Marine resource feature extraction method based on high-quality data set

Abstract

The application discloses a sea area resource feature extraction method based on a high-quality data set, relates to the technical field of sea area resource identification, and comprises the following steps: obtaining a high-quality data set of a sea area resource, which is named a resource data set, wherein the resource data set comprises a data set and an image set; performing feature extraction on the data set in the resource data set to obtain data features of the sea area resource; performing feature extraction on the image set in the resource data set to obtain image features of the sea area resource; and integrating the data features and the image features, preliminarily screening the sea area resource in the sea by the image features, and verifying the sea area resource according to the data features. The application is used for solving the problem that the existing sea area resource identification technology is not reasonable enough in identifying the sea area resource, and thus leads to excessive waste of human resources and material resources when searching for the sea area resource.

Description

Technical Field

[0001] This invention relates to the field of marine resource identification technology, specifically a method for extracting marine resource features based on high-quality datasets. Background Technology

[0002] Marine resource identification technology refers to the technology that uses methods such as space-based Earth observation, marine field detection, computer science, and artificial intelligence to discover, locate, classify, and quantify natural resources (such as fishery resources, oil and gas resources, and mineral resources) and spatial resources (such as ports, waterways, tourist areas, and aquaculture areas) in marine space.

[0003] Existing marine resource identification technologies typically require probing every area of ​​the ocean to discover marine biological resources. Although it is only necessary to detect nearshore biological resources, the nearshore area is very large, making comprehensive detection impossible. Therefore, selective detection is necessary, but existing marine resource identification technologies cannot perform selective detection. Furthermore, most existing marine resource identification technologies use preset threshold comparisons to identify marine resources. However, different factors within the ocean are interrelated; different values ​​of one parameter affect the applicable range of other parameters. Existing marine resource identification technologies also suffer from insufficient rationality in identifying marine resources, leading to excessive waste of human and material resources when searching for marine resources. Summary of the Invention

[0004] This invention aims to at least partially address one of the technical problems in the prior art. It obtains a high-quality dataset of marine resources, named a resource dataset, which includes a data set and an image set. Then, it statistically analyzes different monitoring data within the data set to obtain the individual ranges of the monitoring data. Features are then extracted from these individual ranges to obtain the data features of the marine resources. Simultaneously, features are extracted from regional images in the image set to obtain preliminary image features of the marine resources. These preliminary image features are then calibrated based on images of sea surfaces where no marine resources exist, resulting in image features. Finally, the data features and image features are integrated. Marine resources are initially screened within the ocean using the image features, and then verified based on the data features. This addresses the problem that existing marine resource identification technologies are not sufficiently accurate in identifying marine resources, leading to excessive waste of human and material resources when searching for them.

[0005] To achieve the above objectives, this application provides a method for extracting marine resource features based on high-quality datasets, comprising the following steps:

[0006] A high-quality dataset of marine resources is obtained and named the resource dataset, which includes a data set and an image set;

[0007] Feature extraction is performed on the dataset within the resource dataset to obtain data features of marine resources;

[0008] Feature extraction is performed on the image set within the resource dataset to obtain the image features of marine resources;

[0009] The data features and image features are integrated. Marine resources are initially screened based on image features, and then verified based on data features.

[0010] Furthermore, the marine resources are biological resources, and the resource dataset centrally records monitoring data, which includes seawater temperature, chlorophyll concentration, salinity, water quality, water depth, and regional images. The set of seawater temperature, chlorophyll concentration, salinity, water quality, and water depth constitutes the dataset, and the set of regional images constitutes the image set.

[0011] Furthermore, feature extraction is performed on the dataset within the resource dataset to obtain data features of marine resources, including the following sub-steps:

[0012] By statistically analyzing the different monitoring data within the dataset, the individual range of the monitoring data can be obtained.

[0013] Feature extraction is performed on individual areas to obtain data features of marine resources.

[0014] Furthermore, statistical analysis is performed on the different monitoring data within the dataset to obtain the individual range of the monitoring data, including the following sub-steps:

[0015] The area where biological resources are located is named a biological region. Each biological region has a resource dataset, and the resource datasets of different biological regions exist independently of each other.

[0016] Obtain the extent of each monitoring data point in the resource dataset, name it the "Biomonitoring Extent," and number the bioregions, labeling them as BY. n , where n is a positive integer and n is the sequence number of BY;

[0017] The biological monitoring range includes seawater temperature range, chlorophyll concentration range, salinity range, water quality range, and water depth range.

[0018] The range of seawater temperature, chlorophyll concentration, salinity, water quality, and water depth within the biological monitoring range constitutes the individual range.

[0019] BY n The range of individuals is denoted as BF(n,m), where m is a positive integer and (n,m) is the index of BF. Furthermore, BF(n,m) represents BY in ascending order of m. nThe range of seawater temperature, chlorophyll concentration, salinity, water quality, and water depth.

[0020] Furthermore, feature extraction is performed on individual areas to obtain data features of marine resources, including the following sub-steps:

[0021] Calculate the median of each BF(n,m) and label it as BG(n,m). Establish a two-dimensional coordinate system with m as the X-axis and BG(n,m) as the Y-axis, and name it Resource Feature Analysis Chart. Enter BG(n,m) into the Resource Feature Analysis Chart according to m.

[0022] The coordinate points in the resource feature analysis graph are named feature analysis points. Adjacent feature analysis points are connected by straight lines to obtain feature polylines. Each value of n corresponds to a feature polyline, and all feature polylines are located in the same resource feature analysis graph.

[0023] Obtain the feature analysis points corresponding to the maximum and minimum values ​​of the Y-axis for each value on the X-axis in the resource feature analysis graph, and name them the maximum feature point and the minimum feature point;

[0024] Connect the largest and smallest feature points with the same X-axis value by using straight lines to form a surface, which is called a data feature.

[0025] Furthermore, feature extraction is performed on the image set within the resource dataset to obtain image features of marine resources, including the following sub-steps:

[0026] Feature extraction is performed on regional images in the image set to obtain preliminary image features of marine resources;

[0027] The image features are obtained by calibrating the preliminary features of the sea surface image where no marine resources are available.

[0028] Furthermore, feature extraction is performed on the regional images in the image set to obtain preliminary image features of marine resources, including the following sub-steps:

[0029] Label the pixel in the i-th row and j-th column of the region image as PT(i,j), convert the region image to a grayscale image, and label the grayscale value at PT(i,j) as GV(i,j);

[0030] Obtain the water depth in the resource dataset to which the regional image belongs, and name it the regional depth. Establish a two-dimensional coordinate system with the regional depth as the horizontal axis and GV(i,j) as the vertical axis, and name it the resource image analysis map. Enter GV(i,j) into the resource image analysis map according to the regional depth, and name the coordinate points of the resource image analysis map the resource image analysis points.

[0031] Perform function regression on the resource image analysis map, and name the function obtained by function regression the depth gray level relationship function;

[0032] The residual between the resource image analysis points and the depth grayscale relationship function is obtained and named the relationship fluctuation value. The relationship fluctuation value is the preliminary feature of the image.

[0033] Furthermore, based on images of sea surfaces lacking marine resources, preliminary image features are calibrated to obtain image features, including the following sub-steps:

[0034] Randomly select an area on the sea surface that has no marine resources, name it the "no-resource area", obtain an image of the "no-resource area", and name it the "no-resource image".

[0035] Extract the GV(i,j) of the resource-free image and label it as W(i,j). At the same time, obtain the water depth of the resource-free region and name it the resource-free depth.

[0036] Enter W(i,j) into the resource image analysis map according to the resource-free depth, and name the coordinates formed by W(i,j) and the resource-free depth as the resource-free point;

[0037] Obtain the residual between the resource-free point and the depth grayscale relationship function, and name it the resource-free fluctuation value. Obtain the minimum value of the resource-free fluctuation value, and name it the resource-free fluctuation limit. Obtain the maximum value of the relationship fluctuation value, and name it the relationship fluctuation limit.

[0038] Determine whether the no-resource fluctuation limit is greater than or equal to the relationship fluctuation limit. If so, use the average of the no-resource fluctuation limit and the relationship fluctuation limit as the image feature. If not, execute the image feature calibration scheme.

[0039] Furthermore, the image feature calibration scheme includes the following sub-steps:

[0040] The fluctuation values ​​without resources and those with relationships are collectively referred to as analytical fluctuation values. These analytical fluctuation values ​​are then evenly divided into a first-order range, named the fluctuation range. The fluctuation ranges are then sorted and numbered in ascending order, using the symbol H. t This indicates that t is a positive integer and t is the index of H;

[0041] Statistical H t The number of resource fluctuation values ​​and the number of relationship fluctuation values ​​are respectively labeled as FA. t and FB t ;

[0042] With H t Create a histogram with the X-axis as the x-axis and the Y-axis as the y-axis, name it the Feature Calibration Map, and then use the FA... t and FB t According to H tEnter the feature calibration map, and then input the FA. t With H t The resulting histogram is named the No Resource Analysis Column, and FB is... t With H t The resulting histogram is named the relational analysis column;

[0043] By continuously increasing the first quantity by one, the number of fluctuation ranges increases, and eventually the histogram bars can approach a smooth curve. The smooth curve formed by the resource-free analysis bars is named the resource-free analysis curve, and the smooth curve formed by the relationship analysis bars is named the relationship analysis curve.

[0044] Obtain the intersection point of the resource-free analysis curve and the relationship analysis curve, name it the difference intersection point, and set the H at the difference intersection point. t Name it as an image feature.

[0045] Furthermore, the data features and image features are integrated. Marine resources are initially screened based on image features, and then verified based on data features, including the following sub-steps:

[0046] Obtain an ocean dataset, which records the water depth in different areas of the ocean and is named the reference depth. The area corresponding to the reference depth is named the reference area.

[0047] Substitute the reference depth into the depth-gray-level relationship function to obtain the reference gray level. Add the reference gray level to the image features to obtain the upper reference limit, and subtract the reference gray level from the image features to obtain the lower reference limit.

[0048] Pixels with gray values ​​between the lower and upper reference limits within the reference area are marked as suspected points, and the area formed by consecutive adjacent suspected points is named the suspected region.

[0049] Dispatch unmanned surface vessels to the suspected area to collect monitoring data, which is named reference data. Analyze the characteristic polyline of the reference data and name it reference polyline.

[0050] Substitute the reference line into the resource feature analysis chart. If the reference line is within the data features, then mark the reference area as containing biological resources.

[0051] The beneficial effects of this invention are as follows: This invention obtains a high-quality dataset of marine resources, named a resource dataset, which includes a data set and an image set. Then, it statistically analyzes different monitoring data within the data set to obtain the individual ranges of the monitoring data. Furthermore, it extracts features from the individual ranges to obtain the data features of marine resources. The advantage is that the data features reveal the range relationships between different monitoring data. Biological resources are all within the data features. If the monitoring data is obtained, it is possible to specifically determine whether biological resources exist in the marine area, thereby improving the accuracy and effectiveness of marine resource identification.

[0052] This invention extracts features from regional images in an image set to obtain preliminary image features of marine resources. These preliminary features are then calibrated based on images of sea surfaces where no marine resources exist, resulting in further image features. Finally, the data features and image features are integrated. Marine resources are initially screened using image features, and then verified based on data features. The advantage lies in the ability to perform preliminary screening of marine resources using image features, followed by detection of suspected areas after the initial screening. This selective detection, rather than blindly probing within the marine area, improves the accuracy and rationality of marine resource identification. Attached Figure Description

[0053] Figure 1 This is a flowchart of the steps of the method of the present invention;

[0054] Figure 2 This is a schematic diagram of the resource feature analysis diagram of the present invention;

[0055] Figure 3 This is a schematic diagram of the feature analysis points and feature polylines of the present invention;

[0056] Figure 4 This is a schematic diagram illustrating the data features of the present invention;

[0057] Figure 5 This is a schematic diagram of the resource image analysis diagram of the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Example 1, please refer to Figure 1 As shown, this application provides a method for extracting marine resource features based on high-quality datasets, including the following steps:

[0060] Step S1: Obtain a high-quality dataset of marine resources, named the resource dataset. The resource dataset includes a data set and an image set. Marine resources are biological resources. The resource dataset records monitoring data, including seawater temperature, chlorophyll concentration, salinity, water quality, water depth, and regional images. The set of seawater temperature, chlorophyll concentration, salinity, water quality, and water depth is the data set, and the set of regional images is the image set.

[0061] In specific implementation, this embodiment is applicable to finding areas rich in biological resources, that is, finding marine areas suitable for building marine ranches. When analyzing whether a marine area is suitable for building a marine ranch, it is necessary to refer to whether the monitoring data in the data set listed in this embodiment meets the requirements. At the same time, although the main factor affecting the color depth of the regional image is water depth, other monitoring data will slightly affect the color depth of the regional image. Therefore, marine resources can be effectively identified through the data set and image set. When monitoring other marine resources, such as mineral resources, monitoring data related to mineral resources are selected. The monitoring data listed in this embodiment is only applicable to finding marine ranches in this embodiment.

[0062] Step S2 involves extracting features from the dataset within the resource dataset to obtain data features of marine resources. Step S2 includes the following sub-steps:

[0063] Step S201: Statistically analyze the different monitoring data within the dataset to obtain the individual range of the monitoring data;

[0064] Step S201 includes the following sub-steps:

[0065] Step S201.1: The region where the biological resources are located is named a biological region. Each biological region has a resource dataset, and the resource datasets of different biological regions exist independently of each other.

[0066] In specific implementation, the biological regions in this embodiment are known marine ranches, each marine ranch is a biological region, and each biological region stores an independent resource dataset. During the analysis process, each resource dataset is analyzed independently.

[0067] Step S201.2: Obtain the range of each monitoring data item in the resource dataset, name it the biological monitoring range, and number the biological regions, marking them as BY. n , where n is a positive integer and n is the sequence number of BY;

[0068] Step S201.3, the biological monitoring range includes seawater temperature range, chlorophyll concentration range, salinity range, water quality range, and water depth range;

[0069] Step S201.4, the range of seawater temperature, chlorophyll concentration, salinity, water quality, and water depth in the biological monitoring range is the individual range;

[0070] Step S201.5, BY n The range of individuals is denoted as BF(n,m), where m is a positive integer and (n,m) is the index of BF. Furthermore, BF(n,m) represents BY in ascending order of m. n The range of seawater temperature, chlorophyll concentration, salinity, water quality, and water depth;

[0071] In specific implementation, taking salinity range as an example, for instance, the salinity in marine ranch BY1 is as low as 30ppt and as high as 33ppt, so the salinity range is [30ppt, 33ppt], and the individual range corresponding to the salinity is BF(n, 3), that is, BF(1, 3) is [30ppt, 33ppt]. The statistics of other biological monitoring ranges are the same as those of salinity ranges, and will not be described in detail in this embodiment.

[0072] Step S202: Extract features from the individual range to obtain the data features of marine resources;

[0073] Step S202 includes the following sub-steps:

[0074] Please see Figure 2 As shown, in step S202.1, the median of each BF(n,m) is calculated and labeled as BG(n,m). A two-dimensional coordinate system is established with m as the X-axis and BG(n,m) as the Y-axis, named the resource feature analysis chart. BG(n,m) is entered into the resource feature analysis chart according to m.

[0075] Please see Figure 3 As shown in step S202.2, the coordinate points in the resource feature analysis graph are named feature analysis points. Adjacent feature analysis points are connected by straight lines to obtain feature polylines. Each value of n corresponds to a feature polyline, and all feature polylines are in the same resource feature analysis graph.

[0076] Step S202.3: Obtain the feature analysis points corresponding to the maximum and minimum values ​​of the Y-axis for each value of the X-axis in the resource feature analysis graph, and name them as the maximum feature point and minimum feature point;

[0077] Please see Figure 4 As shown, in step S202.4, the maximum and minimum feature points with the same X-axis value are connected by a straight line to form a surface, which is named the data feature.

[0078] In specific implementation, taking BF(1,3) as an example, the calculated BG(1,3) is 31.5 ppt. Since different monitoring data have different dimensions, it is necessary to unify the dimensions of BG(n,m). A normalization algorithm can be directly used. In all resource datasets, the maximum salinity is 35 ppt, and the minimum is 29 ppt. Therefore, the actual BG(1,3) is (31.5-29) / (35-29) = 0.4167. The calculation result is rounded to four decimal places. Normalization is performed on each BG(n,m). Since all BG(n,m) are constrained between 0 and 1, the Y-axis of the constructed resource feature analysis map is actually between 0 and 1. Figure 2 As shown, Figure 3 This example demonstrates the feature analysis points and feature polylines formed after inputting BY1's BG(1,m) into the resource feature analysis graph. Different BY... n Different feature lines can be obtained, and then the maximum and minimum feature points on each value of the X-axis can be acquired. These maximum and minimum feature points represent the maximum and minimum acceptable ranges of relationships between different monitoring data of the marine ranch. Connecting the maximum and minimum feature points, consecutive adjacent lines form a surface, ultimately yielding the data features as shown below. Figure 4 As shown.

[0079] Step S3 involves extracting features from the image set within the resource dataset to obtain image features of the marine resources. Step S3 includes the following sub-steps:

[0080] Step S301: Extract features from the regional images in the image set to obtain preliminary image features of marine resources;

[0081] Step S301 includes the following sub-steps:

[0082] Step S301.1: Mark the pixel in the i-th row and j-th column of the region image as PT(i,j), convert the region image to a grayscale image, and mark the grayscale value at PT(i,j) as GV(i,j);

[0083] Please see Figure 5 As shown, in step S301.2, the water depth in the resource dataset to which the regional image belongs is obtained and named the regional depth. A two-dimensional coordinate system is established with the regional depth as the horizontal axis and GV(i,j) as the vertical axis and named the resource image analysis map. GV(i,j) is entered into the resource image analysis map according to the regional depth, and the coordinate points of the resource image analysis map are named resource image analysis points.

[0084] Step S301.3: Perform function regression on the resource image analysis map, and name the function obtained by function regression as the depth grayscale relationship function;

[0085] Step S301.4: Obtain the residual between the resource image analysis points and the depth grayscale relationship function, and name it the relationship fluctuation value. The relationship fluctuation value is the preliminary feature of the image.

[0086] In practice, the regional image is a satellite image of the marine ranch, and the resulting resource image analysis map is as follows. Figure 5 As shown, the depth-grayscale relationship function obtained from the analysis is DY=-19.69×ln(DX)+132.19, where DY is the grayscale value, DX is the region depth, and the residual can be directly extracted through the function regression model. In this embodiment, it will not be explained in detail. The residual is the relationship fluctuation value, which is the preliminary feature of the image.

[0087] Step S302: Based on the image of the sea surface where there are no marine resources, the preliminary features of the image are calibrated to obtain the image features;

[0088] Step S302 includes the following sub-steps:

[0089] Step S302.1: Randomly select an area on the sea surface where there are no marine resources, name it the "no-resource area", obtain an image of the "no-resource area", and name it the "no-resource image".

[0090] In practice, when selecting images without resources, the contours of the regional images in the image set are usually used to select areas in the sea that are known to be unsuitable for building marine ranches. Images within the selected area are then used to obtain images without resources. This step is to ensure that the number of analysis points without resources and resource images are consistent, so as to provide an equal amount of data as a reference for the execution of the image feature calibration scheme.

[0091] Step S302.2: Extract GV(i,j) of the resource-free image and label it as W(i,j). At the same time, obtain the water depth of the resource-free area and name it as the resource-free depth.

[0092] Step S302.3: Input W(i,j) into the resource image analysis map according to the resource-free depth, and name the coordinates formed by W(i,j) and the resource-free depth as the resource-free point;

[0093] Step S302.4: Obtain the residual of the relationship function between the resource-free point and the depth grayscale, and name it the resource-free fluctuation value; obtain the minimum value of the resource-free fluctuation value, and name it the resource-free fluctuation limit; obtain the maximum value of the relationship fluctuation value, and name it the relationship fluctuation limit.

[0094] In practice, although the main factor affecting the color depth of the regional image is water depth, other monitoring data will also slightly affect the color depth of the regional image. The depth grayscale relationship function is the relationship between grayscale values ​​and water depth within the marine ranch area. Generally, the residual of the depth grayscale relationship function between the resource-free points formed by areas unsuitable for constructing marine ranches will be greater than the relationship fluctuation limit, but there may also be an intersection. Therefore, further judgment and analysis are required. For example, in this embodiment, the relationship fluctuation limit is 4.25, while the resource-free fluctuation limit for resource-free points is 3.88, indicating that there is an intersection between the two. That is, there is a difficult-to-distinguish boundary between marine ranches and non-marine ranches, and an image feature calibration scheme needs to be implemented.

[0095] Step S302.5: Determine whether the no-resource fluctuation limit is greater than or equal to the relationship fluctuation limit. If yes, use the average of the no-resource fluctuation limit and the relationship fluctuation limit as the image feature. If no, execute the image feature calibration scheme.

[0096] Step S302.5 includes the following sub-steps:

[0097] Step S302.5.a: Collectively refer to the fluctuation values ​​without resources and the fluctuation values ​​of relationships as the analysis fluctuation values. Divide the analysis fluctuation values ​​evenly into a first-order range, named the fluctuation range, and sort and number the fluctuation ranges in ascending order, using the symbol H. t This indicates that t is a positive integer and t is the index of H;

[0098] Step S302.5.b, Statistical analysis of H t The number of resource fluctuation values ​​and the number of relationship fluctuation values ​​are respectively labeled as FA. t and FB t ;

[0099] Step S302.5.c, with H t Create a histogram with the X-axis as the x-axis and the Y-axis as the y-axis, name it the Feature Calibration Map, and then use the FA... t and FB t According to H t Enter the feature calibration map, and then input the FA. t With H t The resulting histogram is named the No Resource Analysis Column, and FB is... t With H t The resulting histogram is named the relational analysis column;

[0100] In step S302.5.d, the first quantity is continuously increased by one, so that the number of fluctuation ranges increases, and eventually the histogram bars can approach a smooth curve. The smooth curve formed by the no-resource analysis bars is named the no-resource analysis curve, and the smooth curve formed by the relationship analysis bars is named the relationship analysis curve.

[0101] Step S302.5.e: Obtain the intersection point of the resource-free analysis curve and the relationship analysis curve, name it the difference intersection point, and set the H at the difference intersection point. t Name it as an image feature;

[0102] In practice, there are no specific requirements for setting the first quantity; analysts can set it themselves. In this embodiment, the first quantity is set to 10. For example, if the maximum fluctuation value is 10 and the minimum is 0, then the range [0,10] is divided into 10 fluctuation ranges, namely H1 to H2. 10 At the same time, statistics are collected for each H. t The number of resource-free fluctuation values ​​and the number of relationship fluctuation values ​​are used to obtain FA. t and FB t By continuously increasing the initial quantity, the number of fluctuation ranges increases, the width of each histogram narrows, and the tops of the histograms become smoother. Through continuous increases, the tops of the histograms eventually approach a curve, thus obtaining the resource-free analysis curve and the relationship analysis curve. The intersection of their differences represents that when water depth is an image feature, there is a 50% probability that this is a non-marine ranch and a 50% probability that it is a marine ranch. When the residual is less than the image feature, it means that this is likely a marine ranch. Although there are small errors, it can significantly reduce the preliminary work in finding suitable marine ranches.

[0103] Step S4 involves integrating data features and image features, initially screening marine resources based on image features, and then verifying the marine resources based on data features. Step S4 includes the following sub-steps:

[0104] Step S401: Obtain the ocean dataset, which records the water depth in different areas of the ocean and is named the reference depth. The area corresponding to the reference depth is named the reference area.

[0105] Step S402: Substitute the reference depth into the depth-gray-level relationship function to obtain the reference gray level. Add the reference gray level to the image features to obtain the upper reference limit. Subtract the reference gray level from the image features to obtain the lower reference limit.

[0106] Step S403: Mark the pixels in the reference area whose gray values ​​are between the lower and upper reference limits as suspected points, and name the area formed by consecutive adjacent suspected points as a suspected area.

[0107] Step S404: Dispatch an unmanned surface vessel to the suspected area to collect monitoring data, name it as reference data, analyze the characteristic polyline of the reference data, and name it as reference polyline;

[0108] Step S405: Substitute the reference polyline into the resource feature analysis map. If the reference polyline is within the data features, mark the reference area as containing biological resources.

[0109] In specific implementation, assuming the image feature in this embodiment is 6.42, for example, the reference depth of a certain sea area is 200m, substituting it into the solution yields a reference grayscale of 27.87. The calculation result is rounded to two decimal places and then added to the image feature to obtain a reference upper limit of 23 and a reference lower limit of 21. The calculation result is rounded to the nearest integer, indicating that sea areas with grayscale values ​​between 21 and 23 are likely suitable for building marine ranches. At this time, unmanned vessels are dispatched to the suspected area to collect monitoring data and obtain reference data. If the reference line of the reference data is within the data feature, it means that the reference data conforms to the relationship between various monitoring data of marine ranches and is suitable for building marine ranches, i.e., there are biological resources. The opposite is also true.

[0110] Example 2: This application provides an electronic device, which may include a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other via the communication bus. The memory stores computer-readable instructions, and the processor can call these instructions. When the processor executes a computer-readable instruction, it performs steps such as those in the "Marine Resource Feature Extraction Method Based on High-Quality Dataset" to achieve the following functions: acquiring a high-quality dataset of marine resources, named a resource dataset, which includes a data set and an image set; performing feature extraction on the data set within the resource dataset to obtain data features of the marine resources; performing feature extraction on the image set within the resource dataset to obtain image features of the marine resources; integrating the data features and image features; initially screening marine resources in the ocean using image features, and then verifying the marine resources based on the data features.

[0111] Furthermore, when the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0112] Example 3: This application also provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the marine resource feature extraction method based on a high-quality dataset provided by the above methods. The method includes: acquiring a high-quality dataset of marine resources, named a resource dataset, which includes a data set and an image set; performing feature extraction on the data set within the resource dataset to obtain data features of marine resources; performing feature extraction on the image set within the resource dataset to obtain image features of marine resources; integrating the data features and image features, initially screening marine resources in the ocean using the image features, and then verifying the marine resources based on the data features.

[0113] Example 4: This application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it performs the steps of the above-described method for extracting marine resource features based on a high-quality dataset to achieve the following functions: obtaining a high-quality dataset of marine resources, named a resource dataset, which includes a data set and an image set; extracting features from the data set within the resource dataset to obtain data features of marine resources; extracting features from the image set within the resource dataset to obtain image features of marine resources; integrating the data features and image features, initially screening marine resources in the ocean using the image features, and then verifying the marine resources based on the data features.

[0114] Based on the above description of the embodiments, the embodiments of the present invention can be provided as methods, systems, or computer program products. Based on this understanding, the above technical solutions, in essence or in terms of their contribution to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or certain parts of the embodiments.

[0115] In the embodiments provided in this application, it should be understood that the disclosed system or method can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules or units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interfaces. The indirect coupling or communication connection between systems, modules, and units may be electrical, mechanical, or other forms.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for extracting marine resource features based on high-quality datasets, characterized in that, Includes the following steps: A high-quality dataset of marine resources is obtained and named the resource dataset, which includes a data set and an image set; Feature extraction is performed on the dataset within the resource dataset to obtain data features of marine resources; Feature extraction is performed on the image set within the resource dataset to obtain the image features of marine resources; The data features and image features are integrated. Marine resources are initially screened based on image features, and then verified based on data features. Extracting features from the image set within the resource dataset to obtain image features of marine resources includes the following sub-steps: Feature extraction is performed on regional images in the image set to obtain preliminary image features of marine resources; The preliminary features of the image are calibrated based on the image of the sea surface where no marine resources exist, and the image features are obtained. Extracting features from regional images in an image set to obtain preliminary image features of marine resources includes the following sub-steps: Label the pixel in the i-th row and j-th column of the region image as PT(i,j), convert the region image to a grayscale image, and label the grayscale value at PT(i,j) as GV(i,j); Obtain the water depth in the resource dataset to which the regional image belongs, and name it the regional depth. Establish a two-dimensional coordinate system with the regional depth as the horizontal axis and GV(i,j) as the vertical axis, and name it the resource image analysis map. Enter GV(i,j) into the resource image analysis map according to the regional depth, and name the coordinate points of the resource image analysis map the resource image analysis points. Perform function regression on the resource image analysis map, and name the function obtained by function regression the depth gray level relationship function; Obtain the residual between the resource image analysis points and the depth grayscale relationship function, and name it the relationship fluctuation value. The relationship fluctuation value is the preliminary feature of the image. The initial feature calibration of images of sea surfaces lacking marine resources involves the following sub-steps: Randomly select an area on the sea surface that has no marine resources, name it the "no-resource area", obtain an image of the "no-resource area", and name it the "no-resource image". Extract the GV(i,j) of the resource-free image and label it as W(i,j). At the same time, obtain the water depth of the resource-free region and name it the resource-free depth. Enter W(i,j) into the resource image analysis map according to the resource-free depth, and name the coordinates formed by W(i,j) and the resource-free depth as the resource-free point; Obtain the residual between the resource-free point and the depth grayscale relationship function, and name it the resource-free fluctuation value. Obtain the minimum value of the resource-free fluctuation value, and name it the resource-free fluctuation limit. Obtain the maximum value of the relationship fluctuation value, and name it the relationship fluctuation limit. Determine whether the no-resource fluctuation limit is greater than or equal to the relationship fluctuation limit. If so, use the average of the no-resource fluctuation limit and the relationship fluctuation limit as the image feature. If not, execute the image feature calibration scheme.

2. The method for extracting marine resource features based on high-quality datasets according to claim 1, characterized in that, The marine resources mentioned are biological resources. The resource dataset contains monitoring data, including seawater temperature, chlorophyll concentration, salinity, water quality, water depth, and regional images. The set of seawater temperature, chlorophyll concentration, salinity, water quality, and water depth constitutes the dataset, and the set of regional images constitutes the image set.

3. The method for extracting marine resource features based on high-quality datasets according to claim 2, characterized in that, Extracting features from the dataset within the resource dataset to obtain data features of marine resources includes the following sub-steps: By statistically analyzing the different monitoring data within the dataset, the individual range of the monitoring data can be obtained. Feature extraction is performed on individual areas to obtain data features of marine resources.

4. The method for extracting marine resource features based on high-quality datasets according to claim 3, characterized in that, To statistically analyze the different monitoring data within the dataset and obtain the individual range of the monitoring data, the following sub-steps are involved: The area where biological resources are located is named a biological region. Each biological region has a resource dataset, and the resource datasets of different biological regions exist independently of each other. Obtain the extent of each monitoring data point in the resource dataset, name it the "Biomonitoring Extent," and number the bioregions, labeling them as BY. n , where n is a positive integer and n is the sequence number of BY; The biological monitoring range includes seawater temperature range, chlorophyll concentration range, salinity range, water quality range, and water depth range. The range of seawater temperature, chlorophyll concentration, salinity, water quality, and water depth within the biological monitoring range constitutes the individual range. BY n The range of individuals is denoted as BF(n,m), where m is a positive integer and (n,m) is the index of BF. Furthermore, BF(n,m) represents BY in ascending order of m. n The range of seawater temperature, chlorophyll concentration, salinity, water quality, and water depth.

5. The method for extracting marine resource features based on high-quality datasets according to claim 4, characterized in that, Extracting features from individual areas to obtain data features of marine resources includes the following sub-steps: Calculate the median of each BF(n,m) and label it as BG(n,m). Establish a two-dimensional coordinate system with m as the X-axis and BG(n,m) as the Y-axis, and name it Resource Feature Analysis Chart. Enter BG(n,m) into the Resource Feature Analysis Chart according to m. The coordinate points in the resource feature analysis graph are named feature analysis points. Adjacent feature analysis points are connected by straight lines to obtain feature polylines. Each value of n corresponds to a feature polyline, and all feature polylines are located in the same resource feature analysis graph. Obtain the feature analysis points corresponding to the maximum and minimum values ​​of the Y-axis for each value on the X-axis in the resource feature analysis graph, and name them the maximum feature point and the minimum feature point; Connect the largest and smallest feature points with the same X-axis value by using straight lines to form a surface, which is called a data feature.

6. The method for extracting marine resource features based on high-quality datasets according to claim 5, characterized in that, The image feature calibration scheme includes the following sub-steps: The fluctuation values ​​without resources and those with relationships are collectively referred to as analytical fluctuation values. These analytical fluctuation values ​​are then evenly divided into a first-order range, named the fluctuation range. The fluctuation ranges are then sorted and numbered in ascending order, using the symbol H. t This indicates that t is a positive integer and t is the index of H; Statistical H t The number of resource fluctuation values ​​and the number of relationship fluctuation values ​​are respectively labeled as FA. t and FB t ; With H t Create a histogram with the X-axis as the x-axis and the Y-axis as the y-axis, name it the Feature Calibration Map, and then use the FA... t and FB t According to H t Enter the feature calibration map, and then input the FA. t With H t The resulting histogram is named the No Resource Analysis Column, and FB is... t With H t The resulting histogram is named the relational analysis column; By continuously increasing the first quantity by one, the number of fluctuation ranges increases, and eventually the histogram bars can approach a smooth curve. The smooth curve formed by the resource-free analysis bars is named the resource-free analysis curve, and the smooth curve formed by the relationship analysis bars is named the relationship analysis curve. Obtain the intersection point of the resource-free analysis curve and the relationship analysis curve, name it the difference intersection point, and set the H at the difference intersection point. t Name it as an image feature.

7. The method for extracting marine resource features based on high-quality datasets according to claim 6, characterized in that, The process of integrating data features and image features, initially screening marine resources based on image features, and then verifying these resources based on data features includes the following sub-steps: Obtain an ocean dataset, which records the water depth in different areas of the ocean and is named the reference depth. The area corresponding to the reference depth is named the reference area. Substitute the reference depth into the depth-gray-level relationship function to obtain the reference gray level. Add the reference gray level to the image features to obtain the upper reference limit, and subtract the reference gray level from the image features to obtain the lower reference limit. Pixels with gray values ​​between the lower and upper reference limits within the reference area are marked as suspected points, and the area formed by consecutive adjacent suspected points is named the suspected region. Dispatch unmanned surface vessels to the suspected area to collect monitoring data, which is named reference data. Analyze the characteristic polyline of the reference data and name it reference polyline. Substitute the reference line into the resource feature analysis chart. If the reference line is within the data features, then mark the reference area as containing biological resources.

Images