Marine biological species prediction method, device, equipment, and storage medium

Abstract

The present invention belongs to the technical field of image recognition, and discloses a method and device, equipment, and storage medium for predicting marine biological species. Feature extraction is performed on a received cave image to be tested to obtain target feature values, which include size feature values, shape feature values, cave feature values, humidity feature values, and biological trace feature values. According to the target feature values ​​and feature weight coefficients, a predicted comprehensive score of the cave image to be tested is calculated, and the target marine biological species are predicted. Compared with the manual visual inspection method, the present invention can improve the efficiency and accuracy of marine biological species prediction by extracting multiple different feature values, thereby reducing blind sea hunting and enhancing the user's sea hunting fun.

Description

Technical Field

[0001] The present invention belongs to the field of image recognition technology, and specifically relates to a method and device, equipment, and storage medium for predicting marine biological species. Background Art

[0002] Nowadays, more and more people like to go to the beach to dig various marine life such as crabs and shellfish. In fact, taking crabs as an example, many crabs living in the intertidal zone are burrowing. They will build their own caves to hide and reproduce here. Different types of crabs, such as horn-eyed sand crabs, white southern fiddlers, saw-eyed mud crabs, and pleasing big-eyed crabs, also have their own different types of caves.

[0003] However, most people have little experience and are often confused when they encounter caves on the beach. They are not sure what kind of marine creatures the caves are. Often, because they cannot predict what kind of marine creatures they are, they use the wrong excavation methods and miss a lot of resources, resulting in no gains and no experience of the fun of sea hunting.

[0004] A small number of people use manual visual observation to observe the types of caves on the beach, predict the types of marine life in the caves, and then dig them. They have a rich harvest and a strong sense of sea-hunting experience. However, this manual visual prediction method of marine life species still has the problems of low efficiency and low accuracy. Summary of the invention

[0005] The purpose of the present invention is to provide a method and device, equipment, and storage medium for predicting marine biological species, which can improve the efficiency and accuracy of marine biological species prediction, thereby reducing blind sea-hunting and enhancing users' sea-hunting fun.

[0006] The first aspect of the present invention discloses a method for predicting marine biological species, comprising:

[0007] Extracting features from the received cave image to be tested to obtain target feature values, wherein the target feature values ​​include size feature values, shape feature values, cave feature values, humidity feature values, and biological trace feature values;

[0008] Calculate the predicted comprehensive score of the cave map to be tested according to the target feature value and the feature weight coefficient; wherein the feature weight coefficient is obtained by training a convolutional neural network;

[0009] According to the predicted comprehensive score, the corresponding target marine biological species are predicted.

[0010] In some embodiments, it also includes:

[0011] Collect multiple cave sample maps and label each cave sample map with the marine life species label;

[0012] Performing feature extraction on each of the cave sample images to obtain sample feature values, wherein the sample feature values ​​include size feature values, shape feature values, cave feature values, humidity feature values, and biological trace feature values;

[0013] The sample feature values ​​of each cave sample image and their corresponding marine life species labels are input into a convolutional neural network for training and cross-validation to obtain feature weight coefficients.

[0014] In some embodiments, according to the predicted comprehensive score, the corresponding target marine biological species are predicted, including:

[0015] Calculate the sample comprehensive score of each cave sample map according to the sample characteristic value and the characteristic weight coefficient;

[0016] Averaging the sample comprehensive scores of the cave sample images marked with the same marine biological species label to obtain a species comprehensive score for each marine biological species;

[0017] Obtaining the comprehensive score of the category closest to the predicted comprehensive score as the comprehensive score of the target category;

[0018] The marine biological species corresponding to the target species comprehensive score is determined as the target marine biological species.

[0019] In some embodiments, the predicted comprehensive score of the cave map to be tested is calculated according to the target feature value and the feature weight coefficient, including:

[0020] The predicted composite score is calculated using the following formula:

[0021]

[0022] Where W represents the predicted comprehensive score; f i is the i-th target feature value, n is the number of features, n = 5; w i is the weight coefficient corresponding to the i-th feature; α is the preset adjustment coefficient; g(f i ,w i )=sin(f i *w i ).

[0023] In some embodiments, feature extraction is performed on the received cave image to be tested, including:

[0024] The received cave image to be tested is converted into a grayscale image and image smoothing is performed to obtain a smoothed image;

[0025] detecting cave contours and reference object contours in the smoothed image;

[0026] The size characteristic value of the cave is calculated according to the cave outline, the reference object outline and the preset size of the reference object.

[0027] In some embodiments, extracting features from the received cave image to be tested further includes:

[0028] According to the cave outline, identifying the cave shape;

[0029] The cave shape is mapped to shape feature values.

[0030] In some embodiments, extracting features from the received cave image to be tested further includes:

[0031] Convert the received cave image to be tested into a color histogram;

[0032] Judging whether there is water or sand in the cave according to the color histogram, to obtain a first judgment result;

[0033] The first judgment result is mapped into a characteristic value inside the hole.

[0034] A second aspect of the present invention discloses a prediction device for marine biological species, comprising:

[0035] An extraction unit is used to extract features from the received cave image to be tested, and obtain target feature values, wherein the target feature values ​​include size feature values, shape feature values, cave feature values, humidity feature values, and biological trace feature values;

[0036] A calculation unit, used to calculate the predicted comprehensive score of the cave map to be tested according to the target feature value and the feature weight coefficient; wherein the feature weight coefficient is obtained by training a convolutional neural network;

[0037] The prediction unit is used to predict the corresponding target marine biological species according to the prediction comprehensive score.

[0038] The third aspect of the present invention discloses an electronic device, comprising a memory storing executable program code and a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the marine biological species prediction method disclosed in the first aspect.

[0039] A fourth aspect of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the method for predicting marine biological species disclosed in the first aspect.

[0040] The beneficial effect of the present invention lies in that, by performing feature extraction on the received cave map to be tested, target feature values ​​are obtained, and the target feature values ​​include size feature values, shape feature values, cave feature values, humidity feature values ​​and biological trace feature values; according to the target feature values ​​and feature weight coefficients, the predicted comprehensive score of the cave map to be tested is calculated, and the corresponding target marine life species are predicted. Therefore, compared with the manual visual inspection method, the present invention can improve the efficiency and accuracy of marine life species prediction by extracting multiple different feature values ​​to predict marine life species, thereby reducing blind sea hunting and enhancing users' sea hunting fun. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings herein show specific examples of the technical solutions described in the present invention, and together with the specific implementation methods constitute a part of the specification, and are used to explain the technical solutions, principles and effects of the present invention.

[0042] Unless otherwise specified or defined, the same reference numerals in different drawings represent the same or similar technical features, and the same or similar technical features may also be represented by different reference numerals.

[0043] Figure 1 is a flow chart of a method for predicting marine biological species disclosed in an embodiment of the present invention;

[0044] Figure 2 It is a schematic diagram of the structure of a device for predicting marine biological species disclosed in an embodiment of the present invention;

[0045] Figure 3 It is a structural schematic diagram of an electronic device disclosed in an embodiment of the present invention.

[0046] Description of reference numerals:

[0047] 201, first extraction unit; 202, calculation unit; 203, prediction unit; 301, memory; 302, processor. DETAILED DESCRIPTION

[0048] Unless otherwise specified or defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art. In the case of combining the technical solution of the present invention with realistic scenarios, all technical and scientific terms used herein may also have meanings corresponding to the purpose of implementing the technical solution of the present invention. The "first, second..." used herein is only used to distinguish the names and does not represent a specific quantity or order. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0049] It should be noted that when a component is considered to be "fixed to" another component, it can be directly fixed to the other component or there can be a central component; when an component is considered to be "connected to" another component, it can be directly connected to the other component or there can be a central component at the same time; when an component is considered to be "installed on" another component, it can be directly installed on the other component or there can be a central component at the same time. When an component is considered to be "set on" another component, it can be directly set on the other component or there can be a central component at the same time.

[0050] Unless otherwise specified or defined, the "said" and "the" used in this document refer to the technical features or technical contents mentioned or described before the corresponding position, and the technical features or technical contents may be the same as or similar to the technical features or technical contents mentioned therein. In addition, the terms "including" and "having" and any variations thereof used in this document are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0051] The embodiment of the present invention discloses a method for predicting marine biological species, which can be implemented by computer programming. The execution subject of the method can be an electronic device such as a computer, a laptop, a tablet computer, or a prediction device for marine biological species embedded in an electronic device, and the present invention is not limited to this. In order to facilitate the understanding of the present invention, the specific embodiments of the present invention will be described in more detail below with reference to the drawings of the specification.

[0052] like Figure 1 As shown, the method includes the following steps 110 to 130:

[0053] 110. Perform feature extraction on the received cave test image to obtain target feature values, which include size feature values, shape feature values, cave feature values, humidity feature values ​​and biological trace feature values.

[0054] Specifically, feature extraction is performed on each cave sample map, including the following steps 1101 to 1108:

[0055] 1101. Perform image smoothing processing on the received cave image to be tested to obtain a smoothed image.

[0056] 1102. Detect cave contours and reference object contours in the smoothed image.

[0057] Among them, the Canny edge detection algorithm can be used to detect the smooth image to identify the cave outline and the reference object outline.

[0058] 1103. Calculate the size characteristic value of the cave according to the cave outline, the reference object outline and the preset size of the reference object.

[0059] When collecting the cave map to be tested, place a reference object (such as a 1-yuan coin) next to each cave and take a photo. You can use the OpenCV image processing library to calculate the cave size characteristic value (with the 1-yuan coin as the reference, the preset size of the reference object is the size of the 1-yuan coin). By calculating the relative size relationship between the cave and the reference object, and based on the preset size of the reference object, the actual size of the cave can be calculated, and then the size characteristic value of the cave can be calculated based on the actual size. For example, the long side of the minimum circumscribed rectangle of the cave is taken as the actual size of the cave, and mapped to the size characteristic value f1 through the following formula (1):

[0060]

[0061] Among them, max() is a function that takes a larger value.

[0062] 1104. Identify the cave shape according to the cave outline, and map the cave shape into shape feature values.

[0063] Specifically, the cave shape includes circular, elliptical or other shapes, which are mapped into the shape feature value f2 by the following formula (2):

[0064]

[0065] 1105. Convert the cave image to be tested into a color histogram.

[0066] 1106. Determine whether there is water or sand in the cave according to the color histogram, obtain a first determination result, and map the first determination result to a characteristic value in the cave.

[0067] Among them, the color histogram of the three channels of RGB (red, green and blue) is obtained for analysis. If the pixel value of the blue channel is higher than the pixel values ​​of the red and green channels, it can be inferred that the cave in the picture is mainly water; if the pixel values ​​of the red and green channels are both higher than the pixel value of the blue channel, it can be inferred that the cave in the picture is mainly sand, thereby obtaining a first judgment result. That is, the first judgment result includes that there is water in the cave, or there is sand in the cave.

[0068] When the first judgment result is that there is water in the cave, the characteristic value of the cave obtained by mapping is determined to be 1; when the first judgment result is that there is sand in the cave, the characteristic value of the cave obtained by mapping is determined to be 0.5.

[0069] Specifically, the first judgment result is mapped into the hole characteristic value f3 by the following formula (3):

[0070]

[0071] 1107. Perform texture analysis on the cave image to be tested, obtain the humidity level of the area surrounding the cave, and map the humidity level into a humidity level characteristic value.

[0072] Among them, the Gray Level Co-occurrence Matrix (GLCM) can be used to extract image texture features, and then texture analysis can be performed on the image texture features to obtain the humidity level of the area around the cave. The humidity level includes humid, semi-humid or dry.

[0073] When the humidity is wet, the humidity characteristic value obtained by mapping is determined to be 1; when the humidity is semi-humid, the humidity characteristic value obtained by mapping is determined to be 0.7; when the humidity is dry, the humidity characteristic value obtained by mapping is determined to be 0.2. The humidity is mapped to the humidity characteristic value f4 as shown in the following formula (4):

[0074]

[0075] 1108. Determine whether there are biological traces around the cave according to the cave test map, obtain a second judgment result, and map the second judgment result into a biological trace feature value.

[0076] Among them, the second judgment result includes the presence of biological traces or the absence of biological traces; when the second judgment result is the presence of biological traces, the biological trace characteristic value obtained by mapping is determined to be 1; when the second judgment result is the absence of biological traces, the biological trace characteristic value obtained by mapping is determined to be 0.

[0077] Specifically, the second judgment result is mapped into the biological trace feature value f5 by the following formula (5):

[0078]

[0079] 120. Calculate the predicted comprehensive score of the cave map to be tested based on the target feature value and the feature weight coefficient; wherein the feature weight coefficient is obtained by training a convolutional neural network.

[0080] Specifically, the training process of the feature weight coefficient may include the following steps 101 to 103 (not shown):

[0081] 101. Collect multiple cave sample images and label the marine life species of each cave sample image.

[0082] In an embodiment of the present invention, several cave photos can be collected from the beach as cave sample images. During the collection process, a reference object (such as a 1-yuan coin) is placed next to each cave to take photos, and the specific biological species in each cave are marked as marine biological species labels corresponding to the cave sample images.

[0083] 102. Feature extraction is performed on each cave sample image to obtain sample feature values, which include size feature values, shape feature values, cave feature values, humidity feature values ​​and biological trace feature values.

[0084] Specifically, feature extraction is performed on each cave sample map, including the following steps 1021 to 1028:

[0085] 1021. Convert each cave sample image into a grayscale image and perform image smoothing to obtain a smoothed image.

[0086] 1022. Detect cave contours and reference object contours in the smoothed image.

[0087] 1023. Calculate the size characteristic value of the cave according to the cave contour, the reference object contour and the preset size of the reference object.

[0088] 1024. Identify the cave shape according to the cave outline, and map the cave shape into shape feature values.

[0089] 1025. Convert the cave sample image into a color histogram.

[0090] 1026. Determine whether there is water or sand in the cave according to the color histogram, obtain a first determination result, and map the first determination result to a characteristic value in the cave.

[0091] 1027. Perform texture analysis on the cave sample image to obtain the humidity level of the area surrounding the cave, and map the humidity level into a humidity level feature value.

[0092] 1028. Determine whether there are biological traces around the cave according to the cave sample map, obtain a second judgment result, and map the second judgment result into a biological trace feature value.

[0093] It should be noted that the feature extraction of the cave sample image is similar to the extraction of the cave sample image in the above steps 1101 to 1108, and will not be repeated here. The sample feature values ​​in the extraction process also include at least five feature values: size feature value, shape feature value, cave feature value, humidity feature value, and biological trace feature value. After the above standardization of these five feature values, it is easy to train the machine learning model, and after the training is completed, the unknown image can be predicted.

[0094] 103. The sample feature values ​​of each cave sample image and their corresponding marine biological species labels are input into the convolutional neural network for training and cross-validation to obtain the feature weight coefficient.

[0095] The feature weight coefficient includes the weight coefficient corresponding to each feature, including size feature, shape feature, cave feature, humidity feature and biological trace feature.

[0096] Specifically, in step 120, the predicted comprehensive score can be calculated by the following formula:

[0097]

[0098] Where W is the comprehensive score (i.e., predicted comprehensive score or sample comprehensive score), which is used to determine the biological category; f i is the i-th eigenvalue, n is the number of features, n = 5; w i is the weight coefficient corresponding to the i-th feature. α is a preset adjustment coefficient used to adjust the second term g(f i ,w i )'s weight on the total score. i ,w i )=sin(f i *w i ) is used to further adjust the contribution of each feature to the total score. Other functions can also be selected according to actual conditions.

[0099] 130. According to the predicted comprehensive score, the corresponding target marine biological species are predicted.

[0100] Step 130 may include the following steps 1301 to 1303, which are not shown in the figure:

[0101] 1301. Calculate the comprehensive sample score of each cave sample map based on the sample characteristic value and characteristic weight coefficient.

[0102] The above formula (6) can be used to calculate the sample comprehensive score; or the following formula (7) can be used to calculate the sample comprehensive score:

[0103] W=w1*f1+w2*f2+w3*f3+w4*f4+ w5*f5 (7)

[0104] Among them, W is the comprehensive score, f1, f2, f3, f4, and f5 represent the characteristic values ​​of size, shape, cave, humidity, and biological traces, respectively; w1, w2, w3, w4, and w5 represent the corresponding characteristic weight coefficients, respectively.

[0105] 1302. The sample comprehensive scores of the cave sample images marked with the same marine biological species label are averaged to obtain the species comprehensive score of each marine biological species.

[0106] Some of the multiple cave sample maps may have the same marine life species label. Therefore, for the cave sample maps marked with the same marine life species label, after calculating the sample comprehensive score of each cave sample map, it is necessary to calculate the average score as the species comprehensive score of the marine life species.

[0107] 1303. Obtain the species comprehensive score closest to the predicted comprehensive score as the target species comprehensive score, and determine the marine biological species corresponding to the target species comprehensive score as the target marine biological species.

[0108] For the cave map to be tested, its predicted comprehensive score is compared with the comprehensive scores of the marine organism species corresponding to multiple marine organism species labels. If the predicted comprehensive score is closest to the comprehensive score of a certain marine organism species, then the marine organism species of the cave to be tested is similar to that marine organism species.

[0109] In summary, by implementing the embodiments of the present invention, the system can automatically extract feature values ​​based on the photos of caves on the beach submitted by users, calculate feature weights based on the feature values ​​and annotations of each photo, and finally predict marine life based on the comprehensive score formula for biological species, thereby reducing blind sea-hunting and enhancing users' fun in sea-hunting.

[0110] For a variety of marine life, the accuracy of predicting the species of marine life based on caves on the beach can be improved by increasing data collection and labeling and adjusting the weight of feature data.

[0111] like Figure 2 As shown, the embodiment of the present invention discloses a prediction device for marine biological species, including a first extraction unit 201, a calculation unit 202 and a prediction unit 203, wherein:

[0112] The first extraction unit 201 is used to extract features from the received cave image to obtain target feature values, where the target feature values ​​include size feature values, shape feature values, cave feature values, humidity feature values, and biological trace feature values;

[0113] A calculation unit 202 is used to calculate the predicted comprehensive score of the cave map to be tested according to the target feature value and the feature weight coefficient; wherein the feature weight coefficient is obtained by training a convolutional neural network;

[0114] The prediction unit 203 is used to predict the corresponding target marine biological species according to the prediction comprehensive score.

[0115] As an optional implementation, the prediction device may further include the following units not shown in the figure:

[0116] A collection unit for collecting multiple cave sample maps;

[0117] Annotation unit, used to label the marine life species of each cave sample map;

[0118] The second extraction unit is used to extract features from each cave sample image to obtain sample feature values, which include size feature values, shape feature values, cave feature values, humidity feature values, and biological trace feature values;

[0119] The training unit is used to input the sample feature values ​​of each cave sample image and its corresponding marine organism species labels into the convolutional neural network for training and cross-validation to obtain feature weight coefficients.

[0120] As an optional implementation, the prediction unit 203 includes the following sub-units (not shown):

[0121] A calculation subunit is used to calculate the sample comprehensive score of each cave sample map according to the sample characteristic value and the characteristic weight coefficient;

[0122] The averaging subunit is used to average the sample comprehensive scores of the cave sample images labeled with the same marine biological species label to obtain the species comprehensive score of each marine biological species;

[0123] An acquisition subunit is used to acquire the comprehensive score of the category closest to the predicted comprehensive score as the comprehensive score of the target category;

[0124] A subunit is determined to determine the marine biological species corresponding to the comprehensive score of the target species as the target marine biological species.

[0125] Optionally, the first extraction unit 201 includes a first mapping subunit not shown in the figure, which is used to convert the received cave image to be tested into a grayscale image for image smoothing to obtain a smoothed image; detect the cave contour and the reference object contour in the smoothed image; and calculate the size characteristic value of the cave based on the cave contour, the reference object contour and the preset size of the reference object.

[0126] Furthermore, the first extraction unit 201 also includes a second mapping subunit (not shown) for identifying the cave shape according to the cave outline; and mapping the cave shape into a shape feature value.

[0127] Furthermore, the first extraction unit 201 also includes a third mapping subunit not shown in the figure, which is used to convert the received cave test image into a color histogram; determine whether there is water or sand in the cave according to the color histogram to obtain a first judgment result; and map the first judgment result into a feature value in the cave.

[0128] Furthermore, the first extraction unit 201 also includes a fourth mapping subunit (not shown), which is used to perform texture analysis on the received cave image to be tested, obtain the humidity level of the area surrounding the cave, and map the humidity level into a humidity level feature value.

[0129] Furthermore, the first extraction unit 201 also includes a fifth mapping subunit not shown in the figure, which is used to determine whether there are biological traces around the cave based on the received cave test image to obtain a second judgment result; and map the second judgment result into a biological trace feature value.

[0130] like Figure 3 As shown, an embodiment of the present invention discloses an electronic device, including a memory 301 storing executable program codes and a processor 302 coupled to the memory 301;

[0131] The processor 302 calls the executable program code stored in the memory 301 to execute the marine biological species prediction method described in the above embodiments.

[0132] An embodiment of the present invention further discloses a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the method for predicting marine biological species described in the above embodiments.

[0133] The purpose of the above embodiments is to exemplarily reproduce and deduce the technical solution of the present invention, and to fully describe the technical solution, purpose and effect of the present invention. Its purpose is to make the public understand the disclosed content of the present invention more thoroughly and comprehensively, and it does not limit the scope of protection of the present invention.

[0134] The above embodiments are not exhaustive enumerations of the present invention, and there may be multiple other implementations not listed. Any replacement and improvement made without violating the concept of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for predicting marine biological species, characterized in that: include: Extracting features from the received cave image to be tested to obtain target feature values, wherein the target feature values ​​include size feature values, shape feature values, cave feature values, humidity feature values, and biological trace feature values; Calculate the predicted comprehensive score of the cave map to be tested according to the target feature value and the feature weight coefficient; wherein the feature weight coefficient is obtained by training a convolutional neural network; According to the prediction comprehensive score, the corresponding target marine biological species are predicted; The method further comprises: Collect multiple cave sample maps and label each cave sample map with the marine life species label; Performing feature extraction on each of the cave sample images to obtain sample feature values, wherein the sample feature values ​​include size feature values, shape feature values, cave feature values, humidity feature values, and biological trace feature values; Inputting the sample feature value of each cave sample image and its corresponding marine biological species label into a convolutional neural network for training and cross-validation to obtain a feature weight coefficient; According to the predicted comprehensive score, the corresponding target marine species are predicted, including: Obtaining the comprehensive score of the category closest to the predicted comprehensive score as the comprehensive score of the target category; Determine the marine biological species corresponding to the comprehensive score of the target species as the target marine biological species; According to the target characteristic value and the characteristic weight coefficient, the predicted comprehensive score of the cave map to be tested is calculated, including: The predicted composite score is calculated using the following formula: Where W represents the predicted comprehensive score; f i is the i-th target feature value, n is the number of features, n = 5; w i is the weight coefficient corresponding to the i-th feature; α is the preset adjustment coefficient; g(f i ,w i )=sin(f i *w i ); Feature extraction is performed on the received cave test image, including: The received cave image to be tested is converted into a grayscale image and image smoothing is performed to obtain a smoothed image; detecting cave contours and reference object contours in the smoothed image; Calculating a size characteristic value of the cave according to the cave outline, the reference object outline and the preset size of the reference object; Feature extraction of the received cave test image also includes: Convert the received cave image to be tested into a color histogram; Judging whether there is water or sand in the cave according to the color histogram, to obtain a first judgment result; The first judgment result is mapped into a characteristic value inside the hole.

2. The method for predicting marine biological species according to claim 1, characterized in that: Feature extraction of the received cave test image also includes: According to the cave outline, identifying the cave shape; The cave shape is mapped to shape feature values.

3. A prediction device for marine biological species, characterized in that: include: An extraction unit is used to extract features from a received cave image to be tested, and obtain target feature values, wherein the target feature values ​​include size feature values, shape feature values, cave feature values, humidity feature values, and biological trace feature values; the extraction unit is specifically used to convert the received cave image to be tested into a grayscale image for image smoothing to obtain a smoothed image; detect the cave contour and reference object contour in the smoothed image; calculate the size feature value of the cave according to the cave contour, the reference object contour, and the preset size of the reference object; and convert the received cave image to be tested into a color histogram; Determine whether there is water or sand in the cave according to the color histogram to obtain a first determination result; map the first determination result into a characteristic value in the cave; A calculation unit is used to calculate the predicted comprehensive score of the cave map to be tested according to the target feature value and the feature weight coefficient; wherein the feature weight coefficient is obtained by training a convolutional neural network; the calculation unit is specifically used to calculate the predicted comprehensive score by the following formula: Where W represents the predicted comprehensive score; f i is the i-th target feature value, n is the number of features, n = 5; w i is the weight coefficient corresponding to the i-th feature; α is the preset adjustment coefficient; g(f i ,w i )=sin(f i *w i ); A prediction unit is used to predict the corresponding target marine biological species according to the predicted comprehensive score; the prediction unit is specifically used to obtain the species comprehensive score closest to the predicted comprehensive score as the target species comprehensive score; determine the marine biological species corresponding to the target species comprehensive score as the target marine biological species; The prediction device also includes: A collection unit for collecting multiple cave sample maps; Annotation unit, used to label the marine life species of each cave sample map; The second extraction unit is used to extract features from each cave sample image to obtain sample feature values, which include size feature values, shape feature values, cave feature values, humidity feature values, and biological trace feature values; The training unit is used to input the sample feature values ​​of each cave sample image and its corresponding marine organism species labels into the convolutional neural network for training and cross-validation to obtain feature weight coefficients.

4. An electronic device, characterized in that: It comprises a memory storing executable program code and a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the method for predicting marine biological species as claimed in claim 1 or 2.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute the method for predicting marine biological species according to claim 1 or 2.

Images