A training method, identification method, and device for a marine trace fossil identification model.

Abstract

This invention discloses a training method, identification method, and apparatus for a marine trace fossil identification model. The method includes: acquiring and organizing raw core image data containing trace fossils of multiple genera and species to generate a core image database; preprocessing the image data in the core image database and adding marine trace fossil tag information corresponding to the image data to generate a marine trace fossil dataset; dividing the marine trace fossil dataset into a training set and a test set; training a pre-set neural network model using the training and test sets, calculating the loss rate of the training and test sets, adjusting the parameters of the neural network model, and selecting the neural network model with the lowest loss rate based on the training and test sets as the marine trace fossil identification model. This invention solves the problems of high workload, cumbersome procedures, and susceptibility to misjudgments in traditional marine trace fossil identification methods.

Description

Technical Field

[0001] This invention relates to the field of geographic information technology, and in particular to a training method, identification method and device for a marine trace fossil identification model. Background Technology

[0002] Trace fossils are traces left by biological activity in sediments and other substrates during geological history. Because trace fossils are mostly buried in situ, they are more advantageous in strata where fossil bodies and other sedimentary structures are underdeveloped. As disturbance structures, trace fossils often have a destructive effect on inorganic sedimentary structures. Trace fossils play an important role in reconstructing paleoenvironments, paleontology, paleoecology, and paleoclimate, as well as in the exploration and development of oil, gas, and solid mineral resources.

[0003] Traditional methods of trace fossil research mainly involve core description and field outcrop investigation. Core data is highly reliable data directly drilled from underground geological bodies and serves as the basis for oil and gas field exploration and development. Traditional methods mainly involve manual identification. Manual identification of trace fossils in cores is labor-intensive and tedious. Furthermore, limited by a single research method, the results obtained from traditional trace fossil research are generally affected by human subjectivity and cannot observe the characteristics of trace fossils inside the strata. Summary of the Invention

[0004] In view of the above problems, the present invention is proposed to provide a training method, identification method and apparatus for a marine trace fossil identification model that overcomes or at least partially solves the above problems.

[0005] In a first aspect, embodiments of the present invention provide a training method for a marine trace fossil identification model, comprising:

[0006] Obtain and organize original core image data containing trace fossils of multiple genera and species to generate a core image database containing trace fossils of multiple genera and species;

[0007] The image data in the core image database is preprocessed, and marine trace fossil tag information corresponding to the image data is added to generate a marine trace fossil dataset.

[0008] The marine trace fossil dataset was divided into a training set and a test set.

[0009] The preset neural network model is trained using training and test sets, and the loss rate of the training and test sets is calculated. The parameters of the neural network model are adjusted, and the neural network model with the smallest loss rate based on the training and test sets is selected as the marine trace fossil identification model.

[0010] The original core image data was acquired and processed to generate a core image database containing trace fossils from multiple genera and species, including:

[0011] The original core image data containing trace fossils of multiple genera and species were segmented according to different genera and species to obtain core image data corresponding to each genera and species.

[0012] Preprocessing the image data in the core image database includes any one or a combination of the following operations:

[0013] Adjust the color, contrast, and brightness of the image data.

[0014] Cropping or rotating image data;

[0015] Remove noisy data from the image data.

[0016] Expand the number of genera and species images that are below the preset limit.

[0017] The marine fossil record tag information corresponding to the image data is generated in the following way:

[0018] Select the target area containing trace fossils in the image data, determine the smallest rectangle of the target area, and record the coordinates of the upper left and lower right corners of the smallest rectangle;

[0019] The coordinate values ​​of the upper left and lower right corners are combined with the genus and species information of the image data to generate the marine trace fossil tag information.

[0020] The preset neural network model is a YOLOv4 network, which includes the following four parts: input, backbone network, neck connection structure, and prediction head output.

[0021] The process involves training a pre-defined neural network model using training and testing sets, calculating the loss rates of the training and testing sets, adjusting the parameters of the neural network model, and selecting the neural network model with the lowest loss rate based on the training and testing sets as the marine fossil identification model. This includes:

[0022] By adjusting the learning rate, batch_size, and epoch parameters of the neural network model, training parameters for multiple sets of neural network models can be obtained.

[0023] The training set is input into a neural network model using different sets of training parameters to train the neural network model, and the test set is input into the trained model. The loss rate of the training set and the loss rate of the test set are calculated respectively.

[0024] The neural network model with the smallest loss rate on the test set was selected as the marine fossil identification model.

[0025] The process of adjusting the learning rate, batch_size, and epoch parameters of the neural network model to obtain training parameters for multiple sets of neural network models includes:

[0026] By controlling the learning rate, batch size, and epoch parameters, we can change one parameter in different groups while keeping the other parameters constant, thus obtaining multiple sets of training parameters.

[0027] To adjust the learning rate, the cosine annealing decay method was used to set different learning rate parameters for different groups.

[0028] Secondly, embodiments of the present invention provide a method for identifying marine trace fossil identification models, comprising:

[0029] The core images to be identified are processed;

[0030] The processed image is input into the marine trace fossil identification model for identification, and the species information and distribution depth of the marine traces contained in the core image to be identified are identified; the marine trace fossil identification model is trained by the training method of the marine trace fossil identification model as described in claims 1-8.

[0031] The core image to be identified is processed, including:

[0032] Based on the size of the core image, determine the image extraction window of a preset width;

[0033] Using the upper left corner of the core image as the origin of coordinates, slide the window from the origin of coordinates and read the image data in the window in sequence until the entire core image is read.

[0034] The process of inputting the processed image into a marine trace fossil identification model for identification includes:

[0035] The extracted image data from each window is input into the marine fossil identification model for identification.

[0036] Thirdly, embodiments of the present invention provide a training device for a marine trace fossil identification model, comprising:

[0037] The acquisition module is used to acquire and organize raw core image data containing trace fossils of multiple genera and species, and generate a core image database containing trace fossils of multiple genera and species.

[0038] The dataset generation module is used to preprocess the image data in the core image database and add marine trace fossil tag information corresponding to the image data to generate a marine trace fossil dataset.

[0039] The training module is used to divide the marine trace fossil dataset into a training set and a test set; train a preset neural network model using the training set and the test set, calculate the loss rate of the training set and the test set, adjust the parameters of the neural network model, and select the neural network model with the smallest loss rate based on the training set and the test set as the marine trace fossil recognition model.

[0040] Fourthly, embodiments of the present invention provide an identification device for a marine trace fossil identification model, comprising:

[0041] The image processing module is used to process the core images to be identified;

[0042] The identification module is used to input the processed image into the marine trace fossil identification model for identification, and to identify the species information and distribution depth of the marine traces contained in the core image to be identified; the marine trace fossil identification model is trained by the training method of the marine trace fossil identification model as described in claims 1-8.

[0043] Fourthly, embodiments of the present invention provide a computing device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the training method for the marine trace fossil identification model as described above or the identification method for the marine trace fossil identification model as described above.

[0044] Fifthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements either the training method for the marine trace fossil identification model as described above or the identification method for the marine trace fossil identification model as described above.

[0045] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following:

[0046] This invention provides a training method, identification method, and apparatus for a marine trace fossil identification model. It involves organizing original core images of trace fossils from multiple genera and species to generate a core image database containing trace fossils from these genera and species. Based on the tag information corresponding to the trace fossils in the images from the core image database, a marine trace fossil dataset is generated. This dataset is divided into a training set and a test set. The marine trace fossil identification model is trained using both sets, and then used to identify marine trace fossils from the core images to be identified. This invention applies the technology of using a neural network model to identify target detection images to the identification of genera and species of trace fossils from cores, providing a fast and accurate method for this purpose. It solves the problems of high workload, tedious work, and susceptibility to subjective human factors in traditional marine trace fossil identification methods, which easily lead to misjudgments. This effectively improves the efficiency of identifying genera and species of trace fossils from cores during the exploration and development stage of oil and gas fields and can be widely applied in geological exploration and other fields.

[0047] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings.

[0048] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0049] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0050] Figure 1 This is a flowchart of the training method for the marine trace fossil identification model in an embodiment of the present invention;

[0051] Figure 2 This is a schematic diagram of the trace fossils that have been initially segmented into 9 genera and species in an embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram illustrating an example of image rotation in an embodiment of the present invention;

[0053] Figure 4 This is a schematic diagram illustrating an example of the characteristic structure of a YOLOv4 network in an embodiment of the present invention;

[0054] Figure 5 This is a schematic diagram of the Mish activation function curve in an embodiment of the present invention;

[0055] Figure 6 This is a schematic diagram illustrating the change in learning rate using the cosine annealing decay method in an embodiment of the present invention.

[0056] Figure 7 This is a schematic diagram of the identification method of the marine trace fossil identification model in an embodiment of the present invention;

[0057] Figure 8 This is a structural block diagram of the training device for the marine trace fossil identification model in this embodiment of the invention;

[0058] Figure 9 This is a structural block diagram of the identification device for the marine trace fossil identification model in an embodiment of the present invention. Detailed Implementation

[0059] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0060] This invention provides a training method for a marine trace fossil identification model, referring to... Figure 1 As shown, the method includes the following steps:

[0061] S11. Obtain and organize the original core image data containing trace fossils of multiple genera and species, and generate a core image database containing trace fossils of multiple genera and species.

[0062] S12. Preprocess the image data in the core image database and add marine trace fossil tag information corresponding to the image data to generate a marine trace fossil dataset.

[0063] S13. Divide the marine trace fossil dataset into a training set and a test set;

[0064] S14. Train the preset neural network model using the training set and test set, calculate the loss rate of the training set and test set, adjust the parameters of the neural network model, and select the neural network model with the smallest loss rate according to the training set and test set as the marine trace fossil identification model.

[0065] In step S11 above, the original core image data containing trace fossils of multiple genera and species is segmented according to different genera and species to obtain core image data corresponding to trace fossils of multiple genera and species respectively.

[0066] The original core images have extremely uneven aspect ratios, with multiple genera and species of trace fossils mixed in a single image. Therefore, it is necessary to perform segmentation processing, dividing the image according to a certain depth value and recording its depth range.

[0067] For example Figure 2 The image shown is of trace fossils that have been preliminarily divided into 9 genera and species.

[0068] In S12 above, the image data in the core image database is preprocessed, including any one or a combination of the following operations:

[0069] Adjust the color, contrast, and brightness of the image data.

[0070] Cropping or rotating image data;

[0071] Remove noisy data from the image data.

[0072] Expand the number of genera and species images that are below the preset limit.

[0073] Specifically, the preprocessing of the data mainly includes uniform adjustment of photo colors, uniform adjustment of contrast, uniform adjustment of brightness, cropping, rotation, and removal of noise data. The image size initially cropped in step S11 is too large and needs to be further cropped, for example, the image data is uniformly cropped to a resolution of 320*320.

[0074] Noise removal primarily involves deleting image data with unclear or unrepresentative category characteristics. After deleting some image data, there's a significant difference in data volume between categories. In this case, it's necessary to augment the existing image data for categories with smaller datasets. The main method is to rotate the images clockwise by 90°, 180°, and 270° and save them (e.g., ...). Figure 3 The example of image rotation shown is as follows: the leftmost image is the original image, the second image from the left is the image rotated 90 degrees, the third image from the left is the image rotated 180 degrees, and the first image from the right is the image rotated 270 degrees. This is to ensure that the amount of image data between different categories does not differ too much, thereby improving the generalization ability of the model.

[0075] In step S12 above, the marine fossil tag information corresponding to the image data is generated in the following way:

[0076] Select the target area containing trace fossils in the image data, determine the smallest rectangle of the target area, and record the coordinates of the upper left and lower right corners of the smallest rectangle;

[0077] The coordinate values ​​of the upper left and lower right corners are combined with the genus and species information of the image data to generate the marine trace fossil tag information.

[0078] For example, when selecting a target area in an image, the coordinates of the top left and bottom right corners of the smallest rectangle are recorded, and the genus information is saved together as a preset format file. The identifier of this file is consistent with the identifier of the image being processed.

[0079] Convolutional neural networks (CNNs) are deep learning algorithms that have been widely used in recent years for image-based target recognition, such as face recognition and vehicle recognition. They can accurately and efficiently identify target types from images based on target structure and color information. Marine trace fossils are a type of rock structure with certain identifiable structural features, making it feasible to use CNNs to identify the target location and fossil type of trace fossils in core images.

[0080] Furthermore, in this embodiment of the invention, the neural network model can be selected from various neural network models, such as the YOLOv4 network, which includes the following four parts: input, backbone network, neck connection structure, and prediction head output. An example of the feature structure of a YOLOv4 network is provided below. Figure 4 As shown.

[0081] The YOLOv4 architecture is characterized by its backbone feature extraction network being the CSPDarkNet53 network, and its activation function being the Mish activation function. The Mish function is:

[0082] Mish=x×tanh(ln(1+e X ))

[0083] Figure 5 The image shows the curve of the Mish activation function.

[0084] Input: Images from the input dataset (training or test set), such as 320*320 images, undergo image preprocessing. During model training, Mosaic data augmentation, CMBN across mini-batch normalization, and SAT automatic training are used. Mosaic is a data augmentation method that combines four training images into one for training. Its effect is to enhance the detection of objects outside the normal background. Each mini-batch contains a large number of images; using Mosaic increases the number of images in a mini-batch to four times, reducing the need for large mini-batches when estimating the mean and variance.

[0085] Cross-mini-batch normalization (cmBN) is an improved version of CBN. It treats the four mini-batches within a large batch as a single unit, isolating them from the outside world, and only performs merging operations within each mini-batch, ensuring that BN updates trainable parameters only once per batch. Self-Adversarial Training (SAT), also known as Automatic Training or Self-Adversarial Training, is a novel data augmentation method. In the first stage, the neural network modifies the original image instead of the network weights. This allows the neural network to perform an adversarial attack on itself, altering the original image to create the illusion of a targetless image. In the second stage, the neural network is trained to perform normal target detection on the modified image.

[0086] Backbone network: Features are extracted using the CSPDarknet53 network, and Mish activation function, Dropblock regularization, and CSP cross-stage partial connections are also used.

[0087] Neck Intermediate Layers: These are layers inserted between the Backbone and the final Head output layer. YOLOv4 adds the SPP module and FPN+PAN structure. The SPP uses max pooling methods of 1x1, 5x5, 9x9, and 13x13 for multi-scale feature fusion.

[0088] Head Output Layer: The anchor box mechanism of the output layer is to extract the scale of the prior boxes through clustering and constrain the position of the predicted bounding boxes, improve the loss function CIOU_Loss during training, and the DIOU_nms for predicting box selection.

[0089] Mosaic data augmentation is used, specifically by stitching together four images from the training or test set, flipping, scaling, and changing the color gamut of each image, arranging them in four directions, combining the images and bounding boxes, and then performing batch normalization (BN) calculations.

[0090] Select Label Smoothing for smoothing; the specific formula is as follows:

[0091] new_onehot_labels=onehot_labels*(1-label_smoothing)+label_smoothing / num_classes

[0092] Label smoothing involves transforming the original labels of 0 and 1 into 0.005 and 0.995 (for binary classification). The goal is to prevent the model from classifying too accurately, effectively avoiding overfitting. CIOU is used as the regression optimization loss. CIOU considers the distance between the target and the anchor, overlap, scale, and penalty term, making the bounding box regression more stable. The formula is as follows:

[0093]

[0094] in, represents the Euclidean distance between the center points of the predicted bounding box and the ground truth bounding box, and 'c' represents the diagonal distance of the smallest closure region that can simultaneously contain both the predicted and ground truth bounding boxes. The formulas for 'a' and 'v' are as follows:

[0095]

[0096]

[0097] The w and h mentioned above represent the width and height of the predicted bounding box, respectively. The width and height of the actual bounding box.

[0098] The formula for calculating the loss of this model is as follows:

[0099] LOSS CIOU =1-CIOU

[0100] Furthermore, in step S14 above, the learning rate parameter, batch_size parameter and epoch parameter of the neural network model are adjusted to obtain training parameters for multiple sets of neural network models.

[0101] During the training process, the training set is input into a neural network model using different sets of training parameters to train the neural network model, and the test set is input into the trained model. The loss rate of the training set and the loss rate of the test set are calculated respectively.

[0102] The neural network model with the smallest loss rate on the test set was selected as the marine fossil identification model.

[0103] Typically, the minimum loss rate of the test set should be less than the preset loss rate threshold, and the corresponding loss rate of the training set should also be less than or equal to the loss rate of the test set.

[0104] Furthermore, by adjusting the learning rate parameter, batch_size parameter, and epoch parameter of the neural network model as described above, multiple sets of training parameters for the neural network model are obtained. In specific implementation, the learning rate parameter, batch_size parameter, and epoch parameter are controlled by changing one parameter in different groups while keeping the other parameters unchanged, thus obtaining multiple sets of training parameters.

[0105] For example, when adjusting the learning rate, the cosine annealing decay method can be used to set different learning rate parameters for different groups.

[0106] The learning rate is set using a cosine annealing decay method. The learning rate will first increase and then decrease, with a linear increase during the increase and a simulated cosine (cosine) decay during the decrease. Examples of training and test set loss rates for different parameter sets with varying learning rates are shown in Table 1.

[0107] Table 1

[0108]

[0109]

[0110] Figure 6 The diagram shows the change in the learning rate when using the cosine annealing decay method.

[0111] Based on the training method of the marine trace fossil identification model described above, this embodiment of the invention also provides an identification method for the marine trace fossil identification model, referring to... Figure 7 ,include:

[0112] S71. Process the core image to be identified;

[0113] S72. Input the processed image into the marine trace fossil identification model for identification, and identify the species information and distribution depth of the marine traces contained in the core image to be identified; the above-mentioned marine trace fossil identification model is trained by the training method of the marine trace fossil identification model as described above.

[0114] Specifically, the processing in S71 above includes: determining an image extraction window of a preset width based on the size of the core image;

[0115] Using the upper left corner of the core image as the origin of coordinates, slide the window from the origin of coordinates and read the image data in the window in sequence until the entire core image is read.

[0116] The extracted image data is then input one by one into the trained marine fossil identification model to identify the species and distribution depth of the marine fossils contained therein.

[0117] For example, in the original core data to be identified, trace fossils are mixed in an image and are vertically distributed. The original image resolution is too high to be directly distinguished. This invention uses a sliding window recognition method. A window of size 280*280 is created based on the width of the original image. The top left corner of the image is used as the origin (0, 0). Starting from the origin, the image data within the window is read sequentially and extracted. A trained marine trace fossil recognition model is used for identification, and the recognition results are recorded in an Excel spreadsheet. Simultaneously, the coordinates of the top left and bottom right corners of the target frame are recorded. The window is moved downwards, and the above steps are repeated until the entire image is identified. Finally, the genera and species of trace fossils and their distribution depth values ​​in the core image can be obtained. Furthermore, the species, quantity, development location, and development density (number / m²) of trace fossils can be calculated. 2 ).

[0118] Based on the same inventive concept, this invention also provides a training device for a marine trace fossil identification model and an identification device for a marine trace fossil identification model. Since the principles by which these devices solve problems are similar to the aforementioned training method and identification method for a marine trace fossil identification model, the implementation of these devices can refer to the implementation of the aforementioned methods, and the repeated parts will not be described again.

[0119] The training device for the marine trace fossil identification model provided in this embodiment of the invention refers to... Figure 8 As shown, it includes:

[0120] The acquisition module 81 is used to acquire and organize original core image data containing trace fossils of multiple genera and species, and generate a core image database containing trace fossils of multiple genera and species.

[0121] The dataset generation module 82 is used to preprocess the image data in the core image database and add marine trace fossil tag information corresponding to the image data to generate a marine trace fossil dataset.

[0122] The training module 83 is used to divide the marine trace fossil dataset into a training set and a test set; to train a preset neural network model using the training set and the test set, to calculate the loss rate of the training set and the test set, to adjust the parameters of the neural network model, and to select the neural network model with the smallest loss rate based on the training set and the test set as the marine trace fossil recognition model.

[0123] The identification device for the marine trace fossil identification model provided in this embodiment of the invention refers to... Figure 9 As shown, it includes:

[0124] Image processing module 91 is used to process the core image to be identified;

[0125] The recognition module 92 is used to input the processed image into the marine trace fossil recognition model for recognition, and to identify the species information and distribution depth of the marine traces contained in the core image to be identified; the marine trace fossil recognition model is trained by the training method of the marine trace fossil recognition model as described in claims 1-8.

[0126] An embodiment of the present invention provides a computing device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements either the training method for the marine trace fossil identification model described above or the identification method for the marine trace fossil identification model described above.

[0127] The present invention provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, which, when executed by a processor, implements the training method for the marine trace fossil identification model as described above or the identification method for the marine trace fossil identification model as described above.

[0128] This invention provides a training method, identification method, and apparatus for a marine trace fossil identification model. It involves organizing original core images of trace fossils from multiple genera and species to generate a core image database containing trace fossils from these genera and species. Based on the tag information corresponding to the trace fossils in the images from the core image database, a marine trace fossil dataset is generated. This dataset is divided into a training set and a test set. The marine trace fossil identification model is trained using both sets, and then used to identify marine trace fossils from the core images to be identified. This invention applies the technology of using a neural network model to identify target detection images to the identification of genera and species of trace fossils from cores, providing a fast and accurate method for this purpose. It solves the problems of high workload, tedious work, and susceptibility to subjective human factors in traditional marine trace fossil identification methods, which can easily lead to misjudgments. This effectively improves the efficiency of identifying genera and species of trace fossils from cores during the exploration and development phase of oil and gas fields and can be widely applied in geological exploration and other fields.

[0129] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0130] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0131] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0132] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0133] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for training a marine ichnological fossil identification model, characterized in that, The method comprises the following steps: obtaining and organizing original core picture data containing multiple generic ichnological fossils to generate a core picture database containing multiple generic ichnological fossils; the step of obtaining and organizing the original core picture data containing multiple generic ichnological fossils to generate the core picture database containing multiple generic ichnological fossils comprises the step of segmenting the original core picture data containing multiple generic ichnological fossils according to different generic ichnological fossils to obtain core picture data corresponding to each generic ichnological fossil; the step of segmenting the original core picture data containing multiple generic ichnological fossils according to different generic ichnological fossils comprises the step of segmenting the original core picture data containing multiple generic ichnological fossils according to the depth value corresponding to each generic ichnological fossil and recording the distribution depth range corresponding to each generic ichnological fossil; performing preprocessing on the picture data in the core picture database and adding marine ichnological fossil label information corresponding to the picture data to generate a marine ichnological fossil data set; dividing the marine ichnological fossil data set into a training set and a test set; training a preset neural network model using the training set and the test set, calculating the loss rate of the training set and the test set, adjusting the parameters of the neural network model, and selecting a neural network model with the minimum loss rate of the training set and the test set as the marine ichnological fossil recognition model; the model is used to predict the generic information and distribution depth of marine ichnological fossils contained in a core image to be recognized.

2. The method of claim 1, wherein, The preprocessing of the picture data in the core picture database comprises any one or a combination of multiple operations: adjusting the color, contrast and brightness of the picture data, cropping or rotating the picture data, removing noise data in the picture data, expanding the number of generic pictures when the number of picture data is lower than a preset condition.

3. The method of claim 1, wherein, The marine ichnological fossil label information corresponding to the picture data is generated in the following way: selecting a target area containing ichnological fossils in the picture data and determining the minimum rectangular frame of the target area, and recording the coordinate values of the upper left corner and the lower right corner of the minimum rectangular frame; generating the marine ichnological fossil label information by combining the coordinate values of the upper left corner and the lower right corner with the generic information of the picture data.

4. The method of claim 1, wherein, The preset neural network model is a YOLOV4 network, and the network comprises the following four parts: an input end, a BackBone main network, a Neck connection structure and a Prediction Head output.

5. The method of claim 1, wherein, Training a preset neural network model using the training set and the test set, calculating the loss rate of the training set and the test set, adjusting the parameters of the neural network model, and selecting a neural network model with the minimum loss rate of the training set and the test set as the marine ichnological fossil recognition model, comprises: adjusting the learning rate parameter, batch_size parameter and epoch parameter of the neural network model to obtain multiple sets of training parameters of the neural network model; inputting the training set into the neural network model using different sets of training parameters to train the neural network model, and inputting the test set into the trained training model to calculate the loss rate of the training set and the loss rate of the test set, respectively; Select the neural network model with the smallest loss rate in the test set as the marine trace fossil identification model.

6. The method of claim 5, wherein, The learning rate parameter, batch_size parameter and epoch parameter of the neural network model are adjusted to obtain a plurality of sets of training parameters of the neural network model, including: The learning rate parameter, batch_size parameter and epoch parameter are changed using the method of control variable, one parameter in different groups is changed, and other parameters remain unchanged to obtain a plurality of sets of training parameters.

7. The method of claim 6, wherein, The learning rate is adjusted using the cosine annealing decay method to set different learning rate parameters in different groups.

8. A method of identifying a marine ichnological fossil identification model, characterized by, Including: processing the core image to be identified; The processed image is input into the marine trace fossil identification model for identification to identify the species information and distribution depth of the marine trace fossils contained in the core image to be identified; the marine trace fossil identification model is trained by the marine trace fossil identification model training method of claims 1-7.

9. The method of claim 8, wherein, processing the core image to be identified, including: According to the size of the core image, determine the image extraction window of the preset width; Taking the upper left corner of the core image as the coordinate origin, sliding the window from the coordinate origin, reading the image data in the window in turn, until the whole core image is read; The processed image is input into the marine trace fossil identification model for identification, including: The image data of each window extracted is input into the marine trace fossil identification model for identification respectively.

10. A device for training a model for identifying marine ichnocoenosium, characterized by, Including: An acquisition module is configured to acquire and organize original core picture data containing a plurality of generic trace fossils, and generate a core picture database containing a plurality of generic trace fossils; The original core picture data containing a plurality of generic trace fossils is segmented according to different genera to obtain core picture data corresponding to a plurality of generic trace fossils respectively; the original core picture data containing a plurality of generic trace fossils is segmented according to the depth value corresponding to each generic trace fossil, and the distribution depth range corresponding to a plurality of generic trace fossils is recorded; A dataset generation module is configured to preprocess picture data in the core picture database and add marine trace fossil label information corresponding to the picture data to generate a marine trace fossil dataset; A training module is configured to divide the marine trace fossil dataset into a training set and a test set; train a preset neural network model using the training set and the test set, calculate the loss rate of the training set and the test set, adjust the parameters of the neural network model, and select a neural network model with the smallest loss rate of the training set and the test set as the marine trace fossil identification model; the model is used to predict the species information and distribution depth of marine traces contained in a core image to be identified.

11. A recognition device of a marine ichnological fossil recognition model, characterized by, Including: An image processing module is configured to process the core image to be identified; The identification module is configured to input the processed image into a marine trace fossil identification model to identify the species information and distribution depth of the marine trace contained in the core image to be identified. The marine trace fossil identification model is obtained by training the marine trace fossil identification model according to the training method of the marine trace fossil identification model in any one of claims 1-7.

12. A computing device, comprising: The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the training method of the marine trace fossil identification model according to any one of claims 1-7 or the identification method of the marine trace fossil identification model according to claim 8 or 9. The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the training method of the marine trace fossil identification model according to any one of claims 1-7 or the identification method of the marine trace fossil identification model according to claim 8 or 9.

13. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the training method of the marine trace fossil identification model according to any one of claims 1-7 or the identification method of the marine trace fossil identification model according to claim 8 or 9.

Images