone network is as follows:

[0026] The informative feature map extracted by the backbone network is expressed as:

[0027]

[0028] Among them, E(·; ·) is the feature mapping function in the detection network, Represents the height h r , width is w r , batch size is n b and the number of channels is n rc The original input sample, θ e represents the parameter to be optimized, represents the feature map, h and w are the height and width of the feature map, n c is the number of embedding feature layers.

[0029] Using the batch normalization method, it is expressed as:

[0030]

[0031] Where b = 1, 2, L, n b and l=1,2,L,n c represents the sample index and feature layer index, γ and β represent the scale and offset coefficients respectively, ∈ is a constant, The coordinates of the lth feature layer in the bth sample are (x m ,y n ), μ l and σ l are the mean and standard deviation of the features indexed at the lth layer, expressed as:

[0032]

[0033] in, and Represents n c The mean and standard deviation of each channel;

[0034] Taking the b-th sample as an example, the construction process of discriminative features is explained. In order to increase the nonlinear mapping capability, Leaky ReLU is used as the activation function, which is expressed as:

[0035]

[0036] Among them, 0<κ<1 represents the slope of the activation function, H l (x m ,y n ) represents the coordinates in the lth feature layer (x m ,y n ) is the activation eigenvalue at .

[0037] Furthermore, the process of estimating the feature value of the region of interest based on the informative feature map transmitted by the informative feature map construction submodule is as follows:

[0038] The coordinates of the feature points in the region of interest are expressed as:

[0039]

[0040] in, is the coordinate of the upper left corner of the region of interest, w bin and h bin Represents the width and height of a single interval, g = 1, L, n p and k=1,L,n q are the column and row dimension indices of the interval, u=1,L,n col and v=1,L,n row Respectively represent the index of the sampling point in a single interval in the column dimension and row dimension, (x g,u ,y k,v ) are the coordinates of the sampling points;

[0041] By means of the bilinear interpolation method, for the i-th instance in the l-th feature layer, the coordinate position is (x g,u ,y n ) at the eigenvalue It is expressed as:

[0042]

[0043] Among them, (x m ,y n ) and (x m+1 ,y n ) is the closest sampling point (x g,u ,y k,v )'s upper left and upper right integer coordinates;

[0044] Similarly, the coordinates (x g,u ,y n+1 ) at the eigenvalue It is expressed as:

[0045]

[0046] Among them, (x m ,yn+1 ) and (x m+1 ,y n+1 ) is the closest to (x g,u ,y k,v )’s lower left and lower right corner integer coordinate positions;

[0047] Furthermore, the coordinates are (x g,u ,y k,v ) at the eigenvalue It is expressed as:

[0048]

[0049] After training m batches, for the i-th instance in the l-th feature layer, the feature value at interval index (g, k) It is expressed as:

[0050]

[0051] in, Represents the embedded feature of the i-th marine benthic organism instance at the l-th layer. Considering all feature layers, the embedded feature f of the i-th marine benthic organism instance i m It is expressed as:

[0052]

[0053] in, represents the embedded features obtained by the dimension transformation operation T(·), d = n q ×n p ×n c , Represents n c The embedded features of the channels.

[0054] Furthermore, the process of designing the loss function based on the eigenvalue of the region of interest estimated by the eigenvalue estimation submodule of the region of interest is as follows:

[0055] Normalized embedding features and weights are expressed as:

[0056]

[0057]

[0058] in, yes The jth column of non-normalized weights, j = 1, L, n tc Represents the category index, n tc represents the total number of categories, ‖·‖2 represents the 2-norm, and Represent the normalized embedding features and normalized weights respectively;

[0059] The i-th embedding feature and the weight of the jth class The angle between them is expressed as:

[0060]

[0061] in,

[0062] Discriminative additional angle margin classification loss function l DAAMCL It is expressed as:

[0063]

[0064] Where s is the scale parameter, δ represents the additional angle margin, Represents embedded features With corresponding category weight The angle between inst Represents the number of sample instances in this batch.

[0065] Furthermore, the distribution loss module of the uneven embedding feature includes:

[0066] The embedding feature center construction submodule is used to transfer the distribution information of the embedded features of marine benthic organisms from the implicit head class to the tail class, propose an uneven embedded feature distribution loss, enhance the diversity of the embedded features of the tail class of marine benthic organisms, and construct the embedded features of each category respectively;

[0067] The intra-class embedding feature distribution variance calculation submodule is used to calculate the embedding feature distribution variance of each category of embedding features constructed by the embedding feature center construction submodule;

[0068] Uneven embedding feature distribution loss submodule: used for uneven embedding feature distribution loss function of all categories based on the embedding feature distribution variance calculated by the intra-class embedding feature distribution variance calculation submodule.

[0069] Furthermore, by migrating the distribution information of the embedded features of the marine benthic organisms from the implicit head class to the tail class, an uneven embedded feature distribution loss is proposed to enhance the diversity of the embedded features of the tail class of the marine benthic organisms. The process of constructing the embedded features of each category is as follows:

[0070] Construct the embedding features of each category and calculate the corresponding embedding feature center, expressed as:

[0071]

[0072] Among them, Ω∈{ec,sc,st,ho}, ec, sc, st and ho represent the categories of sea urchin, scallop, starfish and sea cucumber respectively. represents the i-th embedded feature in the corresponding category after training m batches, represents the center of the embedded feature, Represents the number of samples of the corresponding category after training m batches;

[0073] Adopt the sliding average mechanism to make full use of the embedding feature center of the m-1th batch To update the embedding feature center of the mth batch It is expressed as:

[0074]

[0075] in, λ represents the class center update rate.

[0076] Further: the process of constructing each category of embedded features for the embedding feature center construction submodule to calculate the embedding feature distribution variance is as follows:

[0077] The angle between the embedded feature and its corresponding category center is expressed as:

[0078]

[0079] in, is the i-th embedding feature after training m batches Its corresponding center Angle.

[0080] Average angle distribution of different categories It is expressed as:

[0081]

[0082] Then, the intra-class embedding feature distribution variance is expressed as:

[0083]

[0084] in, and Represents the variance of the intra-class embedding feature distribution of sea urchins, scallops, starfish, and sea cucumbers after training for m batches.

[0085] Further: based on the embedding feature distribution variance calculated by the intra-class embedding feature distribution variance calculation submodule, the process of the uneven embedding feature distribution loss function of all categories is as follows:

[0086] The uneven embedding feature distribution loss is expressed as:

[0087]

[0088] in, and Representatives from the collection The first largest intra-class embedding feature distribution variance, the second largest intra-class embedding feature distribution variance, the third largest intra-class embedding feature distribution variance, and the fourth largest intra-class embedding feature distribution variance, n 1st 、n 2nd 、n 3rd and n 4th is the corresponding number of instances, and Represents the uneven embedding feature distribution loss for different categories.

[0089] Finally, the uneven embedding feature distribution loss l for all categories UEFDL It is expressed as:

[0090]

[0091] The present invention provides a method for detecting benthic organisms based on deep discriminant representation learning under long-tail distribution, which has the following advantages:

[0092] 1. The regional alignment technology can be used to establish the spatial correspondence between the region of interest of marine benthic organisms and the embedded features, avoiding the mapping offset problem caused by quantization operations and ensuring the integrity of the embedded features of benthic organisms;

[0093] 2. By calculating the angle between the embedding features of marine benthic organisms and the corresponding class weights, an additional angle margin is added in the cosine space, which improves the discriminability of the embedding features of marine benthic organisms and enhances the intra-class compactness and inter-class differences;

[0094] 3. Without manually specifying the head class and tail class, the embedded feature distribution information of marine benthic organisms can be adaptively transferred from the implicit head class to the tail class, which improves the diversity of the embedded features of the tail class of marine benthic organisms, thereby enhancing the detection accuracy of marine benthic organisms under long-tail distribution. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0096] Figure 1 A flowchart of the application method;

[0097] Figure 2 It is a sampling point eigenvalue calculation mechanism with a bilinear interpolation strategy;

[0098] Figure 3 An additional angle margin classification loss is added for discriminability;

[0099] Figure 4 Loss for uneven embedding feature distribution;

[0100] Figure 5 Visualization of the embedding feature distribution of different categories; (a) is the feature distribution of the 1st iteration, (b) is the feature distribution of the 20th iteration, (c) is the feature distribution of the 40th iteration, (d) is the feature distribution of the 60th iteration, (e) is the feature distribution of the 80th iteration, and (f) is the feature distribution of the 100th iteration;

[0101] Figure 6 The angle distribution of different categories: (a) the angle distribution of sea urchins, (b) the angle distribution of scallops, (c) the angle distribution of starfish, and (d) the angle distribution of sea cucumbers.

[0102] Figure 7 Performance comparison of the accuracy-recall curves of different categories, including (a) accuracy-recall curve of sea urchins, (b) accuracy-recall curve of scallops, (c) accuracy-recall curve of starfish, and (d) accuracy-recall curve of sea cucumbers;

[0103] Figure 8 Qualitative comparison of detection performance in typical scenarios, where (a) is the dark light situation, (b) is the bright light situation, (c) is the small scale situation, and (d) is the situation where multiple categories of targets coexist. DETAILED DESCRIPTION

[0104] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0105] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0106] Figure 1 A flowchart of the application method;

[0107] A method for detecting marine benthic organisms based on deep discriminant representation learning under long-tail distribution includes the following steps:

[0108] S1: Construct a dataset based on images of any one or more marine benthic organisms including sea urchins, scallops, starfish, and sea cucumbers;

[0109] S2: Divide the data set into training set and test set;

[0110] S3: Construct a marine benthic organism detection model based on deep discriminant representation learning under long-tail distribution;

[0111] S4: Based on the training set data, the marine benthic organism detection model based on deep discriminant representation learning under the long-tail distribution is trained to obtain the trained marine benthic organism detection model based on deep discriminant representation learning under the long-tail distribution;

[0112] S5: Input the test set data into the marine benthic organism detection model based on deep discriminant representation learning under the trained long-tail distribution to realize the detection and identification of marine benthic organisms such as sea urchins, scallops, starfish and sea cucumbers.

[0113] The steps S1 / S2 / S3 / S4 / S5 are performed sequentially;

[0114] Further: The marine benthic organism detection model based on deep discriminant representation learning under the long-tail distribution includes:

[0115] Backbone network: extracts the informative feature graph contained in the image;

[0116] Classification loss module based on discriminative additional angle margin: Based on the informative feature map extracted by the backbone network, combined with the regional alignment mechanism, the bilinear interpolation method is used to construct the characteristic values ​​of the sampling points of the marine benthic organisms in the area of ​​interest, and the average aggregation strategy is used to obtain the single-interval marine benthic organisms. The embedded features of the marine benthic organisms are constructed with the help of dimensional conversion operations, and the angle between the embedded features and the corresponding category weights is calculated. By increasing the angle margin in the cosine space, the discriminability of the embedded features of the marine benthic organisms is improved;

[0117] Uneven embedding feature distribution loss module: A sliding average strategy is used to update the embedding feature centers of various types of marine benthic organisms, and the angles between the embedded features and their corresponding centers are solved in turn to calculate the differences in the embedding feature distribution between the tail class and other head classes of marine benthic organisms.

[0118] The two loss modules jointly optimize the entire detection and recognition network model;

[0119] Further: the classification loss module based on the discriminative additional angle margin includes:

[0120] Informative feature map construction submodule: used to construct the informative feature map extracted by the backbone network;

[0121] A submodule for estimating characteristic values ​​of a region of interest: used for estimating characteristic values ​​of a region of interest based on the informative characteristic map transmitted by the informative characteristic map construction submodule;

[0122] Discriminative additional angle margin classification loss submodule: a loss function is designed based on the eigenvalues ​​of the region of interest estimated by the eigenvalue estimation submodule of the region of interest.

[0123] Further: the process of constructing the informative feature map extracted by the backbone network is as follows:

[0124] The informative feature map extracted by the backbone network is expressed as:

[0125]

[0126] Among them, E(·; ·) is the feature mapping function in the detection network, Represents the height h r , width is w r , batch size is n b and the number of channels is n rc The original input sample, θ e represents the parameter to be optimized, represents the feature map, h and w are the height and width of the feature map, n c is the number of embedding feature layers.

[0127] Using the batch normalization method, it is expressed as:

[0128]

[0129] Where b = 1, 2, L, n b and l=1,2,L,n c represents the sample index and feature layer index, γ and β represent the scale and offset coefficients respectively, ∈ is a constant, The coordinates of the lth feature layer in the bth sample are (x m ,y n ), μ l and σ l are the mean and standard deviation of the features indexed at the lth layer, expressed as:

[0130]

[0131] in, and Represents n cThe mean and standard deviation of each channel;

[0132] Taking the b-th sample as an example, the construction process of discriminative features is explained. In order to increase the nonlinear mapping capability, Leaky ReLU is used as the activation function, which is expressed as:

[0133]

[0134] Among them, 0<κ<1 represents the slope of the activation function, H l (x m ,y n ) represents the coordinates in the lth feature layer (x m ,y n ) is the activation eigenvalue at .

[0135] Figure 2 It is a sampling point eigenvalue calculation mechanism with a bilinear interpolation strategy;

[0136] Further: the process of estimating the feature value of the region of interest based on the informative feature map transmitted by the informative feature map construction submodule is as follows:

[0137] The coordinates of the feature points in the region of interest are expressed as:

[0138]

[0139] in, is the coordinate of the upper left corner of the region of interest, w bin and h bin Represents the width and height of a single interval, g = 1, L, n p and k=1,L,n q are the column and row dimension indices of the interval, u=1,L,n col and v=1,L,n row Respectively represent the index of the sampling point in a single interval in the column dimension and row dimension, (x g,u ,y k,v ) are the coordinates of the sampling points;

[0140] By means of the bilinear interpolation method, for the i-th instance in the l-th feature layer, the coordinate position is (x g,u ,y n ) at the eigenvalue It is expressed as:

[0141]

[0142] Among them, (x m ,y n ) and (x m+1 ,y n ) is the closest sampling point (xg,u ,y k,v )'s upper left and upper right integer coordinates;

[0143] Similarly, the coordinates (x g,u ,y n+1 ) at the eigenvalue It is expressed as:

[0144]

[0145] Among them, (x m ,y n+1 ) and (x m+1 ,y n+1 ) is the closest to (x g,u ,y k,v )’s lower left and lower right corner integer coordinate positions;

[0146] Furthermore, the coordinates are (x g,u ,y k,v ) at the eigenvalue It is expressed as:

[0147]

[0148] After training m batches, for the i-th instance in the l-th feature layer, the feature value at interval index (g, k) It is expressed as:

[0149]

[0150] in, Represents the embedded feature of the i-th marine benthic organism instance at the l-th layer. Considering all feature layers, the embedded feature f of the i-th marine benthic organism instance i m It is expressed as:

[0151]

[0152] in, represents the embedded features obtained by the dimension transformation operation T(·), d = n q ×n p ×n c , Represents n c The embedded features of the channels.

[0153] Figure 3 An additional angle margin classification loss is added for discriminability;

[0154] Furthermore, the process of designing the loss function based on the eigenvalue of the region of interest estimated by the eigenvalue estimation submodule of the region of interest is as follows:

[0155] Normalized embedding features and weights are expressed as:

[0156]

[0157] in, yes The jth column of non-normalized weights, j = 1, L, n tc Represents the category index, n tc represents the total number of categories, ‖·‖2 represents the 2-norm, and Represent the normalized embedding features and normalized weights respectively;

[0158] The i-th embedding feature and the jth class weight The angle between them is expressed as:

[0159]

[0160] in,

[0161] Discriminative additional angle margin classification loss function l DAAMCL It is expressed as:

[0162]

[0163] Where s is the scale parameter, δ represents the additional angle margin, Represents embedded features With corresponding category weight The angle between inst Represents the number of sample instances in this batch.

[0164] Further: the distribution loss module of the uneven embedding feature includes:

[0165] The embedding feature center construction submodule is used to transfer the distribution information of the embedded features of marine benthic organisms from the implicit head class to the tail class, propose an uneven embedded feature distribution loss, enhance the diversity of the embedded features of the tail class of marine benthic organisms, and construct the embedded features of each category respectively;

[0166] In-class embedded feature distribution variance calculation submodule: used to calculate the embedding feature distribution variance of each category of embedded features constructed by the embedding feature center construction submodule;

[0167] Uneven embedding feature distribution loss submodule: used for uneven embedding feature distribution loss function of all categories based on the embedding feature distribution variance calculated by the intra-class embedding feature distribution variance calculation submodule.

[0168] Furthermore, by migrating the distribution information of the embedded features of the marine benthic organisms from the implicit head class to the tail class, an uneven embedded feature distribution loss is proposed to enhance the diversity of the embedded features of the tail class of the marine benthic organisms. The process of constructing the embedded features of each category is as follows:

[0169] Construct the embedding features of each category and calculate the corresponding embedding feature center, expressed as:

[0170]

[0171] Among them, Ω∈{ec,sc,st,ho}, ec, sc, st and ho represent the categories of sea urchin, scallop, starfish and sea cucumber respectively. represents the i-th embedded feature in the corresponding category after training m batches, represents the center of the embedded feature, Represents the number of samples of the corresponding category after training m batches.

[0172] Adopt the sliding average mechanism to make full use of the embedding feature center of the m-1th batch To update the embedding feature center of the mth batch It is expressed as:

[0173]

[0174] in, λ represents the class center update rate.

[0175] Further: the process of constructing each category of embedded features for the embedding feature center construction submodule to calculate the embedding feature distribution variance is as follows:

[0176] The angle between the embedded feature and its corresponding category center is expressed as:

[0177]

[0178] in, is the i-th embedding feature after training m batches Its corresponding center Angle.

[0179] Average angle distribution of different categories It is expressed as:

[0180]

[0181] Then, the intra-class embedding feature distribution variance is expressed as:

[0182]

[0183] in, and Represents the variance of the intra-class embedding feature distribution of sea urchins, scallops, starfish, and sea cucumbers after training for m batches.

[0184] Figure 3 Loss for uneven embedding feature distribution;

[0185] Further: based on the embedding feature distribution variance calculated by the intra-class embedding feature distribution variance calculation submodule, the process of the uneven embedding feature distribution loss function of all categories is as follows:

[0186] The uneven embedding feature distribution loss is expressed as:

[0187]

[0188] in, and Representatives from the collection The first largest intra-class embedding feature distribution variance, the second largest intra-class embedding feature distribution variance, the third largest intra-class embedding feature distribution variance, and the fourth largest intra-class embedding feature distribution variance, n 1st 、n 2nd 、n 3rd and n 4th is the corresponding number of instances, and Represents the uneven embedding feature distribution loss for different categories.

[0189] Finally, the uneven embedding feature distribution loss l for all categories UEFDL It is expressed as:

[0190]

[0191] Embodiment 1:

[0192] 1 Dataset and Experimental Configuration Dataset and Experimental Configuration

[0193] To verify the effectiveness and superiority of the DDRL method, a comprehensive comparative experiment was conducted on the URPC2020 dataset and the DUO dataset, which contain 5543 images and 7782 images respectively. For the URPC2020 dataset, the training set contains 4434 images and the test set contains 1109 images; for the DUO dataset, the training set contains 6671 images and the test set contains 1111 images. The corresponding experimental configuration parameters are shown in Table 1. The constant ∈=10 is set based on experience. -5and scale parameter s = 64. AP, mAP@.5, mAP@.75, and mAP@[.5,.95] are used as detection performance evaluation indicators. A five-fold cross-validation strategy is used, and the average of the five test results is used as the detection accuracy of different models. Videos with a resolution of 1280×720 are used to obtain the inference speed.

[0194] Table 1 Experimental configuration parameters

[0195]

[0196] 2 Model parameter selection

[0197] In order to obtain the optimal number of embedding feature layers n c 、Number of intervals (n p and n q ), number of sampling points (n col and n row ), sampling strategy, angle margin δ and class center update rate λ, and carried out corresponding performance comparison experiments. In the present invention, n is set p and n q Keep consistent, col and n row Keep consistent, the comparison results are shown in Table 2-Table 6.

[0198] Table 2 Performance comparison of different embedding feature layers

[0199]

[0200] Table 3 Performance comparison of different numbers of intervals and sampling points

[0201]

[0202] Table 4 Performance comparison of different sampling strategies

[0203]

[0204] Table 5 Performance comparison of different angle margins

[0205]

[0206] Table 6 Performance comparison of different center update rates

[0207]

[0208] It can be clearly seen from Table 2 that with the increase of the number of embedding feature layers, the detection accuracy of different categories gradually improves, which indicates that more discriminative embedding features of marine benthic organisms are constructed. c =256 to n c=512, the detection accuracy improved is quite limited. In order to save computing resources, n is set in subsequent experiments. c =256.

[0209] In addition, regarding the determination of the optimal number of intervals (n p and n q ) and the number of sampling points (n col and n row ), it can be clearly seen from Table 3 that the detection accuracy is affected by the number of intervals and the number of sampling points, and it is difficult to achieve the optimal detection accuracy of the four categories at the same time under the same set of parameters. Considering the detection accuracy of all categories comprehensively, the number of intervals is finally selected as 4 and the number of sampling points is 5.

[0210] Then, regarding the determination of the optimal sampling strategy, it can be intuitively observed from Table 4 that the detection accuracy that can be achieved by the RoIPool and RoIWarp methods is significantly lower than that of the regional alignment method. The main reason is that the RoIPool and RoIWarp methods use quantization operations, which leads to a mapping offset problem between the region of interest and the constructed embedded features. On the contrary, with the help of bilinear interpolation technology, the regional alignment method avoids the quantization process, so that the spatial correspondence can be constructed pixel by pixel, and the average aggregation strategy can be used to obtain a higher detection accuracy.

[0211] Furthermore, regarding the determination of the optimal angle margin δ, it can be seen from Table 5 that the detection accuracy increases with the increase of the angle margin δ, which indicates that a more discriminative embedding feature is constructed. Although the sea urchin class achieves the highest detection accuracy when δ = 0.55, in order to take into account the performance of mAP@.5, mAP@.75 and mAP@[.5,.95] indicators, δ = 0.45 is set in subsequent experiments.

[0212] In order to determine the optimal class center update rate λ, corresponding performance comparison experiments were carried out, and the results are shown in Table 6. Overall, with the increase of the class center update rate λ, the detection accuracy of all categories gradually improves. However, when λ>0.3, the detection accuracy gradually decreases, and finally λ=0.3 is set.

[0213] 3 Embedding Feature Visualization

[0214] The t-Distributed Stochastic Neighbor Embedding (t-SNE) technique is used to present the distribution of embedded features of different categories. The corresponding results are as follows: Figure 5 As shown, Figure 5Visualization of the distribution of embedded features for different categories; (a) is the feature distribution of the 1st iteration, (b) is the feature distribution of the 20th iteration, (c) is the feature distribution of the 40th iteration, (d) is the feature distribution of the 60th iteration, (e) is the feature distribution of the 80th iteration, and (f) is the feature distribution of the 100th iteration; it is difficult to distinguish the embedded features of different categories from (a) and (b), and there is no obvious boundary between the classes. As the number of iterations increases, DDRL produces more discriminative embedded features as shown in (c) and (d). As shown in (e) and (f), with the help of the proposed DAAMCL module, DDRL can derive more obvious boundaries between different classes. In particular, for the scallop, starfish, and sea cucumber categories, the spatial range derived from the 100th iteration is larger than that of the 20th iteration, which clearly shows that with the help of the proposed UEFDL module, the tail class can gradually learn the intra-class diversity from the implicit head class.

[0215] also, Figure 6 The angle distribution of different categories: (a) the angle distribution of sea urchins, (b) the angle distribution of scallops, (c) the angle distribution of starfish, and (d) the angle distribution of sea cucumbers.

[0216] The initial and final angular distributions of the embedded features and the corresponding class centers are shown. It can be seen that the initial and final angular distributions are Gaussian. Specifically, the initial angular distribution centers of sea urchins, scallops, starfish, and sea cucumbers are approximately 40°, 50°, 43°, and 40°, respectively. In this case, the embedded features lack discriminability. Combining the DAAMCL and UEFDL modules, a more compact intra-class distribution can be achieved after multiple iterations of training, and the final angular distribution centers are approximately 20°, 25°, 23°, and 20°, which shows that DDRL derives more discriminative embedded features.

[0217] 4-module ablation experiment

[0218] In order to fully reveal the contribution of DAAMCL and UEFDL modules to the DDRL method, the present invention selects YOLOv3 as the benchmark model, and the corresponding experimental results are summarized in Table 7. It can be seen that the combination of DAAMCL module can achieve better detection performance than the benchmark model. The key reason is that the additional angle margin helps to construct discriminative embedding features. In addition, the combination of UEFDL module can improve the mAP@.5, mAP@.75 and mAP@[.5,.95] index values, indicating that the head class can transfer the variance of the embedded feature distribution to the tail class. Combining DAAMCL and UEFDL modules, the designed DDRL achieves the best detection performance.

[0219] Table 7 Ablation experiments of different modules

[0220]

[0221] Accordingly, Figure 7 The performance comparison of the accuracy-recall curves of different categories, including (a) the accuracy-recall curve of sea urchins, (b) the accuracy-recall curve of scallops, (c) the accuracy-recall curve of starfish, and (d) the accuracy-recall curve of sea cucumbers; the accuracy-recall curves of different categories are presented, from which it can be seen that the proposed DDRL method can achieve the best balance between accuracy and recall. In particular, for scallops, starfish, and sea cucumbers, the combination of the DAAMCL module and the UEFDL module can greatly improve the detection accuracy. The key reasons include: 1) By increasing the angular margin between the embedded features and the corresponding class weights in the cosine space, the designed DAAMCL module improves the intra-class compactness and inter-class differentiation; 2) The UEFDL module can adaptively transfer the distribution information of the embedded features of marine benthic organisms from the implicit head class to the tail class, improving the detection accuracy of the tail class while maintaining the detection performance of the head class.

[0222] 5 Quantitative comparison of classic long-tail object detection methods

[0223] To demonstrate the effectiveness and superiority of DDRL, the present invention comprehensively considers the reweighting methods (Class-Balanced Softmax, focal loss and LDAM) and data synthesis methods (Cutout, Random Erasing, GridMask and Poisson GAN), selects YOLOv3 as the benchmark model, and Table 8 summarizes the corresponding comparative experimental results. It can be seen that according to the mAP@.5, mAP@.75 and mAP@[.5,.95] indicator values, the Class-Balanced Softmax technology can achieve a higher detection accuracy than the benchmark model. In addition to the detection accuracy of the sea urchin class, the detection method based on focal loss can also surpass the benchmark model. Although the loss function with known label distribution can provide a larger margin for the tail class, the detection accuracy of the LDAM method is still lower than that of the DDRL method, mainly because the embedded feature distribution information of different categories is not considered. Regarding the data synthesis method, by randomly removing a fixed-size area, the Cutout method promotes the detection network to comprehensively learn the global semantic information of the target of interest. Especially when detecting sea cucumbers, the Random Erasing technique can significantly improve the detection performance by arbitrarily selecting square areas and replacing them with random values. Although the GridMask method avoids excessive deletion and retains continuous areas, the detection accuracy of the GridMask method is lower than that of DDRL except for sea urchins and sea cucumbers. Combining Poisson fusion technology and generative adversarial networks, the number of samples in the tail class is increased, and Poisson GAN achieves the best sea cucumber and mAP@.5 indicator values. According to the mAP@[.5,.95] indicator value, the proposed DDRL comprehensively improves the detection performance of marine benthic organisms.

[0224] Table 8 Quantitative comparison of classic long-tail target detection methods on the URPC2020 dataset

[0225]

[0226] Quantitative comparison of 6 special marine benthic organism detection methods

[0227] In order to verify the superiority of DDRL, the present invention considers comparing with a dedicated marine benthic biological detection method, including a lightweight underwater object detection (LUOD) method, a hybrid Transformer-based detection network (HTDN), TVFRD, XC-SSD, BBROABR and FCSA, and the corresponding experimental comparison results are shown in Table 9. It can be seen that since the LUOD method uses a lightweight backbone network MobileNet v2, it achieves the fastest detection speed (44FPS). Compared with the ResNet-50 feature extraction network, the XC-SSD based on ResNeXt-50 achieves higher detection accuracy, mainly because ResNeXt-50 adopts the grouping idea and increases the representation ability of the convolutional neural network. When detecting the sea cucumber category, XC-SSD achieves the best detection accuracy. HTDN uses the Transformer structure to enhance the ability to obtain global and local texture information, and obtains the highest accuracy when detecting sea urchins. Since TVFRD needs to generate a large number of candidate bounding boxes with the help of the region proposal network, the inference speed can only reach 3FPS. In addition, although BBROABR and FCSA enhance the detection performance from the perspectives of improving the bounding box regression positioning performance and recalibrating the weights of related feature maps, respectively, the detection accuracy is lower than that of the proposed DDRL. By developing DAAMCL and UEFDL modules, the proposed DDRL method transfers the variance of the feature distribution of the head class embedding to the tail class, surpassing other methods in terms of mAP@.5, mAP@.75, and mAP@[.5,.95] indicators.

[0228] Table 9 Quantitative comparison of dedicated marine benthic organism detection methods on the DUO dataset

[0229]

[0230] 7 Quantitative comparison of common object detection methods

[0231] Furthermore, the present invention comprehensively considers two-stage detection methods (Faster R-CNN), single-stage detection methods (SSD, EfficientDet, YOLOv4, YOLOv7 and RetinaNet) and anchor-free detection methods (FCOS and CenterNet), and the corresponding comparison results are summarized in Table 10. It can be seen that due to the lack of prior anchor information, anchor-free detection methods (i.e., FCOS and CenterNet) are difficult to achieve superior detection performance. Although YOLOv4 uses a more efficient backbone network CSPDarkNet-53 to extract high-level semantic information, it is difficult for YOLOv4 to construct discriminative embedding features due to the lack of additional angle margin strategy. YOLOv7 utilizes an extended efficient layer aggregation network, a hierarchical label assignment method from coarse-grained to fine-grained, and a convolution reparameterization strategy to effectively improve the feature extraction capability and network learning capability, and achieve the optimal value in terms of mAP@.75 indicators. The SSD and EfficientDet methods do not consider the transfer of embedded feature distribution information, so the detection accuracy of the SSD and EfficientDet methods is lower than that of DDRL. Faster R-CNN using ResNet-50 as the feature extraction network achieves higher detection accuracy than that using VGG-16. However, since Faster R-CNN needs to use the region proposal network to generate a large number of candidate boxes, it can only achieve an inference speed of 5FPS at a resolution of 1000×600. Combining the DAAMCL and UEFDL modules, the proposed DDRL surpasses other methods in terms of mAP@.5 and mAP@[.5,.95] indicators. The improvement in detection accuracy of sea cucumbers and scallops is higher than that of sea urchins, which shows that the UEFDL module plays a vital role in transferring the distribution information of head class embedding features.

[0232] Table 10 Quantitative comparison of general target detection methods on the URPC2020 dataset

[0233]

[0234] 8 Qualitative comparison of marine benthic organism detection

[0235] To demonstrate the detection superiority of DDRL in real environments, this paper considers two-stage detection methods (Faster R-CNN), single-stage detection methods (SSD, EfficientDet, YOLOv4 and YOLOv7) and anchor-free detection methods (FCOS). The corresponding experiments are carried out in dark light scenes, bright light scenes, small-scale target scenes and scenes where multiple categories of targets coexist. Figure 8 Qualitative comparison of detection performance in typical scenes, where (a) is low light, (b) is bright light, (c) is small scale, and (d) is multi-category target coexistence.

[0236] As can be seen from Figure (a), Faster R-CNN achieves good detection results in dark light scenes thanks to the high-quality bounding boxes generated by the region proposal network. Regarding the detection performance in bright light scenes, as shown in Figure (b), Faster R-CNN and EfficientDet misdetect seagrass as sea cucumbers, and SSD and YOLOv4 cannot correctly detect and identify marine benthic organisms located in rock cracks. Regarding the detection performance in small-scale marine benthic organisms, as shown in Figure (c), Faster R-CNN, YOLOv7 and DDRL can greatly reduce false detections. Since FCOS does not use prior anchor box information during training, it is difficult to accurately detect all marine benthic organisms. The SSD and EfficientDet methods do not construct discriminative embedded features, so they can only detect some marine benthic organisms. Regarding the detection performance in multi-category target coexistence scenes, as shown in Figure (d), DDRL can accurately detect marine benthic organisms that are occluded within and between classes.

[0237] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

[0238] [1]Spampinato C,Palazzo S,Joalland P,et al.Fine-grained objectrecognition in underwater visual data[J].Multimedia Tools and Applications,2016,75:1701-1720.

[0239] [2]Qin H, Li

[0240] [3]MaY,Liu M,ZhangY,etal.Imbalanced underwater acoustic targetrecognition with trigonometric loss and attention mechanism convolutionalnetwork[J].Remote Sensing,2022,14(16):4103.

[0241] [4]Xu S,Zhang M,Song W,et al.A systematic review and analysis of deeplearning-based underwater object detection[J].Neurocomputing,2023,527:204-232.

[0242] [5]NieW,RenJ,ChengG,et al.Tackling long-tail data distributionproblem of deep learning based underwater target recognition system[C].OCEANSConference,San Diego,USA,2021:1-6.

[0243] [6]LiY,SunK,Han Z.Vision technology in underwater:Applications,challenges and perspectives[C].The 4th International Conference on Controland Robotics,Guangzhou,China,2022:369-378.

[0244] [7]PangY,Wu C,Wu H,et al.Over-sampling strategy-based class-imbalanced salient object detection and its application in underwater scene[J].The Visual Computer,2023,39(5):1959-1974.

[0245] [8]Zhang Z,Liu Y,Zhu X,et al.DSE-FCOS:Dilated and SE block-reinforcedFCOS for detection of marine benthos[J].The Visual Computer,2023,39(11):5199-5208.

[0246] [9]Wen Y,Zhang K,Li Z,et al.A discriminative feature learningapproach for deep face recognition[C].European Conference on Computer Vision,Amsterdam,Netherlands,2016:499-515.

[0247]

[10] Liu W,Wen Y,Raj B,et al.Sphereface revived:Unifyinghyperspherical face recognition[J].IEEE Transactions on Pattern Analysis andMachine Intelligence,2022,45(2):2458-2474.

[0248]

[11] Wang H,Wang Y,Zhou Z,et al.Cosface:Large margin cosine loss fordeep face recognition[C].IEEE Conference on Computer Vision and PatternRecognition,Salt Lake City,USA,2018:5265-5274.

[0249]

[12] Wang J,Qi S,Wang C,et al.B-YOLOX-S:A lightweight method forunderwater object detection based on data augmentation and multiscale featurefusion[J].Journal of Marine Science and Engineering,2022,10(11):1764.

[0250]

[13] Liu C,Wang Z,Wang S,et al.A new dataset,Poisson GAN and AquaNetfor underwater object grabbing[J].IEEE Transactions on Circuits and Systemsfor Video Technology,2022,32(5):2831-2844.

[0251]

[14] Chen L,Dong J,Zhou H.Class balanced underwater object detectiondataset generated by class-wise style augmentation[J].arXiv preprint:2101.07959,2021.

[0252]

[15] Buda M,Maki A,Mazurowski M A.A systematic study of the classimbalance problem in convolutional neural networks[J].Neural Networks,2018,106:249-259.

Claims

1. A method for detecting benthic organisms based on deep discriminant representation learning under long-tail distribution, characterized in that: The following steps are involved: Construct a dataset based on images of any one or more marine benthic organisms including sea urchins, scallops, starfish, and sea cucumbers; Divide the dataset into training and testing sets; Construct a marine benthic organism detection model based on deep discriminant representation learning under long-tail distribution; Based on the training set data, the marine benthic organism detection model based on deep discriminant representation learning under long-tail distribution is trained to obtain the trained marine benthic organism detection model based on deep discriminant representation learning under long-tail distribution; The test set data is input into the marine benthic organism detection model based on deep discriminant representation learning under the trained long-tail distribution to achieve the detection and identification of marine benthic organisms such as sea urchins, scallops, starfish and sea cucumbers.

2. The method for detecting benthic organisms based on deep discriminant representation learning under long-tail distribution according to claim 1, characterized in that: The marine benthic organism detection model based on deep discriminant representation learning under the long-tail distribution includes: Backbone network: extracts the informative feature graph contained in the image; Classification loss module based on discriminative additional angle margin: Based on the informative feature map extracted by the backbone network, combined with the regional alignment mechanism, the bilinear interpolation method is used to construct the characteristic values ​​of the sampling points of the marine benthic organisms in the area of ​​interest, and the average aggregation strategy is used to obtain the single-interval marine benthic organisms. The embedded features of the marine benthic organisms are constructed with the help of dimensional conversion operations, and the angle between the embedded features and the corresponding category weights is calculated. By increasing the angle margin in the cosine space, the discriminability of the embedded features of the marine benthic organisms is improved; Uneven embedding feature distribution loss module: A sliding average strategy is used to update the embedding feature centers of various types of marine benthic organisms, and the angles between the embedded features and their corresponding centers are solved in turn to calculate the differences in the embedding feature distribution between the tail class and other head classes of marine benthic organisms.

3. The method for detecting benthic organisms based on deep discriminant representation learning under a long-tail distribution according to claim 1, characterized in that: The classification loss module based on discriminative additional angle margin includes: Informative feature map construction submodule: used to construct the informative feature map extracted by the backbone network; A submodule for estimating characteristic values ​​of a region of interest: used for estimating characteristic values ​​of a region of interest based on the informative characteristic map transmitted by the informative characteristic map construction submodule; Discriminative additional angle margin classification loss submodule: a loss function is designed based on the eigenvalues ​​of the region of interest estimated by the eigenvalue estimation submodule of the region of interest.

4. The method for detecting benthic organisms based on deep discriminant representation learning under a long-tail distribution according to claim 1, characterized in that: The process of constructing the informative feature map extracted by the backbone network is as follows: The informative feature map extracted by the backbone network is expressed as: Among them, E(·; ·) is the feature mapping function in the detection network, Represents the height h r , width is w r , batch size is n b and the number of channels is n rc The original input sample, θ e represents the parameter to be optimized, represents the feature map, h and w are the height and width of the feature map, n c is the number of embedding feature layers. Using the batch normalization method, it is expressed as: Where b = 1, 2, L, n b and l=1,2,L,n c represents the sample index and feature layer index, γ and β represent the scale and offset coefficients respectively, ∈ is a constant, The coordinates of the lth feature layer in the bth sample are (x m ,y n ), μ l and σ l are the mean and standard deviation of the features indexed at the lth layer, expressed as: in, and Represents n c The mean and standard deviation of each channel; Taking the b-th sample as an example, the construction process of discriminative features is explained. In order to increase the nonlinear mapping capability, LeakyReLU is used as the activation function, which is expressed as: Among them, 0<κ<1 represents the slope of the activation function, H l (x m ,y n ) represents the coordinates in the lth feature layer (x m ,y n ) is the activation eigenvalue at .

5. The method for detecting benthic organisms based on deep discriminant representation learning under long-tail distribution according to claim 1, characterized in that: The process of estimating the feature value of the region of interest based on the informative feature map transmitted by the informative feature map construction submodule is as follows: The coordinates of the feature points in the region of interest are expressed as: in, is the coordinate of the upper left corner of the region of interest, w bin and h bin Represents the width and height of a single interval, g = 1, L, n p and k=1,L,n q are the column and row dimension indices of the interval, u=1,L,n col and v=1,L,n row Respectively represent the index of the sampling point in a single interval in the column dimension and row dimension, (x g,u ,y k,v ) are the coordinates of the sampling points; By means of the bilinear interpolation method, for the i-th instance in the l-th feature layer, the coordinate position is (x g,u ,y n ) at the eigenvalue It is expressed as: Among them, (x m ,y n ) and (x m+1 ,y n ) is the closest sampling point (x g,u ,y k,v )'s upper left and upper right integer coordinates; Similarly, the coordinates (x g,u ,y n+1 ) at the eigenvalue It is expressed as: Among them, (x m ,y n+1 ) and (x m+1 ,y n+1 ) is the closest to (x g,u ,y k,v )’s lower left and lower right corner integer coordinate positions; Furthermore, the coordinates are (x g,u ,y k,v ) at the eigenvalue It is expressed as: After training m batches, for the i-th instance in the l-th feature layer, the feature value at interval index (g, k) It is expressed as: in, Represents the embedded feature of the i-th marine benthic organism instance at the l-th layer. Considering all feature layers, the embedded feature f of the i-th marine benthic organism instance i m It is expressed as: in, represents the embedded features obtained by the dimension transformation operation T(·), d = n q ×n p ×n c , Represents n c The embedded features of the channels.

6. The method for detecting benthic organisms based on deep discriminant representation learning under long-tail distribution according to claim 1, characterized in that: The process of designing the loss function based on the eigenvalue of the region of interest estimated by the eigenvalue estimation submodule of the region of interest is as follows: Normalized embedding features and weights are expressed as: in, yes The jth column of non-normalized weights, j = 1, L, n tc Represents the category index, n tc represents the total number of categories, ‖·‖2 represents the 2-norm, and Represent the normalized embedding features and normalized weights respectively; The i-th embedding feature and the jth class weight The angle between them is expressed as: in, Discriminative additional angle margin classification loss function l DAAMCL It is expressed as: Where s is the scale parameter, δ represents the additional angle margin, Represents embedded features With corresponding category weight The angle between inst Represents the number of sample instances in this batch.

7. The method for detecting benthic organisms based on deep discriminant representation learning under long-tail distribution according to claim 2, characterized in that: The distribution loss module of the uneven embedding feature includes: The embedding feature center construction submodule is used to transfer the distribution information of the embedded features of marine benthic organisms from the implicit head class to the tail class, propose an uneven embedded feature distribution loss, enhance the diversity of the embedded features of the tail class of marine benthic organisms, and construct the embedded features of each category respectively; In-class embedded feature distribution variance calculation submodule: used to calculate the embedding feature distribution variance of each category of embedded features constructed by the embedding feature center construction submodule; Uneven embedding feature distribution loss submodule: used for uneven embedding feature distribution loss function of all categories based on the embedding feature distribution variance calculated by the intra-class embedding feature distribution variance calculation submodule.

8. The method for detecting benthic organisms based on deep discriminant representation learning under long-tail distribution according to claim 7, characterized in that: By transferring the distribution information of the embedded features of marine benthic organisms from the implicit head class to the tail class, an uneven embedded feature distribution loss is proposed to enhance the diversity of the embedded features of the tail class of marine benthic organisms. The process of constructing the embedded features of each category is as follows: Construct the embedding features of each category and calculate the corresponding embedding feature center, expressed as: Among them, Ω∈{ec,sc,st,ho}, ec, sc, st and ho represent the categories of sea urchin, scallop, starfish and sea cucumber respectively. represents the i-th embedded feature in the corresponding category after training m batches, represents the center of the embedded feature, Represents the number of samples of the corresponding category after training m batches; Adopt the sliding average mechanism to make full use of the embedding feature center of the m-1th batch To update the embedding feature center of the mth batch It is expressed as: in, λ represents the class center update rate.

9. The method for detecting benthic organisms based on deep discriminant representation learning under long-tail distribution according to claim 7, characterized in that: The process of constructing each category of embedded features for the embedding feature center construction submodule to calculate the embedding feature distribution variance is as follows: The angle between the embedded feature and its corresponding category center is expressed as: in, is the i-th embedding feature after training m batches Its corresponding center Angle. Average angle distribution of different categories It is expressed as: Then, the intra-class embedding feature distribution variance is expressed as: in, and Represents the variance of the intra-class embedding feature distribution of sea urchins, scallops, starfish, and sea cucumbers after training for m batches.

10. The method for detecting benthic organisms based on deep discriminant representation learning under long-tail distribution according to claim 7, characterized in that: Based on the embedding feature distribution variance calculated by the intra-class embedding feature distribution variance calculation submodule, the process of the uneven embedding feature distribution loss function for all categories is as follows: The uneven embedding feature distribution loss is expressed as: in, and Representatives from the collection The first largest category of embedded feature distribution variance, the second largest category of embedded feature distribution variance, the third largest category of embedded feature distribution variance and the fourth largest category of embedded feature distribution variance, n 1st 、n 2nd 、n 3rd and n 4th is the corresponding number of instances, and Represents the uneven embedding feature distribution loss for different categories. Finally, the uneven embedding feature distribution loss l for all categories UEFDL It is expressed as:

Images