Ecological restoration dynamic adjustment method and system for waste slag field of salt rock tunnel

Abstract

The invention provides a dynamic adjustment method and system for ecological restoration of a waste slag yard of a salt rock tunnel. The method comprises the following steps: acquiring a current vegetation growth image of the waste slag yard and current vegetation growth description information of the waste slag yard; extracting the multi-modal fusion feature information of the current vegetation growth image of the waste slag field and the text feature information of the vegetation growth description information; calculating a current ecological restoration score of the waste slag field according to the multi-modal fusion feature information and the text feature information; and comparing the ecological restoration score of the waste slag field with a preset target score, and dynamically adjusting the restoration method according to comparison information. According to the method, the vegetation growth image and description information of the waste slag field are obtained, feature extraction is performed on the vegetation growth image and description information, and the ecological restoration score is dynamically calculated in combination with the multi-modal fusion feature and the text feature, so that intelligent assessment of the restoration process is realized. And the restoration strategy is timely optimized according to the ecological restoration score, so that the problems of resource waste and restoration period prolonging are avoided, and the restoration efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of ecological restoration technology, and more specifically, to a method and system for dynamic adjustment of ecological restoration of salt rock tunnel spoil heaps. Background Technology

[0002] With the rapid development of infrastructure construction globally and in my country, transportation networks, represented by railways and highways, are extending and expanding at an unprecedented pace. Tunnel construction is an indispensable and crucial component of these massive projects, especially in mountainous areas with complex terrain and diverse geology. The continuous construction of tunnels has also generated numerous environmental problems. For example, the construction of salt rock tunnels produces a large amount of waste material, and waste disposal sites often suffer from soil salinization and difficulty in vegetation survival, thus requiring ecological restoration. However, traditional restoration methods suffer from long restoration cycles and low efficiency. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for dynamic adjustment of ecological restoration of salt rock tunnel spoil disposal sites, so as to improve the above-mentioned problems.

[0004] To achieve the above objectives, the embodiments of this application provide the following technical solutions: On the one hand, embodiments of this application provide a method for dynamic adjustment of ecological restoration of salt rock tunnel spoil heaps, the method comprising: Obtain the current vegetation growth image and description information of the current vegetation growth at the spoil disposal site; Extract multimodal fusion feature information and textual feature information of vegetation growth description information from the current vegetation growth image of the spoil disposal site; The current ecological restoration score of the spoil heap is calculated based on multimodal fusion feature information and text feature information; the ecological restoration score of the spoil heap is compared with the preset target score, and the restoration method is dynamically adjusted based on the comparison information.

[0005] Secondly, embodiments of this application provide a dynamic adjustment system for ecological restoration of salt rock tunnel spoil heaps, the system comprising: The acquisition module is used to acquire the current vegetation growth image of the spoil disposal site and the current vegetation growth description information of the spoil disposal site. The extraction module is used to extract multimodal fusion feature information and textual feature information of vegetation growth description information from the current vegetation growth image of the spoil disposal site. The repair module is used to calculate the current ecological restoration score of the waste disposal site based on multimodal fusion feature information and text feature information; compare the ecological restoration score of the waste disposal site with the preset target score, and dynamically adjust the repair method based on the comparison information.

[0006] Thirdly, embodiments of this application provide a dynamic adjustment device for ecological restoration of a salt rock tunnel spoil heap, the device comprising a memory and a processor. The memory stores a computer program; the processor executes the computer program to implement the steps of the aforementioned dynamic adjustment method for ecological restoration of a salt rock tunnel spoil heap.

[0007] Fourthly, embodiments of this application provide a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described dynamic adjustment method for ecological restoration of salt rock tunnel spoil heaps.

[0008] The beneficial effects of this invention are as follows: This invention acquires images and descriptive information of vegetation growth at waste disposal sites, extracts features from them, and dynamically calculates ecological restoration scores by combining multimodal fusion features and textual features, thus achieving intelligent evaluation of the restoration process. Based on the ecological restoration scores, restoration strategies are optimized in a timely manner, avoiding resource waste and prolonged restoration cycles, thereby improving restoration efficiency.

[0009] This invention performs multimodal fusion of vegetation growth images and descriptive information. Through dual feature extraction using a convolutional extraction module and an attention mechanism, it achieves complementarity and integration of image and textual information. This multimodal fusion technology effectively overcomes the limitations of single-modal information; for example, images may not fully reflect vegetation growth, while textual descriptions may contain subjective biases. By combining visual information from images with semantic information from text, a more comprehensive set of ecological restoration features is generated. This combined feature set is then input into a scoring detection model, enabling a more accurate assessment of the ecological restoration level of spoil heaps.

[0010] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of the dynamic adjustment method for ecological restoration of salt rock tunnel spoil disposal sites as described in this embodiment of the invention; Figure 2This is a schematic diagram of the dynamic adjustment system for ecological restoration of salt rock tunnel spoil disposal sites as described in this embodiment of the invention; Figure 3 This is a schematic diagram of the dynamic adjustment equipment structure for ecological restoration of the salt rock tunnel spoil disposal site as described in this embodiment of the invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0014] It should be noted that similar reference numerals or letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0015] Example 1 like Figure 1 As shown in the figure, this embodiment provides a dynamic adjustment method for ecological restoration of salt rock tunnel spoil disposal sites, which includes steps S1, S2 and S3.

[0016] Step S1: Obtain the current vegetation growth image of the spoil disposal site and obtain the current vegetation growth description information of the spoil disposal site; In this step, image acquisition devices can be installed at high altitudes to capture images of vegetation growth throughout the entire spoil heap; alternatively, drones can be used to collect vegetation growth images. The vegetation growth description information can be understood as any textual description of vegetation growth; for example, "In the ecological restoration area of ​​the spoil heap, monitoring revealed that the coverage of the target plants has increased from <10% initially to approximately 65% ​​currently. The overall growth is good and healthy, with no obvious pests or diseases. Species richness has increased, and a preliminary community structure has been formed, dominated by the target herbaceous plants and accompanied by a small number of shrubs." Step S2: Extract the multimodal fusion feature information and textual feature information of the vegetation growth description information of the current vegetation growth image of the spoil disposal site; The specific implementation steps of this step include step S21 and step S22; Step S21: Construct a dual feature extraction module, which includes a convolution extraction module and an attention extraction module. The convolution extraction module includes a first convolutional layer and a second convolutional layer, and the attention extraction module includes a first Transformer model and a second Transformer model. Step S22: Use the dual feature extraction module to extract the multimodal fusion feature information of the current vegetation growth image of the spoil disposal site; perform word segmentation on the vegetation growth description information, and extract the text feature information of the vegetation growth description information based on the word segmentation results.

[0017] The specific implementation steps of this step include step S221 and step S222; Step S221: Process the vegetation growth image through the first convolutional layer to generate the first feature information corresponding to each position; feed the vegetation growth image into the first Transformer model to obtain the second feature information corresponding to each position; superimpose the two feature information element by element to obtain the intermediate feature at each position; input all intermediate features into the second convolutional layer to obtain the third feature information; input all intermediate features into the second Transformer model to obtain the fourth feature information. In this step, the two feature information are superimposed element by element to obtain the intermediate feature at each position. This can be understood as: adding the first feature information and the second feature information corresponding to each position to obtain the intermediate feature corresponding to each position. Step S222: Concatenate the first feature information, the second feature information, and the third feature information in sequence to form the first target feature; concatenate the first feature information, the second feature information, and the fourth feature information in sequence to form the second target feature; concatenate the first target feature and the second target feature in sequence to obtain the multimodal fusion feature information corresponding to the vegetation growth image; use the forward maximum matching method to perform word segmentation processing on the vegetation growth description information, and extract the text feature information of the vegetation growth description information based on the word segmentation processing results.

[0018] In this step, the first feature information, the second feature information, and the third feature information are concatenated in sequence to form the first target feature. This can be understood as: the first feature information A, the second feature information B, and the third feature information C are concatenated in sequence to form A, B, C. The concatenation in other steps is similar. The specific implementation steps for extracting text feature information of vegetation growth description information based on word segmentation processing results include step S2221. Step S2221: Use the forward maximum matching method to segment the vegetation growth description information to obtain multiple first fields in the vegetation growth description information; calculate the similarity between each first field and each second field in the preset target field library, and extract the first field whose similarity calculation result is greater than the similarity threshold and record it as the target field; among all the first fields, use a masking language model to randomly mask the first fields other than the target fields to obtain the masked fields; extract features from the masked fields to obtain the text feature information of the vegetation growth description information.

[0019] In this step, firstly, through positive maximum matching and similarity calculation, key professional indicators such as "75% coverage" and "no pests or diseases" are automatically identified and retained as target fields, effectively strengthening core information. Subsequently, a masked language model (MLM) is used to randomly mask non-target fields, filtering out redundant and irrelevant content, reducing noise interference, and enhancing the processing's resistance to overfitting and data security. Finally, the masked text has a clear structure and focused semantics, enabling the feature extraction stage to more efficiently generate highly discriminative and accurate numerical feature vectors.

[0020] Step S3: Calculate the current ecological restoration score of the waste disposal site based on the multimodal fusion feature information and text feature information; compare the ecological restoration score of the waste disposal site with the preset target score, and dynamically adjust the restoration method based on the comparison information.

[0021] The specific implementation steps of this step include step S31; Step S31: Combine multimodal fusion feature information and text feature information to form combined feature information. Input the combined feature information into a preset scoring detection model to obtain the current ecological restoration score. Compare the ecological restoration score with the preset target score. If the ecological restoration score is greater than or equal to the preset target score, no processing is performed. If the ecological restoration score is less than the preset target score, obtain the current soil attribute information of the spoil disposal site. The soil attribute information includes soil salinity, soil temperature, soil pH value, and soil moisture content. Send the current soil attribute information to the staff to prompt them to adjust the restoration method based on the current soil attribute information.

[0022] In this step, the scoring model is obtained by using the combined feature information obtained from historical periods as input and the corresponding ecological restoration score as output to train a convolutional neural network model to obtain the scoring model. In this step, sending the current soil property information to the staff so they can adjust the remediation method based on this information can be understood as sending the current soil property information to the staff, who then manually adjust the remediation method accordingly. For example: Staff can first compare the collected soil property information—soil salinity, soil temperature, soil pH, and soil moisture content—with the suitable ranges for soil salinity, soil temperature, soil pH, and soil moisture content, and obtain the following comparison results: Soil salinity: Slightly high; Soil pH: Slightly alkaline (e.g., pH 8.2, while the suitable range is 6.5-7.5); Soil temperature: High; Soil moisture content: Low; Based on the above comparison results, the repair method can be adjusted as follows: Analyzing data correlations: Staff first combine information on current vegetation growth (low score) with soil salinity, high pH, ​​and low moisture levels to determine if these soil properties are the main causes or contributing factors to poor vegetation restoration. High salinity and alkalinity may inhibit the growth of certain key vegetation species, while insufficient water exacerbates drought stress.

[0023] Targeted adjustment measures: Salinity and pH issues: Soil improvement measures can be taken to address salinity and alkalinity problems. For example: Leaching: Increase the amount of irrigation water to reduce soil salinity and partially dissolved alkaline substances through dilution and leaching.

[0024] Apply soil conditioners: Spread sulfur powder or apply acidic substances (such as organic acid waste, peat moss) to neutralize soil alkalinity and improve pH levels. Alternatively, use specialized soil passivators / stabilizers to fix excessively high salinity.

[0025] Moisture issues: For soils with low moisture levels, irrigation plans can be optimized. For example: Adjust irrigation frequency and depth: Adopt a more scientific irrigation plan to ensure that the soil maintains appropriate moisture.

[0026] Improve drainage: If poor drainage leads to water accumulation or nutrient loss, it is necessary to dredge the drainage ditches or adopt a more efficient drainage system.

[0027] Laying water-retaining materials: Laying straw mulch, plastic film or water-retaining agent on the ground to reduce water evaporation.

[0028] Choose species with stronger adaptability: Before soil conditions are fundamentally improved, consider using vegetation species that are more tolerant of salt and alkali and drought to replant or adjust the vegetation configuration structure.

[0029] Strengthen monitoring and feedback: After implementing adjustment measures, it is necessary to strengthen follow-up monitoring, including regularly collecting soil samples and vegetation images, recalculating ecological restoration scores, verifying the effectiveness of the adjustments, and providing a basis for further optimization of strategies.

[0030] In this way, staff can accurately identify the root cause of the problem based on the soil property data provided by the system, and take targeted measures to optimize and adjust the restoration strategy, thereby improving the efficiency and effectiveness of ecological restoration.

[0031] In this embodiment, an automated and intelligent ecological restoration assessment system was constructed by integrating multimodal data and textual information. This method significantly enhances the comprehensiveness and accuracy of the assessment, effectively overcomes the limitations of a single data source, and achieves a multi-faceted quantitative assessment of the ecological restoration status of spoil heaps. Simultaneously, when the restoration progress fails to meet standards, the system automatically triggers soil property data collection and delivery, providing staff with precise adjustment guidelines, avoiding blind operations, thereby shortening the restoration cycle and reducing resource waste.

[0032] Example 2 like Figure 2 As shown in the figure, this embodiment provides a dynamic adjustment system for ecological restoration of salt rock tunnel spoil disposal sites. The system includes an acquisition module 1, an extraction module 2, and a restoration module 3.

[0033] Module 1 is used to acquire the current vegetation growth image of the spoil disposal site and the current vegetation growth description information of the spoil disposal site. Extraction module 2 is used to extract multimodal fusion feature information and text feature information of vegetation growth description information from the current vegetation growth image of the spoil disposal site; Repair module 3 is used to calculate the current ecological restoration score of the waste disposal site based on multimodal fusion feature information and text feature information; compare the ecological restoration score of the waste disposal site with the preset target score, and dynamically adjust the restoration method based on the comparison information.

[0034] In one specific embodiment of this disclosure, the extraction module 2 further includes a construction unit 21 and a first extraction unit 22.

[0035] Building unit 21 is used to build a dual feature extraction module, which includes a convolutional extraction module and an attention extraction module. The convolutional extraction module includes a first convolutional layer and a second convolutional layer, and the attention extraction module includes a first Transformer model and a second Transformer model. The first extraction unit 22 is used to extract multimodal fusion feature information of the current vegetation growth image of the spoil disposal site using the dual feature extraction module; to perform word segmentation on the vegetation growth description information; and to extract the text feature information of the vegetation growth description information based on the word segmentation results.

[0036] In one specific embodiment of this disclosure, the first extraction unit 22 further includes a second extraction unit 221 and a third extraction unit 222.

[0037] The second extraction unit 221 is used to process the vegetation growth image through the first convolutional layer to generate first feature information corresponding to each position; feed the vegetation growth image into the first Transformer model to obtain second feature information corresponding to each position; superimpose the two feature information element by element to obtain intermediate features for each position; input all intermediate features into the second convolutional layer to obtain third feature information; and input all intermediate features into the second Transformer model to obtain fourth feature information. The third extraction unit 222 is used to sequentially concatenate the first feature information, the second feature information, and the third feature information to form the first target feature; sequentially concatenate the first feature information, the second feature information, and the fourth feature information to form the second target feature; sequentially concatenate the first target feature and the second target feature to obtain the multimodal fusion feature information corresponding to the vegetation growth image; use the forward maximum matching method to perform word segmentation processing on the vegetation growth description information, and extract the text feature information of the vegetation growth description information based on the word segmentation processing results.

[0038] In one specific embodiment of this disclosure, the third extraction unit 222 further includes a fourth extraction unit 2221.

[0039] The fourth extraction unit 2221 is used to perform word segmentation on the vegetation growth description information using the forward maximum matching method to obtain multiple first fields in the vegetation growth description information; to perform similarity calculation on each first field with each second field in the preset target field library, and to extract the first field whose similarity calculation result is greater than the similarity threshold and record it as the target field; to perform random masking on the first fields other than the target fields using a masking language model to obtain the masked fields; and to perform feature extraction on the masked fields to obtain the text feature information of the vegetation growth description information.

[0040] In one specific embodiment of this disclosure, the repair module 3 further includes a repair unit 31.

[0041] The restoration unit 31 is used to combine multimodal fusion feature information and text feature information to form combined feature information, input the combined feature information into a preset scoring detection model to obtain the current ecological restoration score; compare the ecological restoration score with the preset target score, if the ecological restoration score is greater than or equal to the preset target score, no processing is performed; if the ecological restoration score is less than the preset target score, the current soil attribute information of the spoil disposal site is obtained, including soil salinity, soil temperature, soil pH value, and soil moisture content, and the current soil attribute information is sent to the staff to prompt the staff to adjust the restoration method based on the current soil attribute information.

[0042] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0043] Example 3 Corresponding to the above method embodiments, this disclosure also provides a dynamic adjustment device for ecological restoration of salt rock tunnel spoil heaps. The dynamic adjustment device for ecological restoration of salt rock tunnel spoil heaps described below and the dynamic adjustment method for ecological restoration of salt rock tunnel spoil heaps described above can be referred to each other.

[0044] Figure 3 This is a block diagram illustrating a dynamic adjustment device 300 for ecological restoration of a salt rock tunnel spoil heap, according to an exemplary embodiment. Figure 3 As shown, the dynamic adjustment device 300 for ecological restoration of the salt rock tunnel spoil heap may include: a processor 301 and a memory 302. The dynamic adjustment device 300 may also include one or more of the following: a multimedia component 303, an I / O interface 304, and a communication component 305.

[0045] The processor 301 controls the overall operation of the dynamic adjustment device 300 for ecological restoration of the salt rock tunnel spoil heap, in order to complete all or part of the steps in the aforementioned dynamic adjustment method for ecological restoration of the salt rock tunnel spoil heap. The memory 302 stores various types of data to support the operation of the dynamic adjustment device 300 for ecological restoration of the salt rock tunnel spoil heap. This data may include, for example, instructions for any application or method operating on the dynamic adjustment device 300 for ecological restoration of the salt rock tunnel spoil heap, as well as application-related data, such as contact data, sent and received messages, images, audio, video, etc. The memory 302 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 303 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 302 or transmitted via the communication component 305. The audio component also includes at least one speaker for outputting audio signals. I / O interface 304 provides an interface between processor 301 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 305 is used for wired or wireless communication between the dynamic adjustment device 300 for ecological restoration of the salt rock tunnel spoil heap and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 305 may include a Wi-Fi module, a Bluetooth module, and an NFC module.

[0046] In an exemplary embodiment, the dynamic adjustment device 300 for ecological restoration of the salt rock tunnel spoil heap can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the aforementioned dynamic adjustment method for ecological restoration of the salt rock tunnel spoil heap.

[0047] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, these program instructions implement the steps of the above-described dynamic adjustment method for ecological restoration of a salt rock tunnel spoil heap. For example, the computer-readable storage medium may be the memory 302 including program instructions, which may be executed by the processor 301 of the dynamic adjustment device 300 for ecological restoration of the salt rock tunnel spoil heap to complete the above-described dynamic adjustment method for ecological restoration of the salt rock tunnel spoil heap.

[0048] Example 4 Corresponding to the above method embodiments, this disclosure also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the above-described dynamic adjustment method for ecological restoration of salt rock tunnel spoil heaps.

[0049] A readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the dynamic adjustment method for ecological restoration of salt rock tunnel spoil disposal sites as described in the above method embodiments.

[0050] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.

[0051] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for ecological restoration dynamic adjustment of a salt rock tunnel waste dump, characterized in that, The method comprises the following steps: acquiring the current vegetation growth image of the spoil field and acquiring the current vegetation growth description information of the spoil field; extracting the multi-modal fusion feature information of the current vegetation growth image of the spoil field and the text feature information of the vegetation growth description information; calculating the current ecological restoration score of the spoil field according to the multi-modal fusion feature information and the text feature information; comparing the ecological restoration score of the spoil field with a preset target score, and dynamically adjusting the restoration method according to the comparison information.

2. The method according to claim 1, characterized in that, The method for extracting the multi-modal fusion feature information of the current vegetation growth image of the spoil field and the text feature information of the vegetation growth description information comprises the following steps: a double feature extraction module is constructed, the double feature extraction module comprises a convolution extraction module and an attention extraction module, the convolution extraction module comprises a first convolution layer and a second convolution layer, and the attention extraction module comprises a first Transformer model and a second Transformer model; the multi-modal fusion feature information of the current vegetation growth image of the spoil field is extracted by using the double feature extraction module; and the text feature information of the vegetation growth description information is extracted according to the word segmentation processing result of the vegetation growth description information.

3. The method according to claim 2, characterized in that, The method for extracting the multi-modal fusion feature information of the current vegetation growth image of the spoil field by using the double feature extraction module and the text feature information of the vegetation growth description information according to the word segmentation processing result of the vegetation growth description information comprises the following steps: the vegetation growth image is processed by the first convolution layer to generate first feature information corresponding to each position; the vegetation growth image is input into the first Transformer model to obtain second feature information corresponding to each position; the two feature information are element-wise superimposed to obtain intermediate features of each position; all intermediate features are input into the second convolution layer to obtain third feature information; and all intermediate features are input into the second Transformer model to obtain fourth feature information; the first feature information, the second feature information and the third feature information are sequentially spliced to form a first target feature; the first feature information, the second feature information and the fourth feature information are sequentially spliced to form a second target feature; the first target feature and the second target feature are sequentially spliced to obtain the multi-modal fusion feature information corresponding to the vegetation growth image; and the word segmentation processing of the vegetation growth description information is performed by using the forward maximum matching method, and the text feature information of the vegetation growth description information is extracted according to the word segmentation processing result.

4. The method according to claim 3, characterized in that, The method for extracting the text feature information of the vegetation growth description information according to the word segmentation processing result comprises the following steps: the word segmentation processing of the vegetation growth description information is performed by using the forward maximum matching method to obtain a plurality of first fields in the vegetation growth description information; similarity calculation is performed on each first field and each second field in a preset target field library, and a first field with a similarity calculation result greater than a similarity threshold value is extracted and recorded as a target field; in all first fields, a first field other than the target field is randomly masked by using a mask language model to obtain a masked field; and the text feature information of the vegetation growth description information is obtained by performing feature extraction on the masked field.

5. The method according to claim 1, wherein the ecological restoration score of the waste rock field is calculated according to the multi-modal fusion feature information and the text feature information. The ecological restoration score of the waste dump is compared with a preset target score, and the restoration method is dynamically adjusted according to the comparison information, including: The multi-modal fusion feature information and the text feature information are combined to form combined feature information, and the combined feature information is input into a preset score detection model to obtain a current ecological restoration score. The ecological restoration score is compared with the preset target score. If the ecological restoration score is greater than or equal to the preset target score, no processing is performed. If the ecological restoration score is less than the preset target score, the current soil property information of the waste dump is obtained, including soil salinity, soil temperature, soil pH value, and soil moisture content. The current soil property information is sent to the worker to prompt the worker to adjust the restoration method based on the current soil property information.

6. A dynamic adjustment system for ecological restoration of a salt rock tunnel spoil yard, characterized in that, Including: The acquisition module is configured to acquire a current vegetation growth image of the waste dump and acquire current vegetation growth description information of the waste dump. The extraction module is configured to extract multi-modal fusion feature information of the current vegetation growth image of the waste dump and text feature information of the vegetation growth description information. The restoration module is configured to calculate a current ecological restoration score of the waste dump based on the multi-modal fusion feature information and the text feature information, and compare the ecological restoration score of the waste dump with a preset target score, and dynamically adjust the restoration method according to the comparison information.

7. The salt rock tunnel waste dump ecological restoration dynamic adjustment system according to claim 6, characterized in that, The extraction module includes: The construction unit is configured to construct a double feature extraction module, which includes a convolution extraction module and an attention extraction module. The convolution extraction module includes a first convolution layer and a second convolution layer, and the attention extraction module includes a first Transformer model and a second Transformer model. The first extraction unit is configured to extract multi-modal fusion feature information of the current vegetation growth image of the waste dump using the double feature extraction module, and perform word segmentation processing on the vegetation growth description information to extract text feature information of the vegetation growth description information based on the word segmentation processing result.

8. The salt rock tunnel waste dump ecological restoration dynamic adjustment system according to claim 7, characterized in that, The first extraction unit includes: The second extraction unit is configured to process the vegetation growth image through the first convolution layer to generate first feature information corresponding to each position, and input the vegetation growth image into the first Transformer model to obtain second feature information corresponding to each position. The two feature information are element-wise superimposed to obtain intermediate features of each position. All intermediate features are input into the second convolution layer to obtain third feature information. All intermediate features are input into the second Transformer model to obtain fourth feature information. The third extraction unit is configured to sequentially splice the first feature information, the second feature information, and the third feature information to form a first target feature, and sequentially splice the first feature information, the second feature information, and the fourth feature information to form a second target feature. The first target feature and the second target feature are sequentially spliced to obtain multi-modal fusion feature information corresponding to the vegetation growth image. The word segmentation processing is performed on the vegetation growth description information using the forward maximum matching method, and the text feature information of the vegetation growth description information is extracted based on the word segmentation processing result.

9. The salt rock tunnel waste dump ecological restoration dynamic adjustment system according to claim 8, characterized in that, The third extraction unit includes: The fourth extraction unit is configured to perform word segmentation on the vegetation growth description information by using a forward maximum matching method to obtain a plurality of first fields in the vegetation growth description information; perform similarity calculation on each first field and each second field in a preset target field library; extract a first field with a similarity calculation result greater than a similarity threshold value and record the first field as a target field; perform random mask processing on the first fields other than the target field by using a mask language model to obtain a field after mask processing; and perform feature extraction on the field after mask processing to obtain text feature information of the vegetation growth description information.

10. The ecological restoration dynamic adjustment system for a salt rock tunnel waste dump according to claim 6, the restoration module comprising: a restoration unit configured to combine the multi-modal fusion feature information and the text feature information to form combined feature information, input the combined feature information into a preset score detection model, and obtain a current ecological restoration score; compare the ecological restoration score with a preset target score, if the ecological restoration score is greater than or equal to the preset target score, do not perform any processing; if the ecological restoration score is less than the preset target score, obtain current soil property information of the waste dump, the soil property information including soil salinity, soil temperature, soil pH value, and soil moisture content, and send the current soil property information to a worker to prompt the worker to adjust the restoration method based on the current soil property information.

Images