A combined fleet cargo status management system with intelligent monitoring

Abstract

The present invention relates to the technical field of cargo management, and discloses an intelligent monitoring combined fleet cargo status management system, a fleet construction module, used for constructing a fleet according to a number of ships, the fleet comprising a wave-breaking unit, a transport unit and a data acquisition unit; a data acquisition module, used for acquiring a first image, a second image and a third image of a fleet transport unit, a front side of a fleet path and a rear side of a fleet path at fixed time intervals according to the data acquisition unit; a sea wave recognition module, used for jointly recognizing the first image, the second image and the third image through a preset recognition model to obtain a wave-breaking score of the wave-breaking unit; and a dynamic scheduling module, used for dynamically optimizing the travel speed and travel direction of each ship a according to the wave-breaking score of the wave-breaking unit in combination with an optimization algorithm, so as to realize stable transportation of target cargo by the transport unit.

Description

Technical Field

[0001] The present invention relates to the technical field of cargo management, and more particularly to an intelligently monitored combined fleet cargo status management system. Background Art

[0002] At present, the sea transport is highly sensitive to vibration and environment. However, due to the long sailing time, when encountering bad sea conditions, the ship needs to move slowly to maintain the stability of the goods, resulting in insufficient timeliness. In the existing technology, the fleet management system mainly focuses on single-ship monitoring and scheduling, but lacks intelligent collaborative optimization of the fleet as a whole. Summary of the invention

[0003] The present invention provides an intelligent monitoring combined fleet cargo status management system to solve the technical problems raised in the background technology.

[0004] The present invention provides a combined fleet cargo status management system with intelligent monitoring, comprising:

[0005] A fleet building module, used to build a fleet based on a number of ships, the fleet including a wave-breaking unit, a transport unit and a data collection unit;

[0006] A data acquisition module, used for acquiring, according to the data acquisition unit, a first image, a second image and a third image of a fleet transport unit, a front side of a fleet path and a rear side of a fleet path at fixed time intervals;

[0007] A wave recognition module, used to jointly recognize the first image, the second image and the third image through a preset recognition model to obtain a wave breaking score of a wave breaking unit;

[0008] The dynamic scheduling module is used to dynamically optimize the travel speed and travel direction of the wave-breaking unit according to the wave-breaking score of the wave-breaking unit in combination with the optimization algorithm, so as to realize the stable transportation of the target goods by the transport unit.

[0009] Furthermore, the fleet includes a wave-breaking unit, a transport unit and a data collection unit:

[0010] Allocate A ships a as wave-breaking units of the fleet, and the wave-breaking units are used to break waves for transport units based on ships a;

[0011] Allocate B ships b as the transport units of the fleet, and the transport units are used to transport the target goods;

[0012] Multiple drones are assigned as data acquisition units of the fleet, and the data acquisition units are used to collect images.

[0013] Further, the identification model includes:

[0014] According to the shooting time of the first image, the shooting time is taken as the center time of the preset time window, and the first image in the preset time window is obtained. The acceleration vector of ship a is , ;in, Indicates the first The moment The acceleration parameters of ship a, ;

[0015] The wave breaking score corresponding to the first image is calculated based on the acceleration vector a of each ship. The calculation formula is as follows:

[0016] ;

[0017] ;

[0018] in, represents the wave breaking score of the first image, Represents the acceleration vector The inspection score, The index representing the number of acceleration vectors, and Indicates the start and end time of the preset time window. Indicates the total time length of the preset time window. represents the shooting time of the first image, Represents the acceleration vector Middle The acceleration parameter at each moment, Indicates The Euclidean distance between the position of ship a and the center of the fleet, , and Represent the acceleration vectors Middle The components of the acceleration parameters at a certain moment in the horizontal, vertical and vertical directions are: represents the weight parameter of the inspection score, Represents the preset time window weight parameter, Indicates The weight parameter of the position of ship a;

[0019] The first image is used as a training sample, and the wave-breaking score is used as a sample label of the training sample to train a recognition model, wherein the recognition model includes an image processing layer and a hidden layer;

[0020] The output of the recognition model represents the predicted wave-breaking score corresponding to the first image, and the mean error variance of the predicted wave-breaking score and the sample label is used as the loss function of the recognition model to back-propagate and update the hyperparameters of the image processing layer and the hidden layer of the recognition model.

[0021] Furthermore, an image processing layer of the recognition model is used to perform feature conversion on the first image, including:

[0022] Performing semantic segmentation on the first image to obtain a semantic matrix, the semantic matrix including a number of ship elements and sea surface elements; performing eight-neighborhood clustering on the ship elements to obtain a number of ship clusters;

[0023] Taking the central element of each ship cluster as the center, the edge elements of the corresponding ship cluster are amplified according to a preset ratio to obtain an amplified cluster, and the elements in the amplified cluster are replaced by 0 to obtain a segmentation matrix; wherein the preset ratio is greater than 1;

[0024] Detecting the edge of the segmentation matrix to obtain several characteristic edges of the segmentation matrix;

[0025] Fill the edges of the positions of the augmented clusters in the segmentation matrix corresponding to the semantic matrix, specifically including:

[0026] Cluster several feature edges according to their angles, and obtain the edge angle of the maximum cluster of feature edges;

[0027] Obtain the density of feature edges, where the density of feature edges is the ratio of the number of elements of feature edges in the segmentation matrix to the number of elements of non-amplified clusters in the semantic matrix;

[0028] For the position of the augmented cluster corresponding to the semantic matrix in the segmentation matrix, the edges are randomly filled according to the edge angle and density to obtain the first feature transformation map; wherein the edge angle of the filled edge is the same as the edge angle of the maximum cluster of feature edges, and the ratio of the number of elements of the filled edge to the number of elements of the augmented cluster in the semantic matrix is ​​equal to the density.

[0029] Furthermore, the image processing layer of the recognition model is further used to perform feature conversion on the second image and the third image, including:

[0030] Perform edge detection on the second image or the third image to obtain a second feature conversion map or a third feature conversion map, respectively.

[0031] Further, the hidden layer of the identification layer is used to input the first feature conversion map, the second feature conversion map or the third feature conversion map. The calculation formula of the hidden layer is as follows:

[0032] ;

[0033] ;

[0034] in, represents the wave breaking score output by the hidden layer for the first feature conversion map, the second feature conversion map, or the third feature conversion map, represents a first characteristic conversion diagram, a second characteristic conversion diagram, or a third characteristic conversion diagram, represents a left singular matrix, represents the singular value matrix, represents a right singular matrix, represents the weight vector of the hidden layer, represents the bias vector of the hidden layer, express, represents the bias matrix of the hidden layer, represents the weight matrix of the hidden layer, represents the matrix eigenvalue decomposition operation, represents the vector transpose operation, represents the vector concatenation operation, express Activation function, express Activation function.

[0035] Furthermore, the transport unit is used to stably transport the target goods, including:

[0036] Step 1: Initialize and generate several individuals that meet the constraints to construct the target population;

[0037] Individuals pass through the feature matrix It is expressed as: ;in, Indicates The ship's travel angle, Indicates The speed of the ship, ;

[0038] The constraints include: ;in, and They represent the maximum and minimum permissible travel angles of ship b respectively;

[0039] ;in, and They represent the maximum and minimum permissible speeds of ship b respectively;

[0040] in, represents the driving speed threshold, Indicates The speed of the ship, Indicates The speed of the ship, , ;

[0041] Step 2: The objective function is constructed based on the recognition model, which includes:

[0042] Get the value based on each individual The second image at the moment The first image at the moment The third image at a moment;

[0043] Based on the recognition model, the The second image at the moment The first image at the moment and the The breaking wave score of the third image at the moment , and ;

[0044] Objective Function The calculation formula is:

[0045] ;in, and Represent the first penalty term coefficient and the second penalty term coefficient respectively, represents the ideal breaking wave constant;

[0046] Step 3, calculate the objective function value of each individual, and sort them from small to large according to the objective function value to obtain the feature ranking;

[0047] Step 4: retain a preset number of individuals in the feature sorting from front to back, and update the remaining individuals as parents;

[0048] Step 5, loop step 3 and step 4 for a preset number of times, stop updating and iterating, output the individual with the smallest objective function value, and set the driving speed and driving angle of each ship b based on the individual with the smallest objective function value to achieve stable transportation of the target cargo by the transport unit.

[0049] Furthermore, the remaining individuals are used as parents for update iteration, and the update formula is as follows:

[0050] ;

[0051] in, Indicates The update matrix of the parent individual at the iteration, Indicates The matrix of the parent individuals at the iteration, represents the updated weight, Indicates The matrix of the individual with the highest feature ranking at the iteration, Indicates size A sparse matrix whose diagonal elements are 1 and the rest are 0.

[0052] The beneficial effect of the present invention is that by constructing a fleet cargo status management system with intelligent monitoring and dynamic optimization, it integrates multiple module functions such as fleet construction, data collection, wave identification and dynamic scheduling, and realizes efficient cargo transportation stability guarantee, which is particularly suitable for the transportation of highly sensitive cargo. Compared with the traditional single-ship monitoring or static scheduling method, this solution can intelligently and collaboratively optimize the fleet as a whole in complex sea conditions, effectively balancing transportation efficiency and cargo safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a module diagram of an intelligent monitoring combined fleet cargo status management system of the present invention. DETAILED DESCRIPTION

[0054] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0055] like Figure 1 As shown, a combined fleet cargo status management system with intelligent monitoring includes:

[0056] A fleet building module, used to build a fleet based on a number of ships, the fleet including a wave-breaking unit, a transport unit and a data collection unit;

[0057] A data acquisition module, used for acquiring, according to the data acquisition unit, a first image, a second image and a third image of the fleet transport unit, a front side of the fleet path and a rear side of the fleet path at fixed time intervals;

[0058] A wave recognition module, used to jointly recognize the first image, the second image and the third image through a preset recognition model to obtain a wave breaking score of a wave breaking unit;

[0059] The dynamic scheduling module is used to dynamically optimize the travel speed and travel direction of the wave-breaking unit according to the wave-breaking score of the wave-breaking unit in combination with the optimization algorithm, so as to realize the stable transportation of the target goods by the transport unit.

[0060] Specifically, the present invention constructs a fleet and assigns different functions to different ships, including transport ships and wave-breaking ships. The transport ships are used to transport goods that are sensitive to shaking, while the wave-breaking ships are used to break waves for the transport ships, so as to achieve stable operation of the transport ships and further achieve stable transportation management of the goods.

[0061] In one embodiment of the present invention, the fleet includes a wave-breaking unit, a transport unit and a data collection unit:

[0062] Allocate A ships a as wave-breaking units of the fleet, and the wave-breaking units are used to break waves for transport units based on ships a;

[0063] Allocate B ships b as the transport units of the fleet, and the transport units are used to transport the target goods;

[0064] Multiple drones are assigned as data acquisition units of the fleet, and the data acquisition units are used to collect images.

[0065] Specifically, the fleet consists of three units: wave-breaking units, transport units, and data collection units. These units are responsible for different tasks and work together to achieve efficient and safe transportation of goods. The wave-breaking unit consists of A ships. These ships are specially designated as "wave-breaking ships" and are responsible for reducing wave resistance for the transport units of the fleet during navigation by regulating their own position, speed, and heading. The task of the wave-breaking unit is to reduce the direct surge impact on the transport unit by interfering with the waves in front (similar to the "opening" effect). This wave-breaking strategy not only improves the stability of the transport unit, but also reduces the shaking of the cargo, which is particularly suitable for vibration-sensitive cargo. The transport unit consists of B ships, which are responsible for carrying and transporting the target cargo. These ships are the core load units of the fleet, and their main goal is to ensure the safe transportation of cargo. During the entire transportation process, the transport unit mainly relies on the protection of the wave-breaking unit, and adjusts its own speed and direction through the dynamic scheduling module to avoid damage to the cargo due to changes in sea conditions. The data collection unit consists of drones, which are responsible for collecting images and data related to the transportation task from the air.

[0066] The drone takes the first, second and third images of the fleet at fixed time intervals, corresponding to the sea conditions in front of the transport unit and the rear of the fleet path. The real-time images and data collected by the drone are passed to the "wave recognition module" to analyze the wave characteristics and calculate the wave breaking score of the wave breaking unit. These data provide key basis for the dynamic scheduling module and support the optimization of the speed and direction of the transport unit. Through the clear division of functional modules (wave breaking, transportation, data collection), the system realizes the distributed processing of tasks and avoids the limitations of the single-ship management model.

[0067] Preferably, the ship b in the wave-breaking unit can transport some goods that are not sensitive to shaking, thereby increasing revenue.

[0068] In one embodiment of the present invention, the identification model includes:

[0069] According to the shooting time of the first image, the shooting time is taken as the center time of the preset time window, and the first image in the preset time window is obtained. The acceleration vector of ship a is , ;in, Indicates the first The moment The acceleration parameters of ship a, ;

[0070] The wave breaking score corresponding to the first image is calculated based on the acceleration vector a of each ship. The calculation formula is as follows:

[0071] ;

[0072] ;

[0073] in, represents the wave breaking score of the first image, Represents the acceleration vector The inspection score, The index representing the number of acceleration vectors, and Indicates the start and end time of the preset time window. Indicates the total time length of the preset time window. represents the shooting time of the first image, Represents the acceleration vector Middle The acceleration parameter at each moment, Indicates The Euclidean distance between the position of ship a and the center of the fleet, , and Represent the acceleration vectors Middle The components of the acceleration parameters at a certain moment in the horizontal, vertical and vertical directions are: represents the weight parameter of the inspection score, Represents the preset time window weight parameter, Indicates The weight parameter of the position of ship a;

[0074] The first image is used as a training sample, and the wave-breaking score is used as a sample label of the training sample to train a recognition model, wherein the recognition model includes an image processing layer and a hidden layer;

[0075] The output of the recognition model represents the predicted wave-breaking score corresponding to the first image, and the mean error variance of the predicted wave-breaking score and the sample label is used as the loss function of the recognition model to back-propagate and update the hyperparameters of the image processing layer and the hidden layer of the recognition model.

[0076] Specifically, since it is difficult to collect data other than pictures from the second and third images, the acceleration unit (such as an accelerometer) installed on ship a collects acceleration parameters in real time, and then converts the acceleration parameters into actual wave breaking scores (that is, the degree of sea surface turbulence corresponding to the transport unit), and then uses the first image as a training sample for the recognition model, and the actual wave breaking score as a sample label for the recognition model. In this way, the recognition model is obtained by constructing a data set for training.

[0077] In one embodiment of the present invention, an image processing layer of the recognition model, the image processing layer is used to perform feature conversion on the first image, including:

[0078] Performing semantic segmentation on the first image to obtain a semantic matrix, the semantic matrix including a number of ship elements and sea surface elements; performing eight-neighborhood clustering on the ship elements to obtain a number of ship clusters;

[0079] Taking the central element of each ship cluster as the center, the edge elements of the corresponding ship cluster are amplified according to a preset ratio to obtain an amplified cluster, and the elements in the amplified cluster are replaced by 0 to obtain a segmentation matrix; wherein the preset ratio is greater than 1;

[0080] Detecting the edge of the segmentation matrix to obtain several characteristic edges of the segmentation matrix;

[0081] Fill the edges of the positions of the augmented clusters in the segmentation matrix corresponding to the semantic matrix, specifically including:

[0082] Cluster several feature edges according to their angles, and obtain the edge angle of the maximum cluster of feature edges;

[0083] Obtain the density of feature edges, where the density of feature edges is the ratio of the number of elements of feature edges in the segmentation matrix to the number of elements of non-amplified clusters in the semantic matrix;

[0084] For the position of the augmented cluster corresponding to the semantic matrix in the segmentation matrix, the edges are randomly filled according to the edge angle and density to obtain the first feature transformation map; wherein the edge angle of the filled edge is the same as the edge angle of the maximum cluster of feature edges, and the ratio of the number of elements of the filled edge to the number of elements of the augmented cluster in the semantic matrix is ​​equal to the density.

[0085] It should be noted that the semantic matrix includes several cluster elements of ships, as follows:

[0086] ; Assume that 1 represents a ship in the semantic matrix, and 0 represents the sea surface. Then, for the four ship elements, a cluster is formed.

[0087] Specifically, the first image includes ship a, while the second image and the third image are both sea surfaces, and the input samples are not uniform. The first image, the second image, and the third image all need to be uniformly input, so the first image is feature transformed. First, the first image is recognized by the existing semantic segmentation model to obtain a semantic matrix, so that the elements corresponding to the ship can be deleted. Because the ship will generate ripples when it is moving, it is necessary to enlarge the ship and then delete it together with the ripples. After deleting the ship and the ripples, edge detection is performed on the deleted segmentation matrix. Edge detection can detect waves or ripples on the sea surface. Since the waves or ripples are in the same direction in the same area, the angle of the edge can be obtained. At the same time, according to the ratio calculation of the pixel points, the density of the edge can also be obtained, and then the deleted part in the matrix is ​​automatically filled, keeping the density and angle of the edge unchanged, and obtaining the edge matrix (first feature conversion diagram).

[0088] In one embodiment of the present invention, the image processing layer of the recognition model is further used to perform feature conversion on the second image and the third image, including:

[0089] Perform edge detection on the second image or the third image to obtain a second feature conversion map or a third feature conversion map, respectively.

[0090] Specifically, the second feature conversion map and the third feature conversion map are directly generated by edge detection on the second image or the third image, and the input consistency with the first feature conversion map is maintained.

[0091] In one embodiment of the present invention, the hidden layer of the identification layer is used to input the first feature conversion map, the second feature conversion map or the third feature conversion map, and the calculation formula of the hidden layer is as follows:

[0092] ;

[0093] ;

[0094] in, represents the wave breaking score output by the hidden layer for the first feature conversion map, the second feature conversion map, or the third feature conversion map, represents a first characteristic conversion diagram, a second characteristic conversion diagram, or a third characteristic conversion diagram, represents a left singular matrix, represents the singular value matrix, represents a right singular matrix, represents the weight vector of the hidden layer, represents the bias vector of the hidden layer, express, represents the bias matrix of the hidden layer, represents the weight matrix of the hidden layer, represents the matrix eigenvalue decomposition operation, represents the vector transpose operation, represents the vector concatenation operation, express Activation function, express Activation function.

[0095] Specifically, based on the input feature conversion map, the output breaking wave score is calculated through matrix operations and activation functions. First, the input feature conversion map matrix Perform eigenvalue decomposition to obtain the singular value matrix of the matrix , left singular matrix , right singular matrix , and through specific matrix operations and nonlinear transformations (such as Activation function) generates a specific feature expression. At the same time, combined with the weight matrix of the hidden layer , position vector weight , bias matrix and the weight matrix of the output layer , the calculation results are comprehensively weighted and offset, and finally The activation function outputs the breaking wave score This model is able to effectively extract complex features of the input image and quantify their impact on the transportation process.

[0096] In one embodiment of the present invention, the stable transportation of target goods by the transport unit is realized, including:

[0097] Step 1: Initialize and generate several individuals that meet the constraints to construct the target population;

[0098] Individuals pass through the feature matrix It is expressed as: ;in, Indicates The ship's travel angle, Indicates The speed of the ship, ;

[0099] The constraints include: ;in, and They represent the maximum and minimum permissible travel angles of ship b respectively;

[0100] ;in, and They represent the maximum and minimum permissible speeds of ship b respectively;

[0101] in, represents the driving speed threshold, Indicates The speed of the ship, Indicates The speed of the ship, , ;

[0102] Step 2: The objective function is constructed based on the recognition model, which includes:

[0103] Get the value based on each individual The second image at the moment The first image at the moment, The third image at the moment;

[0104] Based on the recognition model, the The second image at the moment The first image at the moment and the The breaking wave score of the third image at the moment , and ;

[0105] Objective Function The calculation formula is:

[0106] ;in, and Represent the first penalty term coefficient and the second penalty term coefficient respectively, represents the ideal breaking wave constant;

[0107] Step 3, calculate the objective function value of each individual, and sort them from small to large according to the objective function value to obtain the feature ranking;

[0108] Step 4: retain a preset number of individuals in the feature sorting from front to back, and update the remaining individuals as parents;

[0109] Step 5, loop step 3 and step 4 for a preset number of times, stop updating and iterating, output the individual with the smallest objective function value, and set the driving speed and driving angle of each ship b based on the individual with the smallest objective function value to achieve stable transportation of the target cargo by the transport unit.

[0110] Specifically, the stable transportation of target goods by the transport unit is achieved through an intelligent optimization algorithm, which specifically includes the following steps. First, a number of individuals that meet the constraints are initialized and generated. Each individual is represented by a feature matrix composed of the ship's driving angle and speed, and is subject to the maximum and minimum driving angles and speeds. Secondly, the wave breaking score and objective function value corresponding to each individual at different times are calculated based on the recognition model. The objective function combines the ideal wave breaking constant and the penalty weight to measure the stability and optimization degree of transportation. Then, the objective function values ​​of all individuals are sorted, and the high-quality individuals are retained as the parent generation, and the parent generation is updated and iterated until the preset number of times is reached. Finally, the individual with the smallest objective function value is output, and the driving speed and angle of the ship are set according to the individual to ensure that the transport unit achieves smooth transportation of goods.

[0111] In one embodiment of the present invention, the remaining individuals are used as parents for update iteration, and the update formula is as follows:

[0112] ;

[0113] in, Indicates The update matrix of the parent individual at the iteration, Indicates The matrix of the parent individuals at the iteration, represents the updated weight, Indicates The matrix of the individual with the highest feature ranking at the iteration, Indicates size A sparse matrix whose diagonal elements are 1 and the rest are 0.

[0114] Specifically, the remaining individuals are iteratively optimized as parents through the update formula. Specifically, the update matrix Indicates The new matrix of the parent individual at the iteration is obtained by passing the current parent matrix The adjustment method includes adjusting the matrix of the best individual in the feature ranking With the current matrix The difference, after sparse matrix The sparse matrix is ​​updated after the weighted transformation. It is a matrix with diagonal elements of 1 and other elements of 0, which is used to constrain the optimization range. Update weights The step size is controlled and adjusted so that the parent matrix gradually approaches the individual with the best characteristics. Through multiple iterative optimizations, this method can effectively improve the quality of individual characteristics and provide optimized matrix parameters for achieving the goal.

[0115] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. An intelligent monitoring combined fleet cargo status management system, characterized in that: include: A fleet building module, used to build a fleet based on a number of ships, the fleet including a wave-breaking unit, a transport unit and a data collection unit; A data acquisition module, used for acquiring, according to the data acquisition unit, a first image, a second image and a third image of the fleet transport unit, a front side of the fleet path and a rear side of the fleet path at fixed time intervals; A wave recognition module, used to jointly recognize the first image, the second image and the third image through a preset recognition model to obtain a wave breaking score of a wave breaking unit; The dynamic scheduling module is used to dynamically optimize the speed and direction of the wave-breaking unit according to the wave-breaking score of the wave-breaking unit in combination with the optimization algorithm to achieve stable transportation of the target goods by the transport unit. Specifically, it includes: Step 1: Initialize and generate several individuals that meet the constraints to construct the target population; Individuals pass through the feature matrix It is expressed as: ;in, Indicates The ship's travel angle, Indicates The speed of the ship, ; The constraints include: ;in, and They represent the maximum and minimum permissible travel angles of ship b respectively; ;in, and They represent the maximum and minimum permissible speeds of ship b respectively; in, represents the driving speed threshold, Indicates The speed of the ship, Indicates The speed of the ship, , ; Step 2: The objective function is constructed based on the recognition model, which includes: Get the value based on each individual The second image at the moment The first image at the moment, The third image at the moment; Based on the recognition model, the The second image at the moment The first image at the moment and the The breaking wave score of the third image at the moment , and ; Objective Function The calculation formula is: ;in, and Represent the first penalty term coefficient and the second penalty term coefficient respectively, represents the ideal breaking wave constant; Step 3, calculate the objective function value of each individual, and sort them from small to large according to the objective function value to obtain the feature ranking; Step 4: retain a preset number of individuals in the feature sorting from front to back, and update the remaining individuals as parents; Step 5, loop step 3 and step 4 for a preset number of times, stop updating and iterating, output the individual with the smallest objective function value, and set the driving speed and driving angle of each ship b based on the individual with the smallest objective function value to achieve stable transportation of the target cargo by the transport unit.

2. According to claim 1, a combined fleet cargo status management system with intelligent monitoring is characterized in that: The fleet includes wave-breaking units, transport units and data collection units: Allocate A ships a as wave-breaking units of the fleet, and the wave-breaking units are used to break waves for transport units based on ships a; Allocate B ships b as the transport units of the fleet, and the transport units are used to transport the target goods; Multiple drones are assigned as data acquisition units of the fleet, and the data acquisition units are used to collect images.

3. The intelligent monitoring combined fleet cargo status management system according to claim 2 is characterized in that: Identify models, including: According to the shooting time of the first image, the shooting time is taken as the center time of the preset time window, and the first image in the preset time window is obtained. The acceleration vector of ship a is , ;in, Indicates the first The moment The acceleration parameters of ship a, ; The wave breaking score corresponding to the first image is calculated based on the acceleration vector a of each ship. The calculation formula is as follows: ； ； in, represents the wave breaking score of the first image, Represents the acceleration vector The inspection score, and Indicates the start and end time of the preset time window. Indicates the total time length of the preset time window. represents the shooting time of the first image, Represents the acceleration vector Middle The acceleration parameter at each moment, Indicates The Euclidean distance between the position of ship a and the center of the fleet, , and Represent the acceleration vectors Middle The components of the acceleration parameters at a certain moment in the horizontal, vertical and vertical directions are: represents the weight parameter of the inspection score, Represents the preset time window weight parameter, Indicates The weight parameter of the position of ship a; The first image is used as a training sample, and the wave-breaking score is used as a sample label of the training sample to train a recognition model, wherein the recognition model includes an image processing layer and a hidden layer; The output of the recognition model represents the predicted wave-breaking score corresponding to the first image, and the mean error variance of the predicted wave-breaking score and the sample label is used as the loss function of the recognition model to back-propagate and update the hyperparameters of the image processing layer and the hidden layer of the recognition model.

4. The intelligent monitoring combined fleet cargo status management system according to claim 3 is characterized in that: The image processing layer of the recognition model is used to perform feature conversion on the first image, including: Performing semantic segmentation on the first image to obtain a semantic matrix, the semantic matrix including a number of ship elements and sea surface elements; performing eight-neighborhood clustering on the ship elements to obtain a number of ship clusters; Taking the central element of each ship cluster as the center, the edge elements of the corresponding ship cluster are amplified according to a preset ratio to obtain an amplified cluster, and the elements in the amplified cluster are replaced by 0 to obtain a segmentation matrix; wherein the preset ratio is greater than 1; Detecting the edge of the segmentation matrix to obtain several characteristic edges of the segmentation matrix; Fill the edges of the positions of the augmented clusters in the segmentation matrix corresponding to the semantic matrix, specifically including: Cluster several feature edges according to their angles, and obtain the edge angle of the maximum cluster of feature edges; Obtain the density of feature edges, where the density of feature edges is the ratio of the number of elements of feature edges in the segmentation matrix to the number of elements of non-amplified clusters in the semantic matrix; For the position of the augmented cluster corresponding to the semantic matrix in the segmentation matrix, the edges are randomly filled according to the edge angle and density to obtain the first feature transformation map; wherein the edge angle of the filled edge is the same as the edge angle of the maximum cluster of feature edges, and the ratio of the number of elements of the filled edge to the number of elements of the augmented cluster in the semantic matrix is ​​equal to the density.

5. The intelligent monitoring combined fleet cargo status management system according to claim 4 is characterized in that: The image processing layer of the recognition model is also used to perform feature conversion on the second image and the third image, including: Perform edge detection on the second image or the third image to obtain a second feature conversion map or a third feature conversion map, respectively.

6. The intelligent monitoring combined fleet cargo status management system according to claim 5 is characterized in that: The hidden layer of the recognition layer is used to input the first feature conversion map, the second feature conversion map or the third feature conversion map. The calculation formula of the hidden layer is as follows: ； ； in, represents the wave breaking score output by the hidden layer for the first feature conversion map, the second feature conversion map, or the third feature conversion map, represents a first characteristic conversion diagram, a second characteristic conversion diagram, or a third characteristic conversion diagram, represents a left singular matrix, represents the singular value matrix, represents a right singular matrix, represents the weight vector of the hidden layer, represents the bias vector of the hidden layer, represents the bias matrix of the hidden layer, represents the weight matrix of the hidden layer, represents the matrix eigenvalue decomposition operation, represents the vector transpose operation, represents the vector concatenation operation, express Activation function, express Activation function.

7. The intelligent monitoring combined fleet cargo status management system according to claim 6 is characterized in that: The remaining individuals are used as parents for update iterations. The update formula is as follows: ； in, Indicates The update matrix of the parent individual at the iteration, Indicates The matrix of the parent individuals at the iteration, represents the updated weight, Indicates The matrix of the individual with the highest feature ranking at the iteration, Indicates size A sparse matrix whose diagonal elements are 1 and the rest are 0.

Images