Rock mineral content detection system and method for geological exploration

Abstract

The invention discloses a rock mineral content detection system and a rock mineral content detection method for geological survey, which are characterized in that after image features and spectral features of rocks and ores are extracted, internal generality of the image features and the spectral features of the rocks and ores is mined by using a data processing technology based on deep learning, and the internal generality of the image features and the spectral features of the rocks and ores is used as a mask; according to the method, unique information in rock and ore image features and spectral features is emphasized, and the general information and the unique information of the rock and ore image features and the spectral features are comprehensively utilized through compensation type interactive fusion, so that accurate identification of rock and ore categories is realized. In this way, the problem of information loss in the multi-modal feature fusion process can be effectively solved, and therefore the accuracy of rock and ore classification and the calculation precision of the mineral content are improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection, and more specifically, to a rock mineral content detection system and method for geological survey. Background Art

[0002] In geological research and mineral resource development, rock mineral content analysis is of great significance for understanding geological structures and evaluating the value of mineral resources. Traditional rock mineral content detection methods mainly rely on laboratory chemical analysis and physical testing. Although these methods are highly accurate, they are time-consuming and costly, and cannot meet the needs of rapid on-site detection.

[0003] In recent years, with the development of remote sensing technology and computer vision, rock mineral content detection technology based on spectral analysis has gradually become a research hotspot. For example, the invention patent with publication number CN115656053A discloses a rock mineral content testing method and system, which first splices the spectral information and image information of the tunnel surrounding rock into a 3D tunnel image spectrum, and extracts the characteristic sensitive band, the image feature vector and spectral feature vector of the rock and mineral from it, and then performs end-member unmixing on the spectral information in the 3D tunnel image spectrum based on the characteristic sensitive wave, and identifies the rock and mineral category based on the rock and mineral image feature vector and spectral feature vector, and calibrates the mineral category obtained by end-member unmixing, so as to calculate the mineral content based on the mineral lithology and pixel spatial distribution, and perform visual display.

[0004] However, when the above scheme uses rock and mineral image features and spectral features to classify rocks and minerals, the rock and mineral category matching and identification is performed through the relationship between the spectral fusion information stored in the pre-built tunnel site area atlas library and the rocks and minerals. This method may encounter the problem of incomplete or untimely update of the atlas library in practical applications, thus affecting the accuracy and real-time performance of the detection. In addition, since rock and mineral image features and spectral features have different physical meanings and expressions, traditional fusion methods such as feature splicing and weighted averaging are difficult to fully explore the complementary information between the two, and there are problems such as information loss, which limits the accuracy and reliability of the detection results.

[0005] Therefore, an optimized rock mineral content detection system and method for geological survey is desired. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a rock mineral content detection system and method for geological survey, which, after extracting the image features and spectral features of rocks and minerals, mines out the inherent commonalities of the rock and mineral image features and spectral features by using data processing technology based on deep learning, and uses the common features of the two as a mask to emphasize the unique information in the rock and mineral image features and spectral features, and through compensatory interactive fusion, comprehensively utilizes the common information and unique information of the rock and mineral image features and spectral features, so as to achieve accurate identification of rock and mineral categories. In this way, the problem of information loss in the process of multimodal feature fusion can be effectively solved, thereby improving the accuracy of rock and mineral classification and the calculation accuracy of mineral content.

[0007] According to one aspect of the present application, a method for detecting rock and mineral content for geological survey is provided, which includes: obtaining spectral information and image information of rock and mineral, and splicing the spectral information and image information of the rock and mineral into a 3D rock and mineral image spectrum map; extracting characteristic sensitive bands, rock and mineral image feature vectors and spectral feature vectors from the 3D rock and mineral image spectrum map; based on the characteristic sensitive bands, performing end-member unmixing on the spectral information in the 3D rock and mineral image spectrum map to obtain an end-member unmixing result; determining the rock and mineral category based on the rock and mineral image feature vector and the spectral feature vector; calibrating the mineral category in the end-member unmixing result based on the rock and mineral category; performing weighted processing on the weight of each mineral in the end-member unmixing of each pixel in the end-member unmixing result and the pixel points occupied by the mineral to obtain the mineral content, characterized in that determining the rock and mineral category based on the rock and mineral image feature vector and the spectral feature vector includes:

[0008] Performing cross-modal interactive fusion on the rock and mineral image feature vector and the spectrum feature vector to obtain a rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector;

[0009] Based on the rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector, a rock and mineral category recognition result is generated.

[0010] According to another aspect of the present application, a rock mineral content detection system for geological survey is provided, which comprises:

[0011] An information acquisition module, used to acquire spectral information and image information of rock minerals, and splice the spectral information and image information of the rock minerals into a 3D rock mineral image spectrum map;

[0012] A feature extraction module, used to extract feature sensitive bands, rock and mineral image feature vectors and spectral feature vectors from the 3D rock and mineral image spectrum map;

[0013] An endmember unmixing module, used for performing endmember unmixing on the spectral information in the 3D rock mineral image spectrum map based on the characteristic sensitive band to obtain an endmember unmixing result;

[0014] A cross-modal interactive fusion module, used for performing cross-modal interactive fusion on the rock and mineral image feature vector and the spectrum feature vector to obtain a rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector;

[0015] A recognition result generation module, used to generate a recognition result of a rock and mineral category based on the rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector;

[0016] A calibration module, used for calibrating the mineral category in the end member unmixing result based on the rock and mineral category;

[0017] The mineral content detection result generation module is used to perform weighted processing on the weight of each mineral in each pixel end member unmixing in the end member unmixing result and the pixel points occupied by the mineral to obtain the mineral content.

[0018] Compared with the prior art, the present application provides a rock mineral content detection system and method for geological survey. After extracting the image features and spectral features of rocks and minerals, the system uses data processing technology based on deep learning to mine the inherent commonality of rock and mineral image features and spectral features, and uses the common features of the two as a mask to emphasize the unique information in the rock and mineral image features and spectral features. Through compensatory interactive fusion, the common information and unique information of the rock and mineral image features and spectral features are comprehensively utilized to achieve accurate identification of rock and mineral categories. In this way, the problem of information loss in the process of multimodal feature fusion can be effectively solved, thereby improving the accuracy of rock and mineral classification and the calculation accuracy of mineral content. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0020] Figure 1 This is a flow chart of a method for detecting rock mineral content for geological survey according to an embodiment of the present application;

[0021] Figure 2 A data flow diagram of a method for detecting rock mineral content for geological survey according to an embodiment of the present application;

[0022] Figure 3It is a flowchart of sub-step S4 of a method for detecting rock mineral content for geological survey according to an embodiment of the present application;

[0023] Figure 4 It is a block diagram of a rock mineral content detection system for geological survey according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0025] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0026] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0027] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0028] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0029] Patent CN115656053A proposes a rock mineral content testing method, which includes: obtaining spectral information and image information of rock minerals, and splicing the spectral information and image information of the rock minerals into a 3D rock mineral image spectrum; extracting characteristic sensitive bands, rock mineral image feature vectors and spectral feature vectors from the 3D rock mineral image spectrum; based on the characteristic sensitive bands, performing end-member unmixing on the spectral information in the 3D rock mineral image spectrum to obtain an end-member unmixing result; based on the rock mineral image feature vector and the spectral feature vector, determining the rock mineral category; based on the rock mineral category, calibrating the mineral category in the end-member unmixing result; and weighting the weight of each mineral in the end-member unmixing of each pixel in the end-member unmixing result with the pixel points occupied by the mineral to obtain the mineral content.

[0030] When the above scheme uses rock and mineral image features and spectral features to classify rocks and minerals, the rock and mineral category matching and identification is performed through the relationship between the spectral fusion information stored in the pre-built tunnel site area atlas library and the rocks and minerals. However, this method may encounter problems such as incomplete or untimely updates of the atlas library in practical applications, thus affecting the accuracy and real-time performance of the detection. In addition, since rock and mineral image features and spectral features have different physical meanings and expressions, traditional fusion methods such as feature splicing and weighted averaging are difficult to fully explore the complementary information between the two, and there are problems such as information loss, which limits the accuracy and reliability of the detection results.

[0031] Based on this, in the technical solution of the present application, a method for detecting rock mineral content for geological survey is proposed. Figure 1 The present invention is a flowchart of a method for detecting rock mineral content for geological survey according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of a method for detecting rock mineral content for geological survey according to an embodiment of the present application. Figure 1 and Figure 2 As shown, a method for detecting rock and mineral content for geological survey according to an embodiment of the present application includes the following steps: S1, acquiring spectral information and image information of rock and mineral, and splicing the spectral information and image information of the rock and mineral into a 3D rock and mineral image spectrum map; S2, extracting characteristic sensitive bands, rock and mineral image feature vectors and spectral feature vectors from the 3D rock and mineral image spectrum map; S3, based on the characteristic sensitive bands, performing end-member unmixing on the spectral information in the 3D rock and mineral image spectrum map to obtain an end-member unmixing result; S4, determining the rock and mineral category based on the rock and mineral image feature vector and the spectral feature vector; S5, calibrating the mineral category in the end-member unmixing result based on the rock and mineral category; S6, weighting the weight of each mineral in the end-member unmixing of each pixel in the end-member unmixing result with the pixel points occupied by the mineral to obtain the mineral content.

[0032] In particular, the S1 obtains the spectral information and image information of the rock minerals, and splices the spectral information and image information of the rock minerals into a 3D rock mineral image spectrogram. It should be understood that through the spectral information and image information of the rock minerals, the characteristics of the rock minerals can be fully understood from multiple dimensions. Spectral information reflects the chemical composition and physical properties of minerals, while image information provides the morphology and spatial distribution of minerals. Splicing the spectral information and image information into a 3D image spectrogram realizes the integration of multimodal data. In particular, the 3D image spectrogram not only contains two-dimensional spatial information, but also spectral information, which can more intuitively display the internal structure and composition distribution of rock minerals.

[0033] In particular, the step S2 extracts characteristic sensitive bands, rock and mineral image feature vectors and spectral feature vectors from the 3D rock and mineral image spectrum graph, wherein the characteristic sensitive bands are helpful for identifying specific minerals.

[0034] In particular, S3, based on the characteristic sensitive band, performs endmember unmixing on the spectral information in the 3D rock mineral image spectrum to obtain an endmember unmixing result. Among them, endmember unmixing is to separate pure spectral signals (endmembers) from complex spectral data and estimate the proportion of each endmember in the mixed pixel. This process helps to remove noise and interference and extract pure mineral information. Through endmember unmixing, the type and distribution of minerals can be preliminarily determined, providing a basis for subsequent accurate classification.

[0035] In particular, the S4 determines the rock and mineral category based on the rock and mineral image feature vector and the spectral feature vector. Considering that in geological surveys, rock and mineral images intuitively display the appearance characteristics of rocks and minerals by capturing visual information such as the morphology, structure, and texture of geological bodies, while spectral technology reveals the chemical composition and mineralogical properties of rocks and minerals by measuring the spectral characteristics of rock and mineral absorption and emission in a specific wavelength range. Both can reflect the category information of rocks and minerals from different angles and have certain commonalities. However, due to the differences in the physical meaning and expression form of rock and mineral image features and spectral features, both also carry certain unique information. Therefore, in order to solve the problem of information loss in multimodal feature fusion and make full use of the complementary characteristics of rock and mineral image features and spectral features, this application proposes a cross-modal interactive compensation fusion method, which extracts the common feature representation of rock and mineral image features and spectral features, and uses this as a mask to compensate and enhance the unique information of each of the two, thereby effectively retaining the common information of rock and mineral image features and spectral features and the unique information in each feature source, improving the effect of feature fusion, and helping to improve the recognition accuracy of rock and mineral categories. Specifically, in a specific example of the present application, Figure 3As shown, the S4 includes: S41, performing cross-modal interactive fusion of the rock and mineral image feature vector and the spectral feature vector to obtain a rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector; S42, generating a rock and mineral category recognition result based on the rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector.

[0036] Specifically, the S41 performs cross-modal interactive fusion on the rock and mineral image feature vector and the spectral feature vector to obtain a rock and mineral image-spectral cross-modal interactive compensation fusion representation vector. In an embodiment of the present application, first, the common features between the rock and mineral image feature vector and the spectral feature vector are extracted to obtain a common feature representation vector between rock and mineral image and spectral features; that is, the rock and mineral image feature vector and the spectral feature vector are input into a common feature extraction network, and a neural network architecture is used to learn the common pattern between the two, and to mine the intrinsic connection between the two, so as to generate a common feature representation vector between rock and mineral image and spectral features. In a specific example, the specific steps of extracting the common features between the rock and mineral image feature vector and the spectral feature vector include: inputting the rock and mineral image feature vector and the spectral feature vector into a common feature extraction network to obtain a common feature representation vector between the rock and mineral image and the spectral feature, wherein the common feature extraction network performs point addition fusion on the rock and mineral image feature vector and the spectral feature vector, and then uses a neural network layer based on a tanh function to extract common features to obtain a common feature representation vector between the rock and mineral image and the spectral feature. Then, based on the common feature representation vector between the rock and mineral image and the spectral feature, guide the rock and mineral image feature vector and the spectral feature vector to perform feature compensation interactive fusion to obtain the rock and mineral image-spectral cross-modal interactive compensation fusion representation vector. In a specific example, first, based on the specific features of the rock and mineral image feature vector and the spectral feature vector relative to the common feature representation vector between the rock and mineral image and the spectral feature, feature compensation modulation is performed on the rock and mineral image feature vector and the spectral feature vector to obtain a rock and mineral image fine-grained feature compensation vector and a spectral fine-grained feature compensation vector; more specifically, the rock and mineral image feature vector, the spectral feature vector and the generated rock and mineral image-spectral feature common feature representation vector are input into a vector probabilization unit based on the Sigmaid function for normalization processing to ensure that the value range of the feature vector is between 0 and 1, so as to better measure the importance of the feature. Then, the specific features of the probabilistic rock and mineral image feature vector and the probabilistic spectral feature vector relative to the probabilistic rock and mineral image-spectral feature common feature representation vector are calculated respectively, and by learning the feature differences between the two, the unique feature information of the rock and mineral image feature and the spectral feature in addition to the common features is mined, thereby generating a rock and mineral image specific feature compensation representation vector and a spectral specific feature compensation representation vector. Subsequently, the rock and mineral image specific feature compensation representation vector and the spectrum specific feature compensation representation vector are used as weights to compensate and strengthen the original rock and mineral image feature vector and spectrum feature vector by means of position point multiplication, so as to ensure that the unique information of the rock and mineral image and spectrum features is retained as much as possible during the fusion process.Furthermore, the rock and mineral image fine-grained feature compensation vector, the spectral fine-grained feature compensation vector and the rock and mineral image-spectral feature common feature representation vector are fused to obtain the rock and mineral image-spectral feature common feature representation vector. That is, by fusing the compensated and enhanced rock and mineral image feature vector, spectral feature vector and rock and mineral image-spectral feature common feature representation vector, the common information of rock and mineral image features and spectral features and the unique information in their respective feature sources are comprehensively considered to achieve comprehensive capture of rock and mineral category features.

[0037] The process of performing feature compensation modulation on the rock and mineral image feature vector and the spectral feature vector based on the specific features of the rock and mineral image feature vector and the spectral feature vector relative to the common feature representation vector between the rock and mineral image and the spectral feature to obtain the rock and mineral image fine-grained feature compensation vector and the spectral fine-grained feature compensation vector comprises: inputting the rock and mineral image feature vector, the spectral feature vector and the common feature representation vector between the rock and mineral image and the spectral feature into the sigmo The vector probabilization unit of the id function is used to obtain a probabilistic rock and mineral image feature vector, a probabilistic spectral feature vector and a probabilistic rock and mineral image-spectral feature common feature representation vector; the specific features of the probabilistic rock and mineral image feature vector relative to the probabilistic rock and mineral image-spectral feature common feature representation vector are calculated to obtain a rock and mineral image specific feature compensation representation vector; the specific features of the probabilistic spectral feature vector relative to the probabilistic rock and mineral image-spectral feature common feature representation vector are calculated to obtain a spectral specific feature compensation representation vector; the rock and mineral image specific feature compensation representation vector and the rock and mineral image feature vector are input into a fine-grained compensation module to obtain the rock and mineral image fine-grained feature compensation vector; the spectral specific feature compensation representation vector and the spectral feature vector are input into the fine-grained compensation module to obtain the spectral fine-grained feature compensation vector. More specifically, the specific features of the probabilistic rock and mineral image feature vector relative to the probabilistic rock and mineral image-spectral feature common feature representation vector are calculated to obtain the rock and mineral image specific feature compensation representation vector, including: calculating the dot division vector between the probabilistic rock and mineral image feature vector and the probabilistic rock and mineral image-spectral feature common feature representation vector, and calculating the base-two logarithmic value of the absolute value of each eigenvalue in the dot division vector to obtain the rock and mineral image specific feature vector; calculating the dot product vector between the rock and mineral image specific feature vector and the probabilistic rock and mineral image feature vector, and calculating the exponential function value with base e as the exponent and each eigenvalue of the dot product vector as the exponent to obtain the indexed rock and mineral image specific feature representation vector; inputting the indexed rock and mineral image specific feature representation vector into the softmax function for normalization to obtain the rock and mineral image specific feature compensation representation vector. More specifically, the rock and mineral image specific feature compensation representation vector and the rock and mineral image feature vector are input into a fine-grained compensation module to obtain the rock and mineral image fine-grained feature compensation vector, including: calculating the Hadamard product between the rock and mineral image specific feature compensation representation vector and the rock and mineral image feature vector to obtain the rock and mineral image fine-grained feature compensation vector.

[0038] More specifically, the process of fusing the rock and mineral image fine-grained feature compensation vector, the spectral fine-grained feature compensation vector and the rock and mineral image-spectral feature common feature representation vector to obtain the rock and mineral image-spectral feature common feature representation vector includes: cascading the rock and mineral image fine-grained feature compensation vector, the spectral fine-grained feature compensation vector and the rock and mineral image-spectral feature common feature representation vector to obtain the rock and mineral image-spectral feature common feature representation vector.

[0039] In summary, in the above embodiment, the rock and mineral image feature vector and the spectral feature vector are cross-modally interactively fused to obtain a rock and mineral image-spectral cross-modal interactive compensation fusion representation vector, including: processing the rock and mineral image feature vector and the spectral feature vector with the following feature compensation interaction formula to obtain a rock and mineral image-spectral cross-modal interactive compensation fusion representation vector, wherein the feature compensation interaction formula is:

[0040] V J =tanh(W VT (V1⊕V2)+b VT )

[0041] V a =sigmoid(V1)

[0042] V b =sigmoid(V2)

[0043] V c =sigmoid(V J )

[0044]

[0045] v 1a =V1⊙v a '

[0046] v 2b =V2⊙v b '

[0047] v f =concat[V J ;v 1a ;v 2b ]

[0048] Wherein, V1 represents the rock and mineral image feature vector, V2 represents the spectral feature vector, V J represents the common feature representation vector between rock and mineral image and spectral features, W VT and b VTRespectively represent the weight parameter matrix and bias term, tanh represents the hyperbolic tangent function, sigmoid represents the sigmoid function, ⊕ represents adding by position point, V a represents the probabilistic rock and mineral image feature vector, V b represents the probabilistic spectral feature vector, V c represents the common feature representation vector between probabilistic rock and mineral image-spectral features, V ai 、V bi and V ci denote the i-th eigenvalue of the probabilistic rock and mineral image feature vector, the i-th eigenvalue of the probabilistic spectral feature vector and the i-th eigenvalue of the common feature representation vector between the probabilistic rock and mineral image and spectral features, log(·) denotes a logarithmic function with base 2, exp(·) denotes a natural exponential function, softmax denotes a normalized exponential function, and v a ′ and v b ′ represents the compensation representation vector of rock and mineral image specific features and the compensation representation vector of spectrum specific features, ⊙ represents the Hadamard product, v 1a and v 2b They represent the fine-grained feature compensation vector of rock and mineral images and the fine-grained feature compensation vector of spectra, respectively. concat[·;·;·] represents the cascade operation. f Represents the common feature representation vector between the rock and mineral image-spectral features.

[0049] Specifically, the S42 generates a recognition result of the rock and mineral category based on the rock and mineral image-spectrum cross-modal interaction compensation fusion representation vector. In a specific example of the present application, the rock and mineral image-spectrum cross-modal interaction compensation fusion representation vector is input into a rock and mineral recognition module based on a classifier to obtain the recognition result, and the recognition result is the rock and mineral category. That is, the rock and mineral category is judged based on the rock and mineral image-spectrum cross-modal interaction features in the rock and mineral image-spectrum cross-modal interaction compensation fusion representation vector. In this way, the problem of information loss in the process of multimodal feature fusion can be effectively solved, thereby improving the accuracy of rock and mineral classification and the calculation accuracy of mineral content.

[0050] In particular, the S5 calibrates the mineral category in the end member unmixing result based on the rock and mineral category. It should be understood that by calibrating the end member unmixing result through known rock and mineral category information, it is possible to correct the misclassification that may occur in the end member unmixing process and improve the accuracy of mineral classification. The calibrated mineral category information can be used to recalculate the proportion of each mineral in each pixel, thereby optimizing the calculation result of the mineral content.

[0051] In particular, in S6, the weight of each mineral in each pixel end member unmixing in the end member unmixing result is weighted with the pixel points occupied by the mineral to obtain the content of the mineral. Here, the weight of each mineral in each pixel end member unmixing in the end member unmixing result is weighted with the pixel points occupied by the mineral to finally obtain the content of the mineral. This process ensures the accuracy and reliability of the mineral content calculation. By calculating the mineral content, quantitative analysis of rock minerals can be achieved, providing important data support for geological surveys.

[0052] In summary, a method for detecting rock and mineral content for geological survey according to an embodiment of the present application is explained, which, after extracting the image features and spectral features of rocks and minerals, mines out the inherent commonalities of the image features and spectral features of rocks and minerals by using data processing technology based on deep learning, and uses the common features of the two as a mask to emphasize the unique information in the image features and spectral features of rocks and minerals, and through compensatory interactive fusion, comprehensively utilizes the common information and unique information of the image features and spectral features of rocks and minerals, so as to achieve accurate identification of rock and mineral categories. In this way, the problem of information loss in the process of multimodal feature fusion can be effectively solved, thereby improving the accuracy of rock and mineral classification and the calculation accuracy of mineral content.

[0053] Furthermore, a rock mineral content detection system for geological survey is also provided.

[0054] Figure 4 FIG. 1 is a block diagram of a rock mineral content detection system for geological survey according to an embodiment of the present application. Figure 4 As shown, a rock mineral content detection system 300 for geological survey according to an embodiment of the present application includes: an information acquisition module 310, used to acquire spectral information and image information of rock minerals, and splice the spectral information and image information of the rock minerals into a 3D rock mineral image spectrum; a feature extraction module 320, used to extract characteristic sensitive bands, rock mineral image feature vectors and spectral feature vectors from the 3D rock mineral image spectrum; an end member unmixing module 330, used to perform end member unmixing on the spectral information in the 3D rock mineral image spectrum based on the characteristic sensitive bands to obtain an end member unmixing result; a cross-modal interactive fusion module 330, used to extract the characteristic sensitive bands, rock mineral image feature vectors and spectral feature vectors from the 3D rock mineral image spectrum ... Block 340 is used to perform cross-modal interactive fusion on the rock and mineral image feature vector and the spectral feature vector to obtain a rock and mineral image-spectral cross-modal interactive compensation fusion representation vector; an identification result generation module 350 is used to generate a rock and mineral category recognition result based on the rock and mineral image-spectral cross-modal interactive compensation fusion representation vector; a calibration module 360 ​​is used to calibrate the mineral category in the end-member unmixing result based on the rock and mineral category; a mineral content detection result generation module 370 is used to weight the weight of each mineral in each pixel end-member unmixing in the end-member unmixing result and the pixel points occupied by the mineral to obtain the mineral content.

[0055] As described above, a rock mineral content detection system 300 for geological survey according to an embodiment of the present application can be implemented in various wireless terminals, such as a server with a rock mineral content detection algorithm for geological survey. In a possible implementation, a rock mineral content detection system 300 for geological survey according to an embodiment of the present application can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the rock mineral content detection system 300 for geological survey can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the rock mineral content detection system 300 for geological survey can also be one of the many hardware modules of the wireless terminal.

[0056] Alternatively, in another example, the rock mineral content detection system 300 for geological survey and the wireless terminal may also be separate devices, and the rock mineral content detection system 300 for geological survey may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0057] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for detecting rock mineral content for geological survey, comprising: Acquire spectral information and image information of rock minerals, and splice the spectral information and image information of the rock minerals into a 3D rock mineral image spectrum; extract characteristic sensitive bands, rock mineral image feature vectors and spectral feature vectors from the 3D rock mineral image spectrum; based on the characteristic sensitive bands, perform endmember unmixing on the spectral information in the 3D rock mineral image spectrum to obtain an endmember unmixing result; determine the rock mineral category based on the rock mineral image feature vector and the spectral feature vector; calibrate the mineral category in the endmember unmixing result based on the rock mineral category; perform weighted processing on the weight of each mineral in the endmember unmixing of each pixel in the endmember unmixing result and the pixel points occupied by the mineral to obtain the mineral content, characterized in that the rock mineral category is determined based on the rock mineral image feature vector and the spectral feature vector, including: Performing cross-modal interactive fusion on the rock and mineral image feature vector and the spectrum feature vector to obtain a rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector; Generate a rock and mineral category recognition result based on the rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector; Among them, the rock and mineral image feature vector and the spectral feature vector are cross-modal interactively fused, including: extracting common features between the rock and mineral image feature vector and the spectral feature vector to obtain a common feature representation vector between rock and mineral image and spectral features; based on the common feature representation vector between rock and mineral image and spectral features, guiding the rock and mineral image feature vector and the spectral feature vector to perform feature compensation interactive fusion to obtain the rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector.

2. A method for detecting rock mineral content for geological survey according to claim 1, characterized in that: Extracting common features between the rock and mineral image feature vector and the spectral feature vector to obtain a common feature representation vector between the rock and mineral image and the spectral feature, including: The rock and mineral image feature vector and the spectral feature vector are input into a common feature extraction network to obtain a common feature representation vector between the rock and mineral image and the spectral features, wherein the common feature extraction network performs point-wise fusion on the rock and mineral image feature vector and the spectral feature vector, and then uses a neural network layer based on a tanh function to extract common features to obtain a common feature representation vector between the rock and mineral image and the spectral features.

3. A method for detecting rock mineral content for geological survey according to claim 2, characterized in that: Based on the common feature representation vector between the rock and mineral image and the spectral feature, guiding the rock and mineral image feature vector and the spectral feature vector to perform feature compensation interactive fusion to obtain the rock and mineral image-spectral cross-modal interactive compensation fusion representation vector, including: Based on the specific features of the rock and mineral image feature vector and the spectral feature vector relative to the common feature representation vector between the rock and mineral image and the spectral feature vector, feature compensation modulation is performed on the rock and mineral image feature vector and the spectral feature vector to obtain a rock and mineral image fine-grained feature compensation vector and a spectral fine-grained feature compensation vector; The rock and mineral image fine-grained feature compensation vector, the spectral fine-grained feature compensation vector and the rock and mineral image-spectral feature common feature representation vector are fused to obtain the rock and mineral image-spectral feature common feature representation vector.

4. A method for detecting rock mineral content for geological survey according to claim 3, characterized in that: Based on the specific features of the rock and mineral image feature vector and the spectral feature vector relative to the common feature representation vector between the rock and mineral image and the spectral feature vector, feature compensation modulation is performed on the rock and mineral image feature vector and the spectral feature vector to obtain a rock and mineral image fine-grained feature compensation vector and a spectral fine-grained feature compensation vector, including: Input the rock and mineral image feature vector, the spectral feature vector and the common feature representation vector between the rock and mineral image and the spectral feature into a vector probabilization unit based on the Sigmoid function to obtain a probabilistic rock and mineral image feature vector, a probabilistic spectral feature vector and a probabilistic rock and mineral image and the common feature representation vector between the spectral feature; Calculate the specific features of the probabilistic rock and mineral image feature vector relative to the common feature representation vector between the probabilistic rock and mineral image and the spectral feature to obtain a rock and mineral image specific feature compensation representation vector; Calculate the specific features of the probabilistic spectral feature vector relative to the common feature representation vector between the probabilistic rock and mineral image and spectral features to obtain a spectral specific feature compensation representation vector; Inputting the rock and mineral image specific feature compensation representation vector and the rock and mineral image feature vector into a fine-grained compensation module to obtain the rock and mineral image fine-grained feature compensation vector; The spectrum specific feature compensation representation vector and the spectrum feature vector are input into the fine-grained compensation module to obtain the spectrum fine-grained feature compensation vector.

5. A method for detecting rock mineral content for geological survey according to claim 4, characterized in that: Calculating the specific features of the probabilistic rock and mineral image feature vector relative to the common feature representation vector between the probabilistic rock and mineral image and the spectral feature to obtain a rock and mineral image specific feature compensation representation vector, including: Calculate the dot division vector between the probabilistic rock and mineral image feature vector and the common feature representation vector between the probabilistic rock and mineral image and the spectral feature, and calculate the base-two logarithm of the absolute value of each eigenvalue in the dot division vector to obtain a rock and mineral image specific feature vector; Calculate the dot product vector between the rock and mineral image specific feature vector and the probabilistic rock and mineral image feature vector, and calculate an exponential function value with e as the base and each eigenvalue of the dot product vector as the exponent to obtain an indexed rock and mineral image specific feature representation vector; The indexed rock and mineral image specific feature representation vector is input into a softmax function for normalization processing to obtain the rock and mineral image specific feature compensation representation vector.

6. A method for detecting rock mineral content for geological survey according to claim 5, characterized in that: Inputting the rock and mineral image specific feature compensation representation vector and the rock and mineral image feature vector into a fine-grained compensation module to obtain the rock and mineral image fine-grained feature compensation vector, including: The Hadamard product between the rock and mineral image specific feature compensation representation vector and the rock and mineral image feature vector is calculated to obtain the rock and mineral image fine-grained feature compensation vector.

7. A method for detecting rock mineral content for geological survey according to claim 6, characterized in that: The method of fusing the rock and mineral image fine-grained feature compensation vector, the spectral fine-grained feature compensation vector and the rock and mineral image-spectral feature common feature representation vector to obtain the rock and mineral image-spectral feature common feature representation vector comprises: The rock and mineral image fine-grained feature compensation vector, the spectral fine-grained feature compensation vector and the rock and mineral image-spectral feature common feature representation vector are cascaded to obtain the rock and mineral image-spectral feature common feature representation vector.

8. A method for detecting rock mineral content for geological survey according to claim 7, characterized in that: Based on the rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector, a rock and mineral category recognition result is generated, including: The rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector is input into a rock and mineral recognition module based on a classifier to obtain the recognition result.

9. A rock mineral content detection system for geological survey, characterized in that: include: An information acquisition module, used to acquire spectral information and image information of rock minerals, and splice the spectral information and image information of the rock minerals into a 3D rock mineral image spectrum map; A feature extraction module, used to extract feature sensitive bands, rock and mineral image feature vectors and spectral feature vectors from the 3D rock and mineral image spectrum map; An endmember unmixing module, used for performing endmember unmixing on the spectral information in the 3D rock mineral image spectrum map based on the characteristic sensitive band to obtain an endmember unmixing result; A cross-modal interactive fusion module, used for performing cross-modal interactive fusion on the rock and mineral image feature vector and the spectrum feature vector to obtain a rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector; A recognition result generation module, used to generate a recognition result of a rock and mineral category based on the rock and mineral image-spectrum cross-modal interactive compensation fusion representation vector; A calibration module, used for calibrating the mineral category in the end member unmixing result based on the rock and mineral category; The mineral content detection result generation module is used to perform weighted processing on the weight of each mineral in each pixel end member unmixing in the end member unmixing result and the pixel points occupied by the mineral to obtain the mineral content.

Images