Coal mine underground target retrieval method and system

Abstract

The invention provides a coal mine underground target retrieval method and system, and belongs to the technical field of image retrieval. The method comprises the following steps: constructing a target database of underground coal mine targets to be retrieved based on the targets to be retrieved; obtaining a high-frequency image and a low-frequency image of a to-be-retrieved target Based on the Markov conversion field, obtaining a low-frequency image Markov graph of the to-be-retrieved target and a Markov graph of a low-frequency target database; extracting features of the target to be retrieved and the target database by adopting a feature extractor, and carrying out binarization processing; and on the basis of the binarized features, respectively calculating the similarity between the target to be retrieved and each target in the target database to obtain similarity measurement, and sorting the similarity measurement to obtain a final retrieval result. The adverse effect caused by an external complex environment is effectively reduced, the accuracy of target description is further improved, and the retrieval efficiency is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of image retrieval, and in particular relates to a method and system for underground coal mine target retrieval. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] In the context of the coal mining industry accelerating towards intelligence, efficient monitoring and precise management of underground coal mine working environments have become key needs. With the continuous expansion of mining depth and scale, underground equipment and facilities are becoming increasingly complex, the scope of personnel activities is wide and complex, and there are many uncertainties in geological conditions. Through the effective retrieval of underground images, specific targets can be quickly and accurately located, whether it is troubleshooting and maintenance supervision of key equipment, or real-time tracking and safety protection of personnel, as well as early warning and dynamic monitoring of geological disaster risk areas, it can provide strong technical support.

[0004] Traditional coal mine underground target retrieval mainly follows the daily scene target retrieval technology, namely text-based image retrieval and content-based image retrieval. In terms of text-based image retrieval, it is often necessary to manually add descriptive text labels to the image in advance. These labels cover information such as the name, model, location characteristics, personnel identification, action behavior keywords, and related terms of geological structure of coal mine underground equipment. When searching, the user enters relevant text keywords, and the system selects the target image based on the degree of match between these keywords and image labels. However, this method has obvious defects. Due to the complex and changeable environment of coal mines, the manually annotated text labels may not be accurate and comprehensive enough, and it is difficult to update in real time to reflect the dynamic changes of underground conditions, which can easily lead to deviations and omissions in retrieval results. In recent years, content-based image retrieval has been widely favored by researchers, focusing on extracting features directly from the visual content of the image itself for retrieval. For example, for coal mine underground equipment images, the shape, color, texture and other visual elements are analyzed to construct feature vectors. For personnel images, the contour, posture, clothing texture and other features of the human body may be focused on. In geological structure images, we focus on the texture orientation, crack shape and distribution of rock formations, etc. However, the special environmental conditions in coal mines, such as dim and unstable lighting, diffuse dust, and obstruction by various equipment and cables, will seriously interfere with the accurate extraction and feature construction of image visual content, making content-based image retrieval face many challenges when applied in coal mines. The accuracy and reliability of its retrieval are greatly reduced, and it is difficult to efficiently meet the stringent requirements of target retrieval tasks such as safety monitoring, equipment management, and personnel positioning in coal mines. Summary of the invention In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a coal mine underground target retrieval method and system. On the one hand, before extracting the features of the coal mine underground image, in order to increase the accuracy of the image description, a coal mine underground target image analysis method based on Fourier transform and Markov algorithm is proposed; on the other hand, in the feature construction stage, in order to improve the retrieval efficiency, a feature binarization method based on Hill diversity is proposed.

[0005] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of the present invention provides a method for underground target retrieval in a coal mine; A method for underground coal mine target retrieval, comprising: Building a target database of targets to be retrieved in coal mines based on the targets to be retrieved, wherein the target database includes a high-frequency target database and a low-frequency target database; Acquire a high-frequency image and a low-frequency image of the target to be retrieved, input the low-frequency image of the target to be retrieved and the low-frequency target database into the Markov transformation field respectively, and acquire a Markov graph of the low-frequency image of the target to be retrieved and a Markov graph of the low-frequency target database; Extracting the features of the target to be retrieved and the features of the target database respectively, and performing binarization processing on the extracted features; wherein the features of the target to be retrieved are composed of the high-frequency image features of the target to be retrieved and the low-frequency image Markov graph features of the target to be retrieved; the features of the target database are composed of the high-frequency target database features and the Markov graph features of the low-frequency target database; Based on the binarized features, the similarity between the target to be retrieved and each target in the target database is calculated respectively to obtain a similarity measure, and the final retrieval result is obtained by sorting the similarity measures.

[0006] As a further technical solution, the process of obtaining a high-frequency target database and a low-frequency target database is: by traversing the target database, obtaining a high-frequency image and a low-frequency image of each target data in the target database corresponding to the target to be retrieved, and constructing a high-frequency target database and a low-frequency target database respectively.

[0007] As a further technical solution, a high-frequency image and a low-frequency image of the target to be retrieved are obtained by performing Fourier transform on the target to be retrieved; specifically: Perform Fourier transform on the target to be retrieved to obtain a spectrum image of the target to be retrieved; the spectrum image is as follows:

[0008] In the formula, Represents the gray value of the pixel at the coordinate (m,n) in the target to be retrieved, represents the corresponding spectrum; M is the width of the target to be retrieved; N is the height of the target to be retrieved; u is the frequency domain horizontal coordinate; v is the frequency domain vertical coordinate; Center the spectrum image and move the zero-frequency part to the center of the spectrum; A Gaussian high-pass filter is used to perform high-pass filtering to obtain a high-pass filtering spectrum diagram; a Gaussian low-pass filter is used to perform low-pass filtering to obtain a low-pass filtering spectrum diagram, and the zero-frequency domain part is moved back to the original position; The high-pass filter spectrum diagram and the low-pass filter spectrum diagram are respectively subjected to inverse Fourier transform to generate a high-frequency image and a low-frequency image of the target to be retrieved.

[0009] As a further technical solution, the low-frequency image of the target to be retrieved is input into the Markov transformation field, and the process of obtaining the Markov graph of the low-frequency image of the target to be retrieved is: Filling the low-frequency image of the target to be retrieved, obtaining the filled image and calculating the size of the filled image; Segmenting the filled image to obtain image blocks; performing statistical processing on each image block to obtain a final description vector of the image block; The final description vector is converted into a Markov field by using a Markov transformation field, and the Markov field is scaled to a size equal to the low-frequency image of the target to be retrieved, thereby obtaining a Markov graph of the low-frequency image of the target to be retrieved.

[0010] As a further technical solution, the features of the target to be retrieved are composed of the high-frequency image features of the target to be retrieved and the low-frequency image Markov graph features of the target to be retrieved. The features of the target to be retrieved are as follows:

[0011] In the formula, is the feature of the target to be retrieved; is a high-frequency image feature; is the low-frequency image Markov graph feature of the target to be retrieved; The characteristics of the target database are composed of the high-frequency target database characteristics and the Markov graph characteristics of the low-frequency target database. The characteristics of the target database are as follows:

[0012] In the formula, is the characteristic of the target database; Represents the Nth feature of the target database; is the Nth feature of the high-frequency target database; is the Nth feature of the Markov graph of the low-frequency target database.

[0013] As a further technical solution, the process of binarizing the extracted features is as follows: Dividing the features of the target to be retrieved and the features of the target database, and finding the neighborhood of each feature component; Hill diversity is calculated based on neighborhood values. The formula for Hill diversity is:

[0014] In the formula, is the characteristic component; N( )for Neighborhood of; is the characteristic component Hill diversity value; for Neighborhood N( ) elements; based on Hill diversity, the features of the target to be retrieved and the features of the target database are binarized to obtain the binarized features.

[0015] As a further technical solution, the similarity measure is:

[0016] Where D is the similarity measure; is the feature of the target to be retrieved after binary processing; is the Nth feature of the target database after binarization.

[0017] A second aspect of the present invention provides a coal mine underground target retrieval system.

[0018] A coal mine underground target retrieval system, comprising: The target database construction module is configured to: construct a target database of targets to be retrieved in coal mines based on the targets to be retrieved, wherein the target database includes a high-frequency target database and a low-frequency target database; The image analysis module is configured to: obtain a high-frequency image and a low-frequency image of the target to be retrieved, input the low-frequency image of the target to be retrieved and the low-frequency target database into the Markov transformation field respectively, and obtain a Markov graph of the low-frequency image of the target to be retrieved and a Markov graph of the low-frequency target database; The feature extraction and binarization processing module is configured to: extract the features of the target to be retrieved and the features of the target database respectively, and perform binarization processing on the extracted features; wherein the features of the target to be retrieved are composed of the high-frequency image features of the target to be retrieved and the low-frequency image Markov graph features of the target to be retrieved; the features of the target database are composed of the high-frequency target database features and the Markov graph features of the low-frequency target database; The retrieval module is configured to: calculate the similarity between the target to be retrieved and each target in the target database based on the binarized features, obtain similarity metrics, and obtain the final retrieval result by sorting the similarity metrics.

[0019] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of a method for underground coal mine target retrieval as described in the first aspect of the present invention.

[0020] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of a method for underground target retrieval in a coal mine as described in the first aspect of the present invention are implemented.

[0021] One or more of the above technical solutions have the following beneficial effects: (1) The present invention uses Fourier transform and Markov algorithm to retrieve target images in coal mines. By combining descriptions in different domains, the adverse effects of complex external environments on the accurate extraction of image visual content and feature construction are effectively reduced, thereby further increasing the accuracy of target description.

[0022] (2) The present invention uses Hill diversity to perform binary processing on features, which effectively solves the problem of heavy computational tasks and low retrieval efficiency when calculating similarity metrics based on extracted features, thereby improving retrieval efficiency while ensuring accuracy.

[0023] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0025] Figure 1 This is a flow chart of the method of the first embodiment.

[0026] Figure 2 This is a flowchart of obtaining a high-frequency image of a target to be retrieved in the first embodiment.

[0027] Figure 3 This is a flow chart of obtaining a low-frequency image Markov graph of a target to be retrieved in the first embodiment.

[0028] Figure 4 It is a schematic diagram of the four-level segmentation of the filled image in the first embodiment.

[0029] Figure 5 This is the feature binarization flow chart in the first embodiment.

[0030] Figure 6 It is a system structure diagram of the second embodiment. DETAILED DESCRIPTION

[0031] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0032] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0033] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0034] The present invention proposes a method and system for underground coal mine target retrieval, which uses Fourier transform and Markov algorithm to analyze underground coal mine target images, and increases the accuracy of target description by combining different domain descriptions. On the other hand, considering the existing method or directly calculating similarity metrics based on extracted features, resulting in low retrieval efficiency, the present invention proposes a feature binarization method based on Hill diversity.

[0035] Embodiment 1 This embodiment discloses a method for searching underground coal mine targets; like Figure 1 As shown, a method for underground coal mine target retrieval includes: Step S1, constructing a target database of targets to be searched underground in a coal mine based on the targets to be searched, wherein the target database includes a high-frequency target database and a low-frequency target database.

[0036] In step S1, a target database of underground coal mine targets to be retrieved is constructed based on the targets to be retrieved. , where , is the target data in the database Q with a capacity of N. Based on the target to be retrieved, traverse the target database Q to obtain the high-frequency image of each target data in the target database corresponding to the target to be retrieved and low frequency images , forming a high-frequency target database and low frequency target database .

[0037] Step S2: Obtain the high-frequency image of the target to be retrieved and low frequency images ; The low-frequency image of the target to be retrieved and low frequency target database Input them into the Markov transformation field respectively to obtain the low-frequency image Markov map of the target to be retrieved and Markov graph of low-frequency target database .

[0038] In step S2, the target to be retrieved is recorded as I. In this embodiment, the target to be retrieved I is obtained under discrete time signals. By performing Fourier transform on the target to be retrieved, a high-frequency image of the target to be retrieved is obtained. and low frequency images .like Figure 2 As shown in the figure, Fourier transform is a powerful tool widely used in mathematics and engineering. Its basic idea is to convert a function in the time domain (or space domain) (for an image, it can be regarded as the distribution of pixel values ​​in space) into the frequency domain for analysis. Specifically, for a complex periodic signal (in an image, it can be analogous to a periodically changing part such as the texture of the image), Fourier transform can decompose it into a combination of sine and cosine functions of different frequencies. In image analysis, the spectrum of the image can be obtained through Fourier transform. The low-frequency part of the spectrum usually corresponds to the smoother area in the image, such as a large area of ​​background; the high-frequency part is related to the details and edges in the image.

[0039] Step S21, Fourier transform is performed on the target to be searched to obtain a spectrum image of the target to be searched. The obtained spectrum image is as follows:

[0040] In the formula, Represents the gray value of the pixel at the coordinate (m,n) in the target to be retrieved, Represents the corresponding frequency spectrum; M is the width of the target to be retrieved; N is the height of the target to be retrieved; u is the horizontal coordinate in the frequency domain; v is the vertical coordinate in the frequency domain.

[0041] Step S22, centering the spectrum image, moving the zero-frequency domain part to the center of the spectrum, so that the zero-frequency component of the spectrum is located in the center, which is convenient for observation and analysis, further reflecting the symmetry of the spectrum, facilitating frequency domain filtering operations, and facilitating intuitive understanding of frequency distribution.

[0042] Step S23, using a Gaussian high-pass filter to perform high-pass filtering, using a Gaussian low-pass filter to perform low-pass filtering, obtaining a high-pass filtering spectrum diagram and a low-pass filtering spectrum diagram, and moving the zero frequency domain part back to the original position to provide a correct input for subsequent frequency domain operations, so as to perform an inverse Fourier transform; wherein the process of using a Gaussian high-pass filter to perform high-pass filtering and obtaining a high-pass filtering spectrum diagram is as follows:

[0043] In the formula, is the high-pass filter spectrum diagram; u is the frequency domain horizontal coordinate; v is the frequency domain vertical coordinate; is a constant, is a point in the frequency domain The distance from the center of the frequency domain rectangle.

[0044] A Gaussian low-pass filter is used for low-pass filtering. The process of obtaining the low-pass filtering spectrum is as follows:

[0045] In the formula, is the low-pass filter spectrum diagram; u is the frequency domain horizontal coordinate; v is the frequency domain vertical coordinate; is a constant, is a point in the frequency domain The distance from the center of the frequency domain rectangle.

[0046] Step S24, respectively perform Fourier inverse transform on the high-pass filter spectrum diagram and the low-pass filter spectrum diagram to obtain a high-frequency image and a low-frequency image of the target to be retrieved. The Fourier inverse transform is:

[0047] In the formula, Represents the gray value of the pixel at the coordinate (m,n) in the target to be retrieved, Represents the corresponding frequency spectrum; M is the width of the target to be retrieved; N is the height of the target to be retrieved; u is the horizontal coordinate in the frequency domain; v is the vertical coordinate in the frequency domain.

[0048] Furthermore, in step S2, for relatively smooth low-frequency images, it is considered that more pixels belong to background information and contribute less to the image description. In order to extract the most concise and comprehensive description, in this embodiment, the Markov transition field is obtained to extract its features. Markov Transition Field (MTF) is a tool for processing time series data. It is developed based on the concept of Markov Chain. Markov chain is a random process in which the future state depends only on the current state and has nothing to do with the past state (given the current state). The Markov transition field extends the state transition information of the Markov chain to a two-dimensional field representation.

[0049] Combination Figure 3 , the low-frequency image of the target to be retrieved and low frequency target database Input them into the Markov transformation field respectively to obtain the low-frequency image Markov map of the target to be retrieved and Markov graph of low-frequency target database The process includes: Step S25: low-frequency image of the target to be retrieved Fill, get the filled image and calculate the size of the filled image; fill the low-frequency image The size of , where W_L_Img, H_L_Img, and C_L_Img are low-frequency images respectively. The width, height and number of channels are calculated in this embodiment. The filling amount in the w direction and the h direction is calculated respectively. , , the filling amount is calculated as follows: ; .

[0050] Add pixel values ​​to the filled area, get the filled image, and calculate the size of the filled image. As shown below: ; Where W_PaddingImg, H_PaddingImg, and C_PaddingImg are the width, height, and number of channels of the padding image, respectively. C_Img is the abbreviation of C_L_Img, which indicates the number of channels of the low-frequency image.

[0051] Step S26, segmenting the filled image to obtain image blocks; performing statistical processing on each image block to obtain a final description vector of the image block; Combination Figure 4 , the padding image PaddingImg is segmented into 4 levels, and the segmented blocks are recorded as P_1, P_2, P_3, P_4, P_5, P_6, P_7, P_8, P_9, P_10, P_12, P_13, P_14, P_15, P_16.

[0052] For each image block , calculate its statistical data, such as the mean ,variance , Standard Deviation , majority , median , skewness , Kurtosis , Very bad , forming a description vector , describing the vector As shown below:

[0053] The description vectors of the image blocks are combined to form the final description vector S of the image, as shown in the following formula:

[0054] In the formula, Describes the vector S The characteristic component value of the dimension.

[0055] Step S27, using the Markov transformation field, converting the final description vector into a Markov field, scaling the Markov field to the same size as the low-frequency image, and obtaining a Markov map of the low-frequency image of the target to be retrieved. .

[0056] First, the number of quantile bins is determined. In this embodiment, the number of quantile bins is set to 32, which is the quantile bin of the data containing timestamp i. and contains the data quantile bins for time axis j The defined Markov field is shown as follows:

[0057] Where M is the defined Markov field, express Finally, the Markov field is scaled to a low-frequency image Equal size, obtain the low-frequency image Markov map of the target to be retrieved Correspondingly, for the low-frequency target database By performing the above operation on each data in, the Markov graph MLQ of the low-frequency target database can be obtained. .

[0058] Step S3, respectively extracting the features of the target to be retrieved and the features of the target database, and performing binarization processing on the extracted features; wherein the features of the target to be retrieved are obtained by high-frequency image The features of and the low-frequency image Markov graph of the target to be retrieved Feature composition; the target database features are composed of high-frequency target database and Markov graph of low-frequency target database Features composition; In step S3, in the process of obtaining the final image description, the number of features is increased by 2 times compared with the image description based on the original image, taking into account the features of the tandem high-frequency image and the Markov image. As the number of features increases, the retrieval efficiency decreases. In order to optimize the retrieval efficiency of the system, in this embodiment, the Hill diversity is used to binarize the features, thereby improving the retrieval efficiency while ensuring the accuracy.

[0059] Hill integrated species richness and relative abundance into a type of diversity measurement index, and called these measurement indicators Hill numbers or effective numbers of species. The calculation of Hill numbers is shown as follows:

[0060] in, is the Hill number, i represents the i-th species, and S is the total number of species. represents the relative abundance of the i-th species. When q=0, Refers to species richness, which does not take into account the relative abundance of species. When q approaches 1, the limit of this formula is the exponent of Shannon entropy, which represents Shannon diversity; when q=1, individuals are counted equally, and the number of species calculated is proportional to their abundance. When q=2, it represents the Simpson index, and the role of other species except dominant species is underestimated, which is interpreted as the effective number of dominant species in the community.

[0061] Combination Figure 5 , m represents the feature dimension, Represents the features to be binarized. For each feature component , find its 3 nearest neighbors , as shown below:

[0062] In the formula, For the neighborhood; , , They are the minimum index value, the second smallest index value, and the maximum index value of the neighborhood respectively; Hill diversity is calculated based on neighborhood values. The Hill diversity formula is:

[0063] In the formula, is the characteristic component Hill diversity value; for The 3-neighborhood N( ) in the element; is the characteristic component; N( )for 3 neighborhoods.

[0064] Based on Hill diversity, the characteristics Binarize and get the binarized features , the parameters in the binary features As shown below:

[0065] In the formula, is the characteristic component Hill diversity value; is the sum of the Hill diversity of all feature components.

[0066] Specifically, the features of the target I to be retrieved are determined by the high-frequency image of the target I to be retrieved. Low-frequency image Markov graph of features and targets to be retrieved Feature composition, extracting high-frequency images of the target I to be retrieved The characteristics are shown below:

[0067] In the formula, is the feature of the high-frequency image of the target to be retrieved.

[0068] Extract the low-frequency image Markov graph of the target to be retrieved The characteristics are shown below:

[0069] In the formula, It is the feature of the low-frequency image Markov graph of the extracted target to be retrieved.

[0070] The characteristics of the target database Q are composed of the high-frequency target database and Markov graph of low-frequency target database Feature composition, extracting high-frequency target database The characteristics are as follows:

[0071] In the formula, is the feature of the extracted high-frequency target database; Representing images The characteristics are as follows:

[0072] Extracting Markov graphs from low-frequency target databases The characteristics are shown below:

[0073] In the formula, The features of the Markov graph of the extracted low-frequency target database; Representing images The characteristics are shown in the following formula:

[0074] In the feature engineering stage, we first fuse the high-frequency images of the target I to be retrieved. Low-frequency image Markov graph of features and targets to be retrieved The features constitute the features of the target I to be retrieved, as shown in the following formula:

[0075] In the formula, is the feature of the target I to be retrieved; The features of the high-frequency image of the target to be retrieved are extracted; It is the feature of the low-frequency image Markov graph of the extracted target to be retrieved.

[0076] Binarize the features of the target I to be retrieved to obtain the binary features of the target I to be retrieved , the binarization process is shown as follows:

[0077] Fusion of high frequency target database and Markov graph of low-frequency target database The features of constitute the features of the target database Q, and the features of the target database Q are shown in the following formula:

[0078] Where FQ is the feature of the target database Q; Represents the Nth feature of the target database; is the Nth feature of the high-frequency target database; is the Nth feature of the Markov graph of the low-frequency target database.

[0079] The features of the target database Q are binarized to obtain the binary features BFQ of the target database Q. The binarization process is shown in the following formula:

[0080] In the formula, is the binary feature of the target database Q; is the Nth feature of the target database after binarization.

[0081] Step S4, based on the binarized features, respectively calculate the similarity between the target to be retrieved and each target in the target database to obtain a similarity measure, and obtain the final retrieval result by sorting the similarity measures.

[0082] Based on the binary features of the target I to be retrieved and the binary feature BFQ of the target database Q, respectively calculate the similarity between the target to be retrieved I and each target in the target database Q, and obtain the similarity measure , as follows:

[0083] Where D is the similarity measure; is the feature of the target to be retrieved after binary processing; is the Nth feature of the target database after binarization.

[0084] Similarity measure Sort and get the final search results.

[0085] Embodiment 2 This embodiment discloses a coal mine underground target retrieval system; like Figure 2 As shown, a coal mine underground target retrieval system includes: The target database construction module is configured to: construct a target database based on the target image of the coal mine underground, wherein the target database includes a high-frequency target database and a low-frequency target database; The image analysis module is configured to: obtain the target to be retrieved and perform Fourier transform to obtain the high-frequency image and the low-frequency image of the target to be retrieved; based on the Markov transition field, obtain the Markov graph of the low-frequency image of the target to be retrieved and the Markov graph of the low-frequency target database; The feature extraction and binarization processing module is configured to: use a feature extractor to extract the features of the target to be retrieved and the features of the target database respectively, and use Hill diversity to perform binarization processing on the extracted features; wherein the features of the target to be retrieved are composed of Markov graph features of the high-frequency image and the low-frequency image of the target to be retrieved; the features of the target database are composed of Markov graph features of the high-frequency target database and the low-frequency target database; The retrieval module is configured to: calculate the similarity between the target to be retrieved and each target in the target database based on the binarized features, obtain similarity metrics, and obtain the final retrieval result by sorting the similarity metrics.

[0086] Embodiment 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0087] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a coal mine underground target retrieval method as described in Example 1.

[0088] Embodiment 4 The purpose of this embodiment is to provide an electronic device.

[0089] An electronic device comprises a memory, a processor and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in a method for underground target retrieval in a coal mine as described in Example 1 are implemented.

[0090] The steps involved in the apparatuses of the above embodiments 2, 3 and 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0091] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0092] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A target retrieval method for underground coal mines, characterized in that: include: Building a target database of targets to be retrieved in coal mines based on the targets to be retrieved, wherein the target database includes a high-frequency target database and a low-frequency target database; Acquire a high-frequency image and a low-frequency image of the target to be retrieved, input the low-frequency image of the target to be retrieved and the low-frequency target database into the Markov transformation field respectively, and acquire a Markov graph of the low-frequency image of the target to be retrieved and a Markov graph of the low-frequency target database; Extracting the features of the target to be retrieved and the features of the target database respectively, and performing binarization processing on the extracted features; wherein the features of the target to be retrieved are composed of the high-frequency image features of the target to be retrieved and the low-frequency image Markov graph features of the target to be retrieved; the features of the target database are composed of the high-frequency target database features and the Markov graph features of the low-frequency target database; Based on the binarized features, the similarity between the target to be retrieved and each target in the target database is calculated respectively to obtain a similarity measure, and the final retrieval result is obtained by sorting the similarity measures.

2. A coal mine underground target retrieval method as claimed in claim 1, characterized in that: The process of obtaining the high-frequency target database and the low-frequency target database is as follows: by traversing the target database, obtaining the high-frequency image and the low-frequency image of each target data in the target database corresponding to the target to be retrieved, and constructing the high-frequency target database and the low-frequency target database respectively.

3. A coal mine underground target retrieval method as claimed in claim 1, characterized in that: By performing Fourier transform on the target to be retrieved, a high-frequency image and a low-frequency image of the target to be retrieved are obtained; specifically: Perform Fourier transform on the target to be retrieved to obtain a spectrum image of the target to be retrieved; the spectrum image is as follows: In the formula, Represents the gray value of the pixel at the coordinate (m,n) in the target to be retrieved, represents the corresponding spectrum; M is the width of the target to be retrieved; N is the height of the target to be retrieved; u is the frequency domain horizontal coordinate; v is the frequency domain vertical coordinate; Center the spectrum image and move the zero-frequency part to the center of the spectrum; A Gaussian high-pass filter is used to perform high-pass filtering to obtain a high-pass filtering spectrum diagram; a Gaussian low-pass filter is used to perform low-pass filtering to obtain a low-pass filtering spectrum diagram, and the zero-frequency domain part is moved back to the original position; The high-pass filter spectrum diagram and the low-pass filter spectrum diagram are respectively subjected to inverse Fourier transform to generate a high-frequency image and a low-frequency image of the target to be retrieved.

4. A coal mine underground target retrieval method as claimed in claim 1, characterized in that: The process of inputting the low-frequency image of the target to be retrieved into the Markov transformation field and obtaining the Markov graph of the low-frequency image of the target to be retrieved is: Filling the low-frequency image of the target to be retrieved, obtaining the filled image and calculating the size of the filled image; Segmenting the filled image to obtain image blocks; performing statistical processing on each image block to obtain a final description vector of the image block; The final description vector is converted into a Markov field by using a Markov transformation field, and the Markov field is scaled to a size equal to the low-frequency image of the target to be retrieved, thereby obtaining a Markov graph of the low-frequency image of the target to be retrieved.

5. A coal mine underground target retrieval method as claimed in claim 1, characterized in that: The features of the target to be retrieved are composed of the high-frequency image features of the target to be retrieved and the low-frequency image Markov graph features of the target to be retrieved. The features of the target to be retrieved are as follows: In the formula, is the feature of the target to be retrieved; is a high-frequency image feature; is the low-frequency image Markov graph feature of the target to be retrieved; The characteristics of the target database are composed of the high-frequency target database characteristics and the Markov graph characteristics of the low-frequency target database. The characteristics of the target database are as follows: In the formula, is the characteristic of the target database; Represents the Nth feature of the target database; is the Nth feature of the high-frequency target database; is the Nth feature of the Markov graph of the low-frequency target database.

6. A method for underground coal mine target retrieval according to claim 1, characterized in that: The process of binarizing the extracted features is as follows: Dividing the features of the target to be retrieved and the features of the target database, and finding the neighborhood of each feature component; The Hill diversity is calculated based on the neighborhood value, and the formula of the Hill diversity is: In the formula, is the characteristic component; N( )for Neighborhood of; is the characteristic component Hill diversity value; for Neighborhood N( ) in the element; Based on Hill diversity, the features of the target to be retrieved and the features of the target database are binarized to obtain the binarized features.

7. A coal mine underground target retrieval method as claimed in claim 1, characterized in that: The similarity measure is: Where D is the similarity measure; is the feature of the target to be retrieved after binary processing; is the Nth feature of the target database after binarization.

8. A target retrieval system for underground coal mines, characterized in that: include: The target database construction module is configured to: construct a target database of targets to be retrieved in coal mines based on the targets to be retrieved, wherein the target database includes a high-frequency target database and a low-frequency target database; The image analysis module is configured to: obtain a high-frequency image and a low-frequency image of the target to be retrieved, input the low-frequency image of the target to be retrieved and the low-frequency target database into the Markov transformation field respectively, and obtain a Markov graph of the low-frequency image of the target to be retrieved and a Markov graph of the low-frequency target database; The feature extraction and binarization processing module is configured to: extract the features of the target to be retrieved and the features of the target database respectively, and perform binarization processing on the extracted features; wherein the features of the target to be retrieved are composed of the high-frequency image features of the target to be retrieved and the low-frequency image Markov graph features of the target to be retrieved; the features of the target database are composed of the high-frequency target database features and the Markov graph features of the low-frequency target database; The retrieval module is configured to: calculate the similarity between the target to be retrieved and each target in the target database based on the binarized features, obtain similarity metrics, and obtain the final retrieval result by sorting the similarity metrics.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps of a coal mine underground target retrieval method as described in any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the coal mine underground target retrieval method as described in any one of claims 1-7 are implemented.

Images