Crude oil exploration method based on image detection

Abstract

The invention provides a crude oil exploration method based on image detection, and relates to the technical field of crude oil exploration. The method comprises the steps of collecting various data; and constructing a reinforced anchor-frame-free detection model optimized based on the giant armadio algorithm, inputting historical crude oil exploration data, and outputting to obtain crude oil feature image data. Analyzing the geological structure image data to obtain key feature image data, extracting the earth surface image data to obtain earth surface image features, combining the key feature image data with the earth surface image features to obtain complete feature image data, comparing the complete feature image data with crude oil feature image data, and judging whether crude oil exists in the target area or not. According to the method, the armadio algorithm is utilized to optimize the anchor-frame-free detection model, the crude oil exploration efficiency and accuracy are improved, the adaptability and generalization ability of the model are enhanced, and complete feature image data obtained through feature extraction and fusion is beneficial to deeply understanding the geological condition of a target area.

Description

Technical Field

[0001] The present invention relates to the technical field of crude oil exploration, and in particular to a crude oil exploration method based on image detection. Background Art

[0002] In the field of crude oil exploration, the method based on image detection has gradually become an important exploration means because it can intuitively obtain and analyze geological information. With the continuous development of exploration technologies, higher requirements are put forward for the accuracy, efficiency and reliability of image detection methods.

[0003] Currently, most traditional crude oil exploration methods based on image detection rely on target detection models to identify features related to crude oil. However, these detection models generally have certain limitations. Among them, many models adopt the detection method based on anchor boxes. This method requires a large number of anchor boxes with different sizes and ratios to be set in advance to cover possible targets, resulting in a significant increase in the amount of calculation, and the training and inference processes of the model become complex and time-consuming. At the same time, the setting of anchor boxes needs to be adjusted according to specific exploration scenarios. It is difficult to find a set of optimal anchor box parameters for targets with different scales and shapes, which makes the adaptability of the model to complex and changeable crude oil exploration images poor, and it is easy to miss detections or make false detections, affecting the accuracy and efficiency of crude oil exploration.

[0004] In addition, in terms of feature extraction and fusion of image data, there are also unreasonable aspects in the existing technologies. Geological structure image data and surface image data contain rich information related to crude oil exploration, but traditional feature extraction methods often cannot fully exploit this information. In the feature fusion stage, there is a lack of accurate evaluation of feature importance and reasonable fusion strategies. This unreasonable feature fusion method makes the fused feature data unable to fully utilize the advantages of each feature, and it is difficult to provide comprehensive and accurate information for crude oil exploration judgment, thus reducing the reliability of exploration results. Summary of the Invention

[0005] The present invention provides a crude oil exploration method based on image detection to solve the defects of limitations in existing detection models and unreasonable feature extraction and fusion of image data.

[0006] On the one hand, the present invention provides a crude oil exploration method based on image detection, including: Collect geological structure image data and historical crude oil exploration data in the target area, and collect surface image data at preset time intervals.

[0007] Construct an anchor-free detection model, and use the giant armadillo algorithm to optimize the anchor-free detection model to obtain a strengthened anchor-free detection model. Input the historical crude oil exploration data, and output the obtained crude oil feature image data.

[0008] The geological structure image data is analyzed using the feature analysis method to obtain key feature image data. The surface image features are extracted from the surface image data according to the multi-scale geometric analysis feature method. The key feature image data and the surface image features are combined according to the feature fusion method to obtain complete feature image data.

[0009] The complete feature image data is compared with the crude oil feature image data to determine whether there is crude oil in the target area.

[0010] According to the crude oil exploration method based on image detection provided by the present invention, the steps of constructing an anchor-free detection model include: Extract the basic features of the images in the crude oil exploration data, and learn the basic features through residual blocks to build a network structure.

[0011] At the output layer of the network structure, a heat map is obtained through convolution operations, and the center point of the network structure is predicted. At the position of the center point, the width and height of the network structure are predicted through a convolutional layer, and the key points of the target are determined according to the center point, width, and height.

[0012] Focal-Loss is used to train the heat map for calculating the center point loss.

[0013] L1-Loss is used to calculate the difference between the actual value and the preset value to obtain the size loss.

[0014] The weighted sum of the center point loss and the size loss is used as the total loss function.

[0015] According to the crude oil exploration method based on image detection provided by the present invention, the steps of obtaining an enhanced anchor-free detection model include: Set the population size of giant armadillos. Each armadillo individual represents a set of parameters of the anchor-free detection model, and within the preset value range, the parameter values of each armadillo individual are randomly initialized.

[0016] For each armadillo individual, new digging individuals are generated through digging behaviors.

[0017] For each armadillo individual, new escaping individuals are generated by perceiving threats and escaping.

[0018] Calculate the fitness values of the initial armadillo individuals, digging individuals, and escaping individuals, and select the armadillo individual with the highest fitness value as the individual for the next iteration.

[0019] When the preset number of iterations is reached, the parameter values represented by the armadillo individual with the highest fitness value are applied to the anchor-free detection model to obtain an enhanced anchor-free detection model.

[0020] According to the oil exploration method based on image detection provided by the present invention, the steps of outputting crude oil characteristic image data include: Convert historical oil exploration data into the input requirements that meet the enhanced anchor-free detection model, and perform normalization processing to obtain exploration processed data.

[0021] Input the exploration processed data into the enhanced anchor-free detection model, perform forward propagation calculations through each layer, and perform feature extraction to obtain feature map data.

[0022] Perform key point and size prediction on the feature map data, fuse the features of different layers, and remove the predictions that do not meet the preset requirements to obtain the prediction results.

[0023] Convert the prediction results into image form and integrate them together to obtain crude oil characteristic image data.

[0024] According to the oil exploration method based on image detection provided by the present invention, the steps of obtaining feature map data include: At the first layer of the network structure, perform convolution operation on the exploration processed data to obtain the convolution result, and use the activation function to perform non-linear transformation to obtain the output feature map.

[0025] At the middle layer of the network structure, repeat the convolution operation and the operation of using the activation function. After reaching the preset number of times, extract high-level abstract features.

[0026] At the last layer of the network structure, output the output feature map and the high-level abstract features as the feature map data.

[0027] According to the oil exploration method based on image detection provided by the present invention, the steps of obtaining key feature image data include: Use the Gaussian filtering algorithm to denoise the geological structure image data, and use the histogram equalization method to enhance the contrast.

[0028] Perform edge detection, texture analysis and morphological feature extraction on the processed geological structure image data to obtain multiple target geological structure features, and calculate the correlation between each target address structure feature to obtain the correlation relationship.

[0029] According to the correlation relationship and the analysis purpose, screen each target geological target structure to obtain the key features.

[0030] Visualize the key features, convert them into image form, and integrate them to obtain the key feature image data.

[0031] According to the oil exploration method based on image detection provided by the present invention, the steps of obtaining the correlation relationship include: Represent multiple target geological structure features in vector form, and use the Pearson correlation coefficient to calculate the correlation between each pair of target geological structure feature vectors respectively, obtaining a correlation matrix, and obtaining the correlation relationship by analyzing the correlation matrix.

[0032] According to the oil exploration method based on image detection provided by the present invention, the steps of extracting surface image features include: Convert the color image in the surface image data into a grayscale image, and perform denoising processing to obtain surface image processing data.

[0033] Use Curvelet to perform multi-scale decomposition on the surface image processing data, and perform directional decomposition at each scale to obtain Curvelet coefficients.

[0034] Perform statistical analysis on the Curvelet coefficients of each scale and direction, extract relevant features, remove the Curvelet coefficients less than a preset threshold, and extract geometric structure features according to the distribution and change of the Curvelet coefficients.

[0035] Fuse the relevant features extracted at different scales and the geometric structure features extracted in different directions to obtain surface image features.

[0036] According to the oil exploration method based on image detection provided by the present invention, the steps of combining to obtain complete feature image data include: Perform standardization processing on the key feature image data and the surface image features, and use the max pooling method to make the dimensions consistent.

[0037] Obtain the preset weights for the key feature image data and the surface image features according to the principal component analysis method.

[0038] Assign different weights to the key feature image data and the surface image features according to the preset weights, and perform weighted summation to obtain complete feature image data.

[0039] According to the oil exploration method based on image detection provided by the present invention, the steps of obtaining the preset weights include: Merge the key feature image data and the surface image features into a feature matrix, and perform standardization processing to obtain a standard feature matrix.

[0040] Calculate the covariance matrix of the standard feature matrix, and perform eigenvalue decomposition to obtain multiple eigenvalues and corresponding eigenvectors.

[0041] Arrange multiple eigenvalue in descending order, select a preset number of eigenvalue and their corresponding eigenvectors to form a principal component matrix, and calculate the contribution weights of each key feature image data and each surface image feature in the principal component matrix to obtain the preset weights.

[0042] The crude oil exploration method based on image detection provided by the present invention optimizes the anchor-free detection model by using the giant armadillo algorithm, solves the problem that traditional object detection models may be difficult to achieve optimal performance in terms of relying on anchor boxes and parameter settings, etc., and achieves the beneficial effects of improving the efficiency and accuracy of crude oil exploration and enhancing the adaptability and generalization ability of the model.

[0043] The crude oil exploration method based on image detection provided by the present invention performs detailed feature extraction and fusion on geological structure image data and surface image data to obtain complete feature image data, solves the problem of difficult data feature extraction and fusion, and achieves the beneficial effects of providing more comprehensive references for geological experts and exploration personnel, helping to deeply understand the geological conditions of the target area, and providing a more powerful basis for crude oil exploration decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0045] Figure 1 is one of the flow diagrams of the crude oil exploration method based on image detection provided by the embodiments of the present invention; Figure 2 is the second flow diagram of the campus patrol device based on the Internet of Things provided by the embodiments of the present invention; Figure 3 is the third flow diagram of the campus patrol device based on the Internet of Things provided by the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] To make the objectives, technical solutions and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention with reference to the drawings in the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0047] The following will be combined with Figures 1 - 3Describe the method for crude oil exploration based on image detection of the present invention.

[0048] As Figure 1 shown, the method for crude oil exploration based on image detection provided by the embodiments of the present invention includes: Collect geological structure image data and historical crude oil exploration data within the target area, and collect surface image data at preset time intervals. Using ground penetrating radar technology, high-frequency electromagnetic waves are emitted underground, and the reflected signals are received. Through the analysis and processing of the signals, high-resolution geological structure images are generated. Such images can clearly show the formation structures, rock types, and possible geological features such as faults and cavities at different depths underground. In the target area, detection points of the ground penetrating radar are arranged according to a certain grid density to ensure that the entire area can be covered and comprehensive geological structure information can be obtained, thereby obtaining geological structure image data.

[0049] From drilling logs, well logging data, core analysis reports, oil testing data, etc. By sorting and analyzing the historical data, important information such as the distribution law of crude oil, physical properties of oil layers, and chemical composition of crude oil within the target area can be understood to obtain historical crude oil exploration data.

[0050] Collecting surface image data at preset time intervals is an important way to monitor the surface changes of the target area in real time. Using an unmanned aerial vehicle equipped with a high-resolution camera, aerial photography of the target area is carried out at a preset time interval of once a week.

[0051] Construct an anchor-free detection model, and use the giant armadillo algorithm to optimize the anchor-free detection model to obtain an enhanced anchor-free detection model. Input the historical crude oil exploration data, and output to obtain crude oil feature image data.

[0052] The steps for constructing the anchor-free detection model include: Extract the basic features of the images in the crude oil exploration data, and learn the basic features through residual blocks, thereby building a network structure, and the formula expression is:

[0053] Among them, is the input of the th layer, is the output of the th layer, is the weight, is the residual function with respect to the input and the weight

[0054] ​At the output layer of the network structure, a heatmap is obtained through convolution operations, and the center point of the network structure is predicted. At the position of the center point, the width and height of the network structure are predicted through a convolutional layer, and the key points of the target are determined based on the center point, width, and height. The formula for the size of the heatmap is expressed as:

[0055] In the formula, is the height of the heatmap, is the width of the heatmap, is the number of classes of the heatmap.

[0056] Focal-Loss is used to train the heatmap for calculating the center point loss, and the formula is expressed as:

[0057] In the formula, is the number of targets in the image, is the probability of the predicted heatmap at position for class , is the true heatmap, and are hyperparameters.

[0058] L1-Loss is used to calculate the difference between the actual value and the preset value to obtain the size loss, and the formula is expressed as:

[0059] In the formula, is the preset value of the th target, is the actual value of the th target.

[0060] The weighted sum of the center point loss and the size loss is used as the total loss function, and the formula is expressed as:

[0061] In the formula, is a hyperparameter used to balance the size loss and the center point loss, is the total loss function.

[0062] As Figure 2 shown, the steps to obtain the enhanced anchor-free detection model include: Set the population size of giant armadillos. Each armadillo individual represents a set of parameters of the anchor-free detection model, and within the preset value range, randomly initialize the parameter values of each armadillo individual.

[0063] For each armadillo individual, a new excavation individual is generated through the excavation behavior, and the formula is expressed as:

[0064] In the formula, is the parameter controlling the step size, and are random numbers between [0, 1], is the individual with the best fitness in the current population, is the th armadillo individual,

[0065]

[0066] For each armadillo individual, a new evasion individual is generated by evading through perceiving threats, and the formula is expressed as:In the formula, is the new evasion individual, is an individual randomly selected from the population, is a random number between [0, 1].

[0067] Calculate the fitness values of the initial armadillo individuals, excavation individuals, and evasion individuals, and select the armadillo individual with the highest fitness value as the individual for the next iteration.

[0068] When the preset number of iterations is reached, select the parameter values represented by the armadillo individual with the highest fitness value and apply them to the anchor-free detection model to obtain the enhanced anchor-free detection model.

[0069] The steps to output the crude oil feature image data include: Convert the historical crude oil exploration data into the input requirements that meet the enhanced anchor-free detection model, and perform normalization processing to obtain the exploration processed data.

[0070] Input the exploration processed data into the enhanced anchor-free detection model, perform forward propagation calculations through each layer, and perform feature extraction to obtain the feature map data.

[0071] The steps to obtain the feature map data include: In the first layer of the network structure, perform a convolution operation on the exploration processed data to obtain the convolution result, and use the activation function for non-linear transformation to obtain the output feature map. Extract the local features of the data through the convolution operation. The convolution kernel multiplies the corresponding area of the input data element by element and sums them to obtain the convolution result.

[0072] In the middle layer of the network structure, repeat the convolution operation and the operation of using the activation function. When the preset number of times is reached, high-level abstract features are extracted.

[0073] At the last layer of the network structure, the output feature map and the high-level abstract features are output as the feature map data.

[0074] Perform key point and size prediction on the feature map data, fuse the features of different layers, and remove the predictions that do not meet the preset requirements to obtain the prediction results.

[0075] Convert the prediction results into an image form and integrate them together to obtain the crude oil feature image data. The crude oil feature image data can visually display the detection results of the model for the potential crude oil features in the historical crude oil exploration data, and can include information such as the location, size, geological structure, and surrounding environmental landforms of the crude oil target.

[0076] Use the feature analysis method to analyze the geological structure image data to obtain the key feature image data, extract the surface image features from the surface image data according to the multi-scale geometric analysis feature method, and combine the key feature image data with the surface image features according to the feature fusion method to obtain the complete feature image data.

[0077] The steps to obtain the key feature image data include: Use the Gaussian filtering algorithm to denoise the geological structure image data, and use the histogram equalization method to enhance the contrast. The Gaussian filtering algorithm can effectively remove Gaussian noise and make the image smoother by performing weighted averaging on the image pixel points and their neighborhoods and determining the weights according to the Gaussian distribution. Histogram equalization enhances the overall contrast of the image and makes the details of the geological structure more obvious by redistributing the pixel values of the image to make the histogram distribution of the image more uniform.

[0078] Perform edge detection, texture analysis, and morphological feature extraction on the processed geological structure image data to obtain multiple target geological structure features, and calculate the correlation between each target geological structure feature to obtain the correlation relationship.

[0079] The steps to obtain the correlation relationship include: Represent multiple target geological structure features in vector form, and use the Pearson correlation coefficient to calculate the correlation between each pair of target geological structure feature vectors respectively to obtain the correlation matrix, and analyze the correlation matrix to obtain the correlation relationship.

[0080] Edge detection: The edges of the geological structure are one of the key features, such as formation boundaries, fault lines, etc. The Canny operator (Canny edge detection operator) is used to calculate the gradient magnitude and direction of the image, perform non-maximum suppression and double-threshold processing, and can detect accurate and continuous edges.

[0081] Texture analysis: Different geological structures have different texture features, such as the texture differences between sandstone and shale. By statistically analyzing the frequency of pixel pairs with different gray values in the image, texture feature parameters such as contrast, correlation, energy, and entropy are calculated to describe the texture characteristics of the geological structure.

[0082] Morphological feature extraction: Using mathematical morphology methods such as dilation, erosion, opening operation, and closing operation to extract the morphological features of the geological structure. For example, through dilation operation, the connected regions in the geological structure can be expanded, erosion operation can remove the burrs on the edges, and opening operation and closing operation can be used to remove small noise blocks and fill small holes respectively, thus highlighting the overall morphological features of the geological structure.

[0083] According to the correlation relationship and analysis purpose, each target geological structure is screened to obtain key features.

[0084] The key features are visualized, converted into image form, and integrated to obtain key feature image data.

[0085] The steps for extracting surface image features include: Convert the color image in the surface image data into a grayscale image and perform denoising to obtain surface image processing data.

[0086] Use Curvelet to perform multi-scale decomposition on the surface image processing data, and at each scale, perform direction decomposition to obtain Curvelet coefficients. The image in the surface image processing data is decomposed into sub-bands of different scales, and each sub-band represents the information of the image at different scales. The sub-bands of large scales contain the low-frequency information of the image, reflecting the overall structure and contour of the image. The sub-bands of small scales contain the high-frequency information of the image, reflecting the details and edges of the image. Geometric structure information in different directions is captured at each scale. Through direction decomposition, multiple direction sub-bands can be obtained, and each sub-band corresponds to the image features in a specific direction.

[0087] Perform statistical analysis on the Curvelet coefficients of each scale and direction, extract relevant features, remove the Curvelet coefficients smaller than the preset threshold, and extract geometric structure features according to the distribution and change of the Curvelet coefficients.

[0088] Fuse the relevant features extracted at different scales and the geometric structure features extracted in different directions to obtain surface image features. Surface image features can include color features, texture features, shape features, spatial features, and spectral features, etc.

[0089] As Figure 3 shown, the steps for combining to obtain complete feature image data include: Standardize the key feature image data and the surface image features, and use the max pooling method to make the dimensions consistent.

[0090] Obtain the preset weights for the key feature image data and the surface image features according to the principal component analysis method.

[0091] Assign different weights to the key feature image data and the surface image features according to the preset weights, and perform weighted summation to obtain the complete feature image data.

[0092] The steps to obtain the preset weights include: Merge the key feature image data and the surface image features into a feature matrix, and perform standardization to obtain a standard feature matrix.

[0093] Calculate the covariance matrix of the standard feature matrix, and perform eigen decomposition to obtain multiple eigenvalues and corresponding eigenvectors.

[0094] Arrange the multiple eigenvalues in descending order, select a preset number of eigenvalues and corresponding eigenvectors to form a principal component matrix, and calculate the contribution weights of each key feature image data and each surface image feature in the principal component matrix to obtain the preset weights.

[0095] Compare the complete feature image data with the crude oil feature image data to determine whether there is crude oil in the target area. Observe the information such as the potential crude oil target position and size shown in the crude oil feature image data, and match them one by one with the geological structure, surface morphology and other features in the complete feature image data. For various feature parameters extracted from the crude oil feature image data and the complete feature image data, such as the contrast and correlation in texture features, and the perimeter and area in shape features, perform quantitative comparison. By calculating the degree of difference between these feature parameters, evaluate the matching degree of the two. Based on the comparison results in all aspects above, make a final conclusion on whether there is crude oil in the target area. If multiple features and parameters show signs related to the presence of crude oil during the comparison process, and the relevance of the geological structure also supports this judgment, then it can be considered that the possibility of crude oil in the target area is relatively high; on the contrary, if there are many contradictions and inconsistencies in the comparison results, or there is a lack of key supporting evidence, then further exploration and research are needed to confirm the presence or absence of crude oil.

[0096] Example 1: 1. Data collection.

[0097] Geological structure image data: Geological structure image data with a resolution of 500×500 pixels in the target area was obtained through equipment such as ground penetrating radar. A total of 100 images were collected, and these images reflect the geological structures at different depths underground.

[0098] Historical crude oil exploration data: The crude oil exploration data of the target area in the past 10 years was collected, including previous drilling records, logging data, and some image data of marked oil-bearing areas, etc., and 80 sets of effective data records were sorted out.

[0099] Surface image data: Surface images were obtained by drone photography at a preset time interval of once a week, with a resolution of 800×800 pixels, and a total of 50 color images were collected.

[0100] 2. Construct and optimize the detection model.

[0101] Construct an anchor-free detection model: Use a residual network structure to build a basic anchor-free detection model, extract the basic features of the images in the crude oil exploration data, and learn the basic features through residual blocks. Assume that the dimension of the input image data is 3×224×224 (number of channels × height × width). After being processed by multiple residual blocks, a heat map with a size of 14×14×10 (height × width × number of classes, assuming the number of classes is 10 to distinguish different feature types) is obtained through convolution operation at the output layer, and the center point of the network structure and the width and height at the center point position are predicted, so as to determine the key points of the target.

[0102] Optimize the model: Set the population size of giant armadillos to 50, and each armadillo individual represents a set of parameters of the anchor-free detection model. Randomly initialize the parameter values of each armadillo individual within the preset value range. Generate new individuals through digging behavior and evasion behavior, and calculate the fitness values of the initial armadillo individuals, digging individuals, and evasion individuals. After 100 preset iteration times, select the parameter values represented by the armadillo individual with the highest fitness value and apply them to the anchor-free detection model to obtain a strengthened anchor-free detection model. Before optimization, the mAP of the model on the validation set was 0.65, and after optimization, it was improved to 0.78.

[0103] 3. Data feature extraction Geological structure image data: Perform denoising processing on 100 pieces of geological structure image data using the Gaussian filtering algorithm (standard deviation set to 1.5), and then use the histogram equalization method to enhance the contrast.

[0104] Use the Canny operator for edge detection, and detect obvious edge features such as formation boundaries and fault lines in 80 images; calculate texture feature parameters using GLCM, and obtain an average contrast of 0.35, an average correlation of 0.68, an average energy of 0.22, and an average entropy of 1.85; extract morphological features through morphological operations such as dilation and erosion, and find that there are closed or semi-closed morphological structures in 60 images that are conducive to crude oil storage.

[0105] Represent multiple target geological structure features in vector form, calculate the correlation between each feature using the Pearson correlation coefficient to obtain a correlation matrix. According to the correlation relationship and analysis purpose, screen out 30 key features, and perform visualization processing and integration to obtain key feature image data.

[0106] Surface image data: Convert 50 color surface images into grayscale images, and perform denoising processing using median filtering to obtain surface image processing data.

[0107] Use Curvelet to perform multi-scale decomposition on the surface image processing data (set the number of scales to 4), and perform directional decomposition at each scale (set the number of directions at each scale to 8) to obtain Curvelet coefficients. Conduct statistical analysis on the Curvelet coefficients for each scale and direction, remove coefficients less than the preset threshold (set to 0.05), and extract geometric structure features based on the distribution and change of the coefficients. Fuse the features extracted from different scales and directions to obtain surface image features. For example, it is found that there are surface texture change features possibly related to crude oil leakage in 35 images.

[0108] 4. Feature fusion Perform standardization processing on the key feature image data and the surface image features, and use the max pooling method to make the dimensions consistent, so that their dimensions both become 1×100 (set as the feature dimension obtained after processing before fusion).

[0109] According to the principal component analysis method, combine the two into a feature matrix, and perform standardization processing to obtain a standard feature matrix. Calculate the covariance matrix of the standard feature matrix, and perform eigen-decomposition to obtain multiple eigenvalues and corresponding eigenvectors. Arrange the eigenvalues from largest to smallest, select the first 10 eigenvalues and corresponding eigenvectors to form a principal component matrix, and calculate that the preset weight of the key feature image data is 0.6, and the preset weight of the surface image features is 0.4.

[0110] Assign weights to the key feature image data and the surface image features according to the preset weights, and perform weighted summation to obtain the complete feature image data.

[0111] 5. Crude oil feature judgment Convert historical crude oil exploration data into the input requirements that meet the enhanced anchor-free detection model, and perform normalization processing to obtain exploration processing data. Input the exploration processing data into the enhanced anchor-free detection model, and through forward propagation calculation and feature extraction of each layer, obtain feature map data. Perform key point and size prediction on the feature map data, fuse the features of different layers and remove predictions that do not meet the preset requirements, obtain the prediction results and convert them into image form, and integrate to obtain crude oil feature image data.

[0112] The crude oil exploration method based on image detection provided in this embodiment, through the optimization of the anchor-free detection model combined with the giant armadillo algorithm, enables the model to adapt to different types of crude oil exploration image data, can more accurately extract crude oil feature image data from historical crude oil exploration data, and combines the analysis and processing of geological structure image data and surface image data to comprehensively judge whether there is crude oil in the target area, improving the efficiency and accuracy of crude oil exploration, reducing blind exploration, and lowering exploration costs. In addition, feature extraction and fusion are performed on the geological structure image data and surface image data, and the obtained complete feature image data contains rich information such as geological structures and surface morphologies, providing a more comprehensive reference for geological experts and exploration personnel, helping to deeply understand the geological conditions of the target area, and providing a more powerful basis for crude oil exploration decisions.

[0113] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0114] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, also by hardware. Based on this understanding, the above technical solutions, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disc, etc., including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An oil exploration method based on image detection, characterized in that, Including: Collecting geological structure image data and historical crude oil exploration data within the target area, and collecting surface image data at preset time intervals; Constructing an anchor-free detection model, and optimizing the anchor-free detection model using the giant armadillo algorithm to obtain a strengthened anchor-free detection model, inputting the historical crude oil exploration data, and outputting crude oil feature image data; Analyzing the geological structure image data using the feature analysis method to obtain key feature image data, extracting surface image features from the surface image data according to the multi-scale geometric analysis feature method, and combining the key feature image data with the surface image features according to the feature fusion method to obtain complete feature image data; Comparing the complete feature image data with the crude oil feature image data to determine whether there is crude oil in the target area.

2. The method for crude oil exploration based on image detection according to claim 1, wherein The steps of constructing the anchor-free detection model include: Extracting the basic features of the images in the crude oil exploration data, and learning the basic features through residual blocks to build a network structure; At the output layer of the network structure, obtaining a heat map through convolution operations, predicting the center point of the network structure, at the position of the center point, predicting the width and height of the network structure through a convolutional layer, and determining the key points of the target according to the center point, the width, and the height; Using Focal-Loss to train the heat map for calculating the center point loss; Using L1-Loss to calculate the difference between the actual value and the preset value to obtain the size loss; Taking the weighted sum of the center point loss and the size loss as the total loss function.

3. The crude oil exploration method based on image detection according to claim 1, wherein The steps of obtaining the strengthened anchor-free detection model include: Setting the population size of the giant armadillo, each armadillo individual representing a set of parameters of the anchor-free detection model, and randomly initializing the parameter values of each armadillo individual within a preset value range; For each armadillo individual, generating a new digging individual through the digging behavior; For each armadillo individual, generating a new escaping individual by perceiving threats and escaping; Calculating the fitness values of the initial armadillo individuals, the digging individuals, and the escaping individuals, and selecting the armadillo individual with the highest fitness value as the individual for the next iteration; After reaching the preset number of iterations, selecting the parameter values represented by the armadillo individual with the highest fitness value and applying them to the anchor-free detection model to obtain the strengthened anchor-free detection model.

4. The crude oil exploration method based on image detection according to claim 2, characterized in that, The steps of outputting the crude oil feature image data include: Converting the historical crude oil exploration data to meet the input requirements of the strengthened anchor-free detection model, and performing normalization processing to obtain exploration processing data; Inputting the exploration processing data into the strengthened anchor-free detection model, performing forward propagation calculations through each layer, and performing feature extraction to obtain feature map data; Performing key point and size predictions on the feature map data, performing fusion operations on the features of different layers, and removing predictions that do not meet the preset requirements to obtain a prediction result; Converting the prediction result into an image form and integrating them together to obtain crude oil feature image data.

5. The method for crude oil exploration based on image detection according to claim 4, wherein The steps of obtaining the feature map data include: At the first layer of the network structure, a convolution operation is performed on the exploration processing data to obtain a convolution result, and a non-linear transformation is performed using an activation function to obtain an output feature map; At the middle layer of the network structure, the convolution operation and the operation of using the activation function are repeated. After reaching the preset number of times, high-level abstract features are extracted; At the last layer of the network structure, the output feature map and the high-level abstract features are output as the feature map data.

6. The method for crude oil exploration based on image detection according to claim 1, wherein The steps for obtaining the key feature image data include: Use the Gaussian filtering algorithm to denoise the geological structure image data, and use the histogram equalization method to enhance the contrast; Perform edge detection, texture analysis, and morphological feature extraction on the processed geological structure image data to obtain multiple target geological structure features, and calculate the correlation between each target geological structure feature to obtain a correlation relationship; According to the correlation relationship and the analysis purpose, each target geological target structure is screened to obtain key features; Visualize the key features, convert them into an image form, and integrate them to obtain the key feature image data.

7. The method for crude oil exploration based on image detection according to claim 6, wherein The steps for obtaining the correlation relationship include: Represent multiple target geological structure features in vector form, and use the Pearson correlation coefficient to calculate the correlation between each two target geological structure feature vectors respectively to obtain a correlation matrix, and analyze the correlation matrix to obtain the correlation relationship.

8. The method for crude oil exploration based on image detection according to claim 1, wherein The steps for extracting the surface image features include: Convert the color image in the surface image data into a grayscale image and perform denoising processing to obtain surface image processing data; Use Curvelet to perform multi-scale decomposition on the surface image processing data, and perform directional decomposition at each scale to obtain Curvelet coefficients; Perform statistical analysis on the Curvelet coefficients of each scale and direction, extract relevant features, remove the Curvelet coefficients smaller than the preset threshold, and extract geometric structure features according to the distribution and change of the Curvelet coefficients; Fuse the relevant features extracted at different scales and the geometric structure features extracted in different directions to obtain the surface image features.

9. The method for crude oil exploration based on image detection according to claim 1, wherein The steps for combining to obtain the complete feature image data include: Perform normalization processing on the key feature image data and the surface image features, and use the max pooling method to make the dimensions consistent; Obtain the preset weights for the key feature image data and the surface image features according to the principal component analysis method; Assign different weights to the key feature image data and the surface image features according to the preset weights, and perform weighted summation to obtain the complete feature image data.

10. The method for crude oil exploration based on image detection according to claim 9, wherein The steps for obtaining the preset weights include: Merge the key feature image data and the surface image features into a feature matrix, and perform normalization processing to obtain a standard feature matrix; Calculate the covariance matrix of the standard feature matrix, and perform eigenvalue decomposition to obtain multiple eigenvalues and corresponding eigenvectors; Arrange multiple eigenvalue in descending order, select a preset number of eigenvalues and the corresponding eigenvectors to form a principal component matrix, and calculate the contribution weights of each key feature image data and each surface image feature in the principal component matrix to obtain the preset weights.

Images