Multidimensional data analysis and calculation method for mine ecological restoration projects based on three-dimensional real scenes

Abstract

The invention relates to the field of image processing, and to a method for analyzing and calculating multidimensional data of a mine ecological restoration project based on three-dimensional real scenes. A video set of preset training healthy mine ecological videos is collected, and after all video sets are cut by a cutting algorithm, the healthy mine ecological grayscale pictures are sorted based on the frequency of the healthy mine ecological grayscale pictures changing in the video set, and an association table of mappings between the healthy mine ecological grayscale pictures and samples of a mine to be reconstructed is generated. A video frame vector of each corresponding video frame in the association table is pre-trained to train vegetation restoration colors, a sequence of healthy mine ecological grayscale pictures of a mine to be reconstructed is input, and a healthy mine ecological mine picture with vegetation restoration colors is output. The invention promotes the exploration of the field of image vegetation restoration colors in this regard. The invention significantly improves the coherence and boundary of colors in a diffusion process by introducing L1-Loss and Edge-Loss.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a multi-dimensional data analysis and measurement method for a mine ecological restoration project based on three-dimensional real scenes. Background Art

[0002] Affected by the severe impact of mining activities, the topsoil is stripped off, the surface morphology is changed, the original surface has lost its original characteristics, and there are also many extreme physical and chemical properties that are harmful to the environment. Its main characteristics are: the topsoil layer is destroyed, the soil lacks moisture, the content of toxic substances increases, etc. Therefore, it is particularly important to restore the ecological environment of the slopes of open-pit mines.

[0003] In order to save riverbank vegetation, restore damaged ecosystems, and curb desertification, water from other rivers or reservoirs can be transported to dry-up rivers through emergency ecological water transfer projects to raise groundwater levels near both sides of the river, saving and restoring natural vegetation on both sides of the river. However, in ecological water transfer projects, it is necessary to collect vegetation indicators such as plant distribution and growth before and after restoration, as well as ecological changes in the surrounding environment such as soil for comparison, which requires large engineering workload and high labor intensity. Summary of the invention

[0004] To achieve the above objectives, this application provides the following technical solutions:

[0005] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a multi-dimensional data analysis and measurement method for mine ecological restoration projects based on three-dimensional real scenes, so as to solve the problem that the reference vegetation restoration colors obtained in the prior art are incomplete or do not meet the ecological requirements of the video frames, so as to provide more reasonable suggestions for mines to be rebuilt.

[0006] To achieve the above-mentioned purpose, the present invention proposes a method for analyzing and calculating multidimensional data of a mine ecological restoration project based on a three-dimensional real scene, comprising the following steps:

[0007] Step S1, through the collected video set of preset training healthy mine ecology videos, after cutting all video sets using a cutting algorithm, sort the healthy mine ecology grayscale pictures based on the frequency of transformation of the healthy mine ecology grayscale pictures in the video set, generate an association table of mappings between the healthy mine ecology grayscale pictures and the mine samples to be restored and reconstructed, and pre-train the video frame vector of each corresponding video frame in the association table;

[0008] Step S2: In the generative model, after completing the video frame cutting for the mine to be reconstructed q to be processed by the user, the repaired mine to be reconstructed sample sequence is obtained by searching the association table, and the video frame vector pre-trained in step S1 is used to convert each repaired mine to be reconstructed sample in the repaired mine to be reconstructed sample sequence into a corresponding video frame vector to obtain the corresponding mine feature vector Hq ;

[0009] Step S3, based on the retrieval algorithm, the video frame healthy mine ecological grayscale image sequence obtained in step S1 is used to retrieve the repaired mine q′ closest to the mine q to be rebuilt from the video collection, and the healthy mine ecological grayscale image corresponding to the closest repaired mine q′ in the video collection. ;

[0010] Step S4, based on the mine feature vector to be reconstructed obtained in step S2, using the transformer convolution head, generates a healthy mine ecological grayscale picture vector related to the video frame ;

[0011] Step S5, the healthy mine ecological grayscale picture retrieved in step S3 is cut by Tokens to obtain a healthy mine ecological grayscale picture sequence, and then the healthy mine ecological grayscale picture sequence is converted into a repaired mine sample sequence to be reconstructed through the association table in step S1, and the Tokens video frame vector corresponding to each repaired mine sample to be reconstructed in the repaired mine sample sequence to be reconstructed is searched to obtain the healthy mine ecological grayscale picture feature vector H a1 , and the mine feature vector H to be reconstructed in step S2 q The model is automatically updated through a mask, and then a dimension correction grid is used to obtain the reference vegetation restoration color vector after the video frame information. ;

[0012] Step S6, the healthy mine ecological grayscale image vector generated in step S4 and the mine feature vector H to be reconstructed in step S2 q Through downsampling processing in turn, the two are combined and then corrected through a dimension grid to obtain the feature vector of the frame information before and after the reference vegetation restoration color vector a2 ;

[0013] Step S7, using the policy gradient algorithm to train the vegetation restoration color is completed, the input is the healthy mine ecological grayscale image sequence of the mine q to be reconstructed, and the output is the healthy mine ecological mine image with the vegetation restoration color completed .

[0014] Furthermore, step S1 further includes:

[0015] Step S100, using a timer to collect training healthy mine ecological videos of historical mines to be rebuilt, and filtering to obtain a video set pair of the historical mines to be rebuilt;

[0016] Step S101, in the pre-training video frame vector stage, the cutting algorithm in the video image analysis is used to cut the ecological grayscale pictures of the mine to be reconstructed and the healthy mine in each sample in the video set, and after removing the duplicates of all the video frames in the video set, an association table is established, and the video frame vectors of the commonly used video frames are obtained by pre-training using the Tokens algorithm;

[0017] Step S4 further comprises:

[0018] Step S401: using the encoder end of the Transformer network, input the mine feature vector H to be reconstructed q , the full text information of the mine to be reconstructed is integrated through the global attention mechanism, and the feature vector H′ of the mine to be reconstructed that integrates the full text information of the video set is further obtained q ;

[0019] Step S402: using the decoder end of the Transformer network, inputting the mine feature vector H′ to be reconstructed that incorporates the full text information of the video set q , sequentially output the corresponding healthy mine ecological grayscale picture vector All video frame vectors in ;

[0020] Furthermore, in step S3, the L1-LOSS and EDGE-LOSS evaluation values ​​are combined as the final similarity judgment to search for the most relevant healthy mine ecological grayscale image sequence from the video collection.

[0021] Further, in step S3,

[0022] During the training phase, grayscale images of healthy mine ecology It is first sent to the convolution head module for dimension correction and downsampling operations to obtain ;

[0023] Where B is the sample size, C1=3 is Channel dimensions, H1=512, W1=512 are The height and width of C2=512 is Channel dimensions, H2=16, W2=16 are The height and width of

[0024] The calculation formulas for the L1-LOSS and EDGE-LOSS evaluation values ​​are as follows:

[0025] L1-Loss is used to calculate Compared with GT, GT refers to Groundtruth, that is, the absolute difference of RGB values ​​at each pixel in the real healthy mine ecological picture;

[0026] The formula is as follows:

[0027]

[0028] in, Indicates the total number of pixels on a picture; Represents the RGB value of the i-th pixel on the GT image; express The RBG value of the i-th pixel;

[0029] Edge-Loss ensures the clarity of the edges of objects in the vegetation restoration color area during color diffusion. The formula is as follows:

[0030]

[0031] Among them, X and Y are and the real GT image, E is the edge feature obtained, H and W are The height and width of the

[0032] Furthermore, in step S3, if the EDGE-LOSS evaluation value calculated between the retrieved mine to be rebuilt and the original mine to be rebuilt is not 0, the EDGE-LOSS evaluation value is adopted; otherwise, the L1-LOSS evaluation value is adopted.

[0033] The invention relates to the field of image processing, and to a method for analyzing and calculating multidimensional data of a mine ecological restoration project based on three-dimensional real scenes. A video set of preset training healthy mine ecological videos is collected, and after all video sets are cut by a cutting algorithm, the healthy mine ecological grayscale pictures are sorted based on the frequency of the healthy mine ecological grayscale pictures changing in the video set, and an association table of mappings between the healthy mine ecological grayscale pictures and samples of a mine to be rebuilt is generated. A video frame vector of each corresponding video frame in the association table is pre-trained to train vegetation restoration colors, and a sequence of healthy mine ecological grayscale pictures of a mine to be rebuilt is input, and a healthy mine ecological mine picture with vegetation restoration colors is output. The invention promotes the exploration of the field of image vegetation restoration colors in this regard. The invention significantly improves the coherence and boundary of colors in a diffusion process by introducing L1-Loss and Edge-Loss. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A workflow diagram of a method for analyzing and calculating multidimensional data of a mine ecological restoration project based on a three-dimensional real scene as claimed in an embodiment of the present application;

[0035] Figure 2 A second workflow diagram of a method for analyzing and calculating multidimensional data of a mine ecological restoration project based on a three-dimensional real scene as claimed in an embodiment of the present application;

[0036] Figure 3 The third workflow diagram of a multi-dimensional data analysis and measurement method for a mine ecological restoration project based on three-dimensional real scenes, which is requested to be protected in an embodiment of the present application. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0038] The terms "first", "second" and "third" in this application are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of this application are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0039] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0040] According to the first embodiment of the present invention, referring to the attached Figure 1 The present invention claims a method for analyzing and calculating multidimensional data of a mine ecological restoration project based on a three-dimensional real scene, comprising the following steps:

[0041] Step S1, through the collected video set of preset training healthy mine ecology videos, after cutting all video sets using a cutting algorithm, sort the healthy mine ecology grayscale pictures based on the frequency of transformation of the healthy mine ecology grayscale pictures in the video set, generate an association table of mappings between the healthy mine ecology grayscale pictures and the mine samples to be restored and reconstructed, and pre-train the video frame vector of each corresponding video frame in the association table;

[0042] Step S2: In the generative model, after completing the video frame cutting for the mine to be reconstructed q to be processed by the user, the repaired mine to be reconstructed sample sequence is obtained by searching the association table, and the video frame vector pre-trained in step S1 is used to convert each repaired mine to be reconstructed sample in the repaired mine to be reconstructed sample sequence into a corresponding video frame vector to obtain the corresponding mine feature vector H q ;

[0043] Step S3, based on the retrieval algorithm, the video frame healthy mine ecological grayscale image sequence obtained in step S1 is used to retrieve the repaired mine q′ closest to the mine q to be rebuilt from the video collection, and the healthy mine ecological grayscale image corresponding to the closest repaired mine q′ in the video collection. ;

[0044] Step S4, based on the mine feature vector to be reconstructed obtained in step S2, using the transformer convolution head, generates a healthy mine ecological grayscale picture vector related to the video frame ;

[0045] Step S5, the healthy mine ecological grayscale picture retrieved in step S3 is cut by Tokens to obtain a healthy mine ecological grayscale picture sequence, and then the healthy mine ecological grayscale picture sequence is converted into a repaired mine sample sequence to be reconstructed through the association table in step S1, and the Tokens video frame vector corresponding to each repaired mine sample to be reconstructed in the repaired mine sample sequence to be reconstructed is searched to obtain the healthy mine ecological grayscale picture feature vector H a1 , and the mine feature vector H to be reconstructed in step S2 q The model is automatically updated through a mask, and then a dimension correction grid is used to obtain the reference vegetation restoration color vector after the video frame information. ;

[0046] Step S6, the healthy mine ecological grayscale image vector generated in step S4 and the mine feature vector H to be reconstructed in step S2 q Through downsampling processing in turn, the two are combined and then corrected through a dimension grid to obtain the feature vector of the frame information before and after the reference vegetation restoration color vector a2 ;

[0047] Step S7, using the policy gradient algorithm to train the vegetation restoration color is completed, the input is the healthy mine ecological grayscale image sequence of the mine q to be reconstructed, and the output is the healthy mine ecological mine image with the vegetation restoration color completed .

[0048] For further information, see Attachment Figure 2 , step S1 further comprises:

[0049] Step S100, using a timer to collect training healthy mine ecological videos of historical mines to be rebuilt, and filtering to obtain a video set pair of the historical mines to be rebuilt;

[0050] Step S101, in the pre-training video frame vector stage, the cutting algorithm in the video image analysis is used to cut the ecological grayscale pictures of the mine to be reconstructed and the healthy mine in each sample in the video set, and after removing the duplicates of all the video frames in the video set, an association table is established, and the video frame vectors of the commonly used video frames are obtained by pre-training using the Tokens algorithm;

[0051] See attached Figure 3 , step S4 further comprises:

[0052] Step S401: using the encoder end of the Transformer network, input the mine feature vector H to be reconstructed q , the full text information of the mine to be reconstructed is integrated through the global attention mechanism, and the feature vector H′ of the mine to be reconstructed that integrates the full text information of the video set is further obtained q ;

[0053] Step S402: using the decoder end of the Transformer network, inputting the mine feature vector H′ to be reconstructed that incorporates the full text information of the video set q , sequentially output the corresponding healthy mine ecological grayscale picture vector All video frame vectors in ;

[0054] Step S5 further comprises:

[0055] Step S501: The feature vector H of the mine to be reconstructed q obtained in step S2 is q , through the loss model, the feature vector of the frame information before and after the video frame is obtained The reference vegetation restoration color a1 obtained by the retrieval algorithm is downsampled to obtain the feature vector of the frame information before and after the reference vegetation restoration color a1 ;

[0056] Step S502: The feature vector of the mine q to be reconstructed and retrieved healthy mine ecological grayscale images The eigenvector of First, we get the attention matrix by matrix multiplication Each element m in the attention matrix i，j Represents the similarity evaluation value between the i-th video frame in the video frame q and the j-th video frame in the reference vegetation restoration color a1. The matrix M1 is weighted and summed in the column dimension and the row dimension in turn. The summed result is and M q and the feature vector of the mine q to be reconstructed Multiply them to get the video frame vector H′ of the reference vegetation restoration color information q ,Will The feature vector of the reference vegetation restoration color a1 Multiply to get the reference vegetation restoration color vector of the video frame information ;

[0057] Step S503: The video frame vector H′ obtained in step S502 is q and the reference vegetation restoration color vector Two types of vector concatenation And through a feedforward neural network including two linear layers, the mine to be reconstructed q and the vegetation restoration color of the reference vegetation restoration color a1 are obtained to complete the healthy mine ecological mine picture ;

[0058] Step S6 further comprises:

[0059] Step S601: Convert the healthy mine ecological grayscale image vector After downsampling, the feature vector of the frame information before and after the reference vegetation restoration color vector a2 is obtained ;

[0060] Step S602: The feature vector H of the mine to be reconstructed q obtained in step S2 is q , through the loss model, the feature vector of the frame information before and after the video frame is obtained Then the feature vector of the previous and next frames of the video frame of the mine q to be reconstructed is obtained And healthy mine ecology grayscale picture vector First, we get the attention matrix by matrix multiplication After the weighted summation of the column dimension and the row dimension of matrix M2, the result after summation is and M q and the feature vector of video frame q Multiply them to get the video frame vector H″ of the reference vegetation restoration color information q ,Will Grayscale picture vector with healthy mine ecology Multiply to get the reference vegetation restoration color vector of the video frame information ;

[0061] Step S603: restore the video frame vector H″ of the reference vegetation color information obtained in step S602 q and the reference vegetation restoration color vector of the video frame information Two types of vector concatenation And through a feed-forward neural network including two linear layers, the vegetation restoration color of the mine to be reconstructed q and the reference vegetation restoration color vector a2 are obtained to complete the healthy mine ecological mine picture ;

[0062] Step S7 further comprises:

[0063] Step S701: compare the healthy mine ecological grayscale image sequence of the reference vegetation restoration color a1 obtained by the retrieval model with the standard reference vegetation restoration color a1 corresponding to the mine q to be reconstructed. * The L1-LOSS and EDGE-LOSS evaluation values ​​are calculated from the grayscale image sequence of the healthy mine ecology, which are f(Y r , Y * ) and g(Y r , Y * ), and the L1-LOSS evaluation value and the EDGE-LOSS evaluation value are weighted and summed to obtain the confidence weight corresponding to the reference vegetation restoration color a1. At the same time, the healthy mine ecological grayscale picture sequence of the reference vegetation restoration color vector a2 obtained by the generation model is compared with the standard reference vegetation restoration color a corresponding to the video frame q. * The L1-LOSS and EDGE-LOSS evaluation values ​​are calculated from the grayscale image sequence of the healthy mine ecology, which are f(Y g , Y * ) and g(Y g , Y * ), weighted sum of the L1-LOSS evaluation value and the EDGE-LOSS evaluation value to obtain the confidence weight corresponding to the reference vegetation restoration color vector a2;

[0064] Step S702: Compare the confidence weights obtained from the two reference vegetation restoration colors to obtain a final confidence weight.

[0065] Furthermore, in step S3, the L1-LOSS and EDGE-LOSS evaluation values ​​are combined as the final similarity judgment to search for the most relevant healthy mine ecological grayscale image sequence from the video collection.

[0066] Furthermore, in step S3, during the training phase, the grayscale image of the healthy mine ecology It is first sent to the convolution head module for dimension correction and downsampling operations to obtain ;

[0067] Where B is the sample size, C1=3 is Channel dimensions, H1=512, W1=512 are The height and width of C2=512 is Channel dimensions, H2=16, W2=16 are The height and width of

[0068] The calculation formulas for the L1-LOSS and EDGE-LOSS evaluation values ​​are as follows:

[0069] L1-Loss is used to calculate Compared with GT, GT refers to Groundtruth, that is, the absolute difference of RGB values ​​at each pixel in the real healthy mine ecological picture;

[0070] The formula is as follows:

[0071]

[0072] in, Indicates the total number of pixels on a picture; Represents the RGB value of the i-th pixel on the GT image; express The RBG value of the i-th pixel;

[0073] Edge-Loss ensures the clarity of the edges of objects in the vegetation restoration color area during color diffusion. The formula is as follows:

[0074]

[0075] Among them, X and Y are and the real GT image, E is the edge feature obtained, H and W are The height and width of the

[0076] Furthermore, in step S3, if the EDGE-LOSS evaluation value calculated between the retrieved mine to be rebuilt and the original mine to be rebuilt is not 0, the EDGE-LOSS evaluation value is adopted; otherwise, the L1-LOSS evaluation value is adopted.

[0077] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0078] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. The above is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the specification and drawings of this application, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present application.

[0079] The specific implementation methods of the invention are described in detail above, but they are only examples, and the present application is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the invention is also within the scope of the present application, and therefore, the equalization, modification, and improvement made without departing from the spirit and principle of the present application should be included in the scope of the present application.

Claims

1. A multi-dimensional data analysis and calculation method for mine ecological restoration projects based on three-dimensional real scenes, comprising the following steps: Step S1, through the collected video set of preset training healthy mine ecology videos, after cutting all video sets using a cutting algorithm, sort the healthy mine ecology grayscale pictures based on the frequency of transformation of the healthy mine ecology grayscale pictures in the video set, generate an association table of mappings between the healthy mine ecology grayscale pictures and the mine samples to be restored and reconstructed, and pre-train the video frame vector of each corresponding video frame in the association table; Step S2: In the generative model, after completing the video frame cutting for the mine to be reconstructed q to be processed by the user, the repaired mine to be reconstructed sample sequence is obtained by searching the association table, and the video frame vector pre-trained in step S1 is used to convert each repaired mine to be reconstructed sample in the repaired mine to be reconstructed sample sequence into a corresponding video frame vector to obtain the corresponding mine feature vector H q ; Step S3, based on the retrieval algorithm, the video frame healthy mine ecological grayscale image sequence obtained in step S1 is used to retrieve the repaired mine q′ closest to the mine q to be rebuilt from the video collection, and the healthy mine ecological grayscale image corresponding to the closest repaired mine q′ in the video collection. ; Step S4, based on the mine feature vector to be reconstructed obtained in step S2, using the transformer convolution head, generates a healthy mine ecological grayscale image vector related to the video frame ; Step S5, the healthy mine ecological grayscale picture retrieved in step S3 is cut by Tokens to obtain a healthy mine ecological grayscale picture sequence, and then the healthy mine ecological grayscale picture sequence is converted into a repaired mine sample sequence to be reconstructed through the association table in step S1, and the Tokens video frame vector corresponding to each repaired mine sample to be reconstructed in the repaired mine sample sequence to be reconstructed is searched to obtain the healthy mine ecological grayscale picture feature vector H a1 , and the mine feature vector H to be reconstructed in step S2 q The model is automatically updated through a mask, and then a dimension correction grid is used to obtain the reference vegetation restoration color vector after the video frame information. ; Step S6, the healthy mine ecological grayscale image vector generated in step S4 and the mine feature vector H to be reconstructed in step S2 q Through downsampling processing in turn, the two are combined and then corrected through a dimension grid to obtain the feature vector of the frame information before and after the reference vegetation restoration color vector a2 ; Step S7, using the policy gradient algorithm to train the vegetation restoration color is completed, the input is the healthy mine ecological grayscale image sequence of the mine q to be reconstructed, and the output is the healthy mine ecological mine image with the vegetation restoration color completed ; In step S3, the L1-LOSS and EDGE-LOSS evaluation values ​​are combined as the final similarity judgment, and the most relevant healthy mine ecological grayscale picture sequence is searched from the video collection; In step S3, during the training phase, the grayscale image of the healthy mine ecology It is first sent to the convolution head module for dimension correction and downsampling operations to obtain ; Where B is the sample size, C1=3 is Channel dimensions, H1=512, W1=512 are The height and width of C2=512 is Channel dimensions, H2=16, W2=16 are The height and width of The calculation formulas for the L1-LOSS and EDGE-LOSS evaluation values ​​are as follows: L1-Loss is used to calculate Compared with GT, GT refers to Groundtruth, that is, the absolute difference of RGB values ​​at each pixel in the real healthy mine ecological picture; The formula is as follows: in, Indicates the total number of pixels on a picture; Represents the RGB value of the i-th pixel on the GT image; express The RBG value of the i-th pixel; Edge-Loss ensures the clarity of the edges of objects in the vegetation restoration color area during color diffusion. The formula is as follows: Among them, X and Y are and the real GT image, E is the edge feature obtained, H and W are The height and width of In step S3, if the EDGE-LOSS evaluation value calculated between the retrieved mine to be rebuilt and the original mine to be rebuilt is not 0, the EDGE-LOSS evaluation value is used; otherwise, the L1-LOSS evaluation value is used; Step S7 includes: Step S701: compare the healthy mine ecological grayscale image sequence of the reference vegetation restoration color a1 obtained by the retrieval model with the standard reference vegetation restoration color a1 corresponding to the mine q to be reconstructed. * The L1-LOSS and EDGE-LOSS evaluation values ​​are calculated from the grayscale image sequence of the healthy mine ecology, which are f(Y r , Y * ) and g(Y r , Y * ), and the L1-LOSS evaluation value and the EDGE-LOSS evaluation value are weighted and summed to obtain the confidence weight corresponding to the reference vegetation restoration color a1. At the same time, the healthy mine ecological grayscale picture sequence of the reference vegetation restoration color vector a2 obtained by the generation model is compared with the standard reference vegetation restoration color a corresponding to the video frame q. * The L1-LOSS and EDGE-LOSS evaluation values ​​are calculated from the grayscale image sequence of the healthy mine ecology, which are f(Y g , Y * ) and g(Y g , Y * ), weighted sum of the L1-LOSS evaluation value and the EDGE-LOSS evaluation value to obtain the confidence weight corresponding to the reference vegetation restoration color vector a2; Step S702: Compare the confidence weights obtained from the two reference vegetation restoration colors to obtain a final confidence weight.

2. A method for analyzing and calculating multidimensional data of a mine ecological restoration project based on three-dimensional real scenes as claimed in claim 1, characterized in that: Step S1 further comprises: Step S100, using a timer to collect training healthy mine ecological videos of the historical mine to be rebuilt, and completing filtering to obtain a video set pair of the historical mine to be rebuilt; Step S101, in the pre-training video frame vector stage, the cutting algorithm in the video image analysis is used to cut the ecological grayscale pictures of the mine to be reconstructed and the healthy mine in each sample in the video set, and after removing the duplicates of all the video frames in the video set, an association table is established, and the video frame vectors of the commonly used video frames are obtained by pre-training using the Tokens algorithm; Step S4 further comprises: Step S401: using the encoder end of the Transformer network, input the mine feature vector H to be reconstructed q , the full text information of the mine to be reconstructed is integrated through the global attention mechanism, and the feature vector H′ of the mine to be reconstructed that integrates the full text information of the video set is further obtained q ; Step S402: using the decoder end of the Transformer network, inputting the mine feature vector H′ to be reconstructed that incorporates the full text information of the video set q , sequentially output the corresponding healthy mine ecological grayscale picture vector All video frame vectors in .

Images