Method and system for identifying and assessing ecological damage of soil heavy metals based on regional scale

Abstract

The present application relates to the field of intelligent detection and provides a method and system for identifying and assessing soil heavy metal ecological damage based on a regional scale. The method divides the area to be assessed into multiple regional units, and uses ICP-MS detection technology to analyze the heavy metal content of each of the multiple regional units to obtain detection results for the multiple regional units. The method also uses deep learning-based data analysis technology to embed and encode the detection results of the heavy metal content of each regional unit. In this way, based on the query matching representation between the semantic embedding features of the heavy metal content of the first regional unit and each region, it is automatically determined whether the soil heavy metal ecological damage of the first regional unit is abnormal. In this way, rapid on-site detection of soil samples can be achieved, and by more finely capturing subtle changes in the heavy metal content in the soil, it is beneficial to improve the accuracy of the assessment.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection, and more specifically, to a method and system for identifying and assessing soil heavy metal ecological damage based on a regional scale. Background Art

[0002] With the accelerated development of industrialization, agricultural activities, and urbanization, soil pollution is becoming an increasingly prominent problem. Heavy metal pollution is a key issue. Once heavy metals such as lead (Pb), cadmium (Cd), mercury (Hg), and arsenic (As) enter the soil, they accumulate in the environment over long periods of time and are difficult to decompose, causing serious damage to the ecosystem. For example, heavy metals can affect plant growth, alter the structure of soil microbial communities, and even enter the human body through the food chain, endangering human health. Therefore, monitoring and assessing heavy metal pollution in soil has become particularly important.

[0003] However, traditional methods for identifying and assessing ecological damage caused by heavy metals in soil rely on manual sampling and laboratory analysis, resulting in a time-consuming and inefficient process. From sampling to laboratory testing to data analysis, each link requires a lot of time and manpower, making it impossible to achieve a rapid assessment. In addition, traditional methods usually use a limited number of sampling points to represent the entire area, resulting in the inability to conduct fine-grained analysis of local areas. Specifically, due to the limited number of sampling points, it is impossible to capture subtle changes in the content of heavy metals in the soil, especially near the pollution source or at the pollution boundary. It is easy to miss high-concentration areas or transition zones at the edge of the pollution, so that the accuracy and reliability of the assessment results are limited.

[0004] Therefore, an optimized identification and assessment scheme for soil heavy metal ecological damage is desired. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, this application provides a method and system for identifying and evaluating soil heavy metal ecological damage based on a regional scale.

[0006] A regional-scale soil heavy metal ecological damage identification and assessment method includes:

[0007] Divide the area to be assessed into multiple regional units;

[0008] Detecting the heavy metal content of each of the plurality of regional units to obtain heavy metal content detection results of the plurality of regional units;

[0009] performing high-dimensional embedding coding on each of the plurality of regional unit heavy metal content detection results to obtain semantic embedding coding vectors of the plurality of regional unit heavy metal content detection results;

[0010] Performing query matching implicit optimization on the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit among the multiple regional units and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results to obtain an implicit optimization representation of the query matching of the first regional unit heavy metal content, including: processing the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit to obtain a semantic value feature vector of the first regional unit detection result; using the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit as a query vector, the semantic value feature vector of the first regional unit detection result as a value vector, and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results as a key vector sequence, performing attention gating analysis based on local domain on the sequence of the query vector, the value vector, and the key vector to obtain an implicit optimization representation of the query matching of the first regional unit heavy metal content;

[0011] An implicit optimization representation is matched based on the query of the heavy metal content of the first regional unit to obtain an evaluation result.

[0012] A soil heavy metal ecological damage identification and assessment system based on regional scale, comprising:

[0013] An area division module for evaluation, used for dividing the area for evaluation into multiple area units;

[0014] A regional unit heavy metal content detection module, configured to detect the heavy metal content of each of the plurality of regional units to obtain a plurality of regional unit heavy metal content detection results;

[0015] A regional unit heavy metal content detection data embedding coding module is used to perform high-dimensional embedding coding on each of the multiple regional unit heavy metal content detection results to obtain multiple regional unit heavy metal content detection result semantic embedding coding vectors;

[0016] An area unit heavy metal content coding vector query matching optimization module is used to perform query matching implicit optimization on the semantic embedding coding vector of the area unit heavy metal content detection result corresponding to the first area unit among the multiple area units and the semantic embedding coding vectors of the multiple area unit heavy metal content detection results to obtain an implicitly optimized representation of the query matching of the first area unit heavy metal content, including: an area unit detection result semantic value feature generation unit, used to process the semantic embedding coding vector of the area unit heavy metal content detection result corresponding to the first area unit to obtain a first area unit detection result semantic value feature vector; an area unit heavy metal content query matching implicitly optimized representation generation unit, used to use the semantic embedding coding vector of the area unit heavy metal content detection result corresponding to the first area unit as a query vector, the first area unit detection result semantic value feature vector as a value vector, and the multiple area unit heavy metal content detection result semantic embedding coding vectors as a key vector sequence, and perform attention gating analysis based on local domain on the sequence of the query vector, the value vector, and the key vector to obtain an implicitly optimized representation of the query matching of the first area unit heavy metal content;

[0017] An evaluation result generation module is used to obtain an evaluation result by querying and matching an implicit optimization representation based on the heavy metal content of the first regional unit.

[0018] This application has significant technical effects due to the adoption of the above technical solutions:

[0019] The present application provides a regional-scale soil heavy metal ecological damage identification and assessment method and system. The method divides the area to be assessed into multiple regional units, analyzes the heavy metal content of each of the multiple regional units using ICP-MS detection technology, and obtains multiple regional unit detection results. Deep learning-based data analysis technology is then used to embed and encode the heavy metal content detection results of each regional unit. This automatically determines whether the soil heavy metal ecological damage in the first regional unit is abnormal based on query matching representations between the semantic embedding features of the heavy metal content of the first regional unit and each region. This allows for rapid on-site testing of soil samples and improves the accuracy of the assessment by more precisely capturing subtle changes in the heavy metal content in the soil. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended 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 drawings, the same reference numerals generally represent the same components or steps.

[0021] Figure 1 Flowchart of a method for identifying and assessing soil heavy metal ecological damage based on a regional scale according to an embodiment of the present application.

[0022] Figure 2 Schematic diagram of data flow for a method for identifying and assessing soil heavy metal ecological damage based on a regional scale according to an embodiment of the present application.

[0023] Figure 3 In a method for identifying and assessing ecological damage of heavy metals in soil based on a regional scale according to an embodiment of the present application, a flowchart is provided for performing implicit optimization of query matching on the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit among the multiple regional units and the semantic embedding coding vectors of the heavy metal content detection results of the multiple regional units to obtain an implicit optimized representation of the query matching of the heavy metal content of the first regional unit.

[0024] Figure 4 In the soil heavy metal ecological damage identification and assessment method based on the regional scale according to an embodiment of the present application, the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit is used as the query vector, the semantic value feature vector of the first regional unit detection result is used as the value vector and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results are used as a sequence of key vectors. The sequence of the query vector, the value vector and the key vector is subjected to attention gating analysis based on the local field to obtain a flowchart of the implicit optimization representation of the query matching of the first regional unit heavy metal content.

[0025] Figure 5 This is a system block diagram of a regional-scale soil heavy metal ecological damage identification and assessment system according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] 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 herein.

[0027] With the rapid development of industrialization, agriculture, and urbanization, heavy metal contamination of soil has become increasingly serious. Among these, lead, cadmium, mercury, and arsenic are particularly critical. These heavy metals are difficult to degrade in soil, accumulating over time and causing persistent damage to ecosystems. They not only affect plant growth and soil microbial structure but can also enter the human body through the food chain, posing a threat to human health. Therefore, effective monitoring and assessment of heavy metal contamination in soil is extremely urgent.

[0028] However, traditional methods for identifying and assessing ecological damage caused by heavy metals in soil rely on manual sampling and laboratory analysis, which is a time-consuming and inefficient process. From sampling to analysis to data interpretation, each step requires a lot of time and manpower, making it difficult to achieve rapid assessment. In addition, traditional methods usually conduct regional representative analysis based on limited sampling points, which limits detailed research on local areas. Due to the limited number of sampling points, this method may not be able to accurately capture subtle changes in the content of heavy metals in the soil, especially near pollution sources or in pollution-edge areas, and it is easy to ignore high-concentration areas or pollution transition zones, thereby affecting the accuracy and reliability of the assessment results.

[0029] To address the above technical issues, the technical concept of this application is to divide the area to be assessed into multiple regional units, and use ICP-MS detection technology to analyze the heavy metal content of each of the multiple regional units to obtain detection results for multiple regional units. Deep learning-based data processing and analysis technology is then used to embed and encode the detection results of the heavy metal content of each regional unit. This automatically determines whether the soil heavy metal ecological damage in the first regional unit is abnormal based on the query matching representation between the semantic embedding features of the heavy metal content of the first regional unit and each region. In this way, through automated sampling equipment and advanced detection technology, rapid on-site testing of soil samples can be achieved. Moreover, by dividing the area into multiple regional units, subtle changes in the heavy metal content in the soil can be analyzed more precisely, capturing the complex relationships between different heavy metal elements, thereby improving the accuracy of the assessment and realizing intelligent soil heavy metal ecological damage assessment.

[0030] Figure 1 Flowchart of a method for identifying and assessing soil heavy metal ecological damage based on a regional scale according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the method for identifying and assessing soil heavy metal ecological damage based on regional scale according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to the embodiment of the present application, the soil heavy metal ecological damage identification and assessment method based on the regional scale includes: S110, dividing the area to be assessed to obtain multiple regional units; S120, detecting the heavy metal content of each regional unit in the multiple regional units to obtain multiple regional unit heavy metal content detection results; S130, performing high-dimensional embedding coding on each regional unit heavy metal content detection result in the multiple regional unit heavy metal content detection results to obtain multiple regional unit heavy metal content detection result semantic embedding coding vectors; S140, performing query matching implicit optimization on the regional unit heavy metal content detection result semantic embedding coding vector corresponding to the first regional unit in the multiple regional units and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results to obtain a first regional unit heavy metal content query matching implicit optimization representation; S150, obtaining an assessment result based on the first regional unit heavy metal content query matching implicit optimization representation.

[0031] In step S110, the area to be assessed is divided into a plurality of regional units. It should be understood that soil properties vary depending on geographic location, land use patterns, and environmental conditions. To help identify these differences and more accurately assess the soil, the area to be assessed can be divided into regions. Regional division divides the area to be assessed into multiple smaller, more manageable units, allowing for more detailed heavy metal testing within each unit and analysis of the heavy metal test results for each unit, thereby improving the accuracy and reliability of the assessment results.

[0032] In step S120, the heavy metal content of each of the multiple regional units is detected to obtain a plurality of regional unit heavy metal content detection results. Specifically, in an embodiment of the present application, the heavy metal content of each of the multiple regional units is detected to obtain a plurality of regional unit heavy metal content detection results, including: using I CP-MS detection technology to analyze the heavy metal content of each of the multiple regional units to obtain the plurality of regional unit heavy metal content detection results. It should be understood that in order to assess the heavy metal pollution status of the soil in a specific area to be assessed, to identify the pattern and potential risk points of heavy metal pollution in the soil, and thus to determine whether the soil is affected by heavy metal pollution, in the technical solution of the present application, the heavy metal content of each of the multiple regional units is analyzed using I CP-MS detection technology to obtain a plurality of regional unit heavy metal content detection results. I CP-MS is an analytical instrument that combines inductively coupled plasma (ICP) technology and mass spectrometry (MS). In the ICP portion of the analyzer, a high-power, high-frequency radio frequency signal applied to an inductor coil creates a high-temperature plasma within the coil. This high-temperature plasma ionizes most elements in the sample, releasing a single electron, forming monovalent positive ions. The MS portion of the instrument detects the intensity of a particular ion by selecting ions with varying mass-to-nuclear ratios to pass through, thereby analyzing and calculating the intensity of the element. ICP-MS technology is extremely sensitive, capable of detecting very low concentrations of heavy metals and supporting the simultaneous determination of multiple elements. Compared to traditional heavy metal detection methods, ICP-MS offers faster analysis speeds, typically taking only a few hours from sample processing to results, making it suitable for large-scale monitoring. Furthermore, the ICP-MS analysis process can be standardized, reducing operational errors and improving the reliability of heavy metal content detection results.

[0033] In step S130, high-dimensional embedding coding is performed on each of the multiple regional unit heavy metal content detection results to obtain multiple regional unit heavy metal content detection result semantic embedding coding vectors. Specifically, in an embodiment of the present application, high-dimensional embedding coding is performed on each of the multiple regional unit heavy metal content detection results to obtain multiple regional unit heavy metal content detection result semantic embedding coding vectors, including: using a heavy metal content embedding matrix to perform high-dimensional embedding coding on each of the multiple regional unit heavy metal content detection results to obtain the multiple regional unit heavy metal content detection result semantic embedding coding vectors. It should be understood that each of the multiple regional unit heavy metal content detection results contains semantic information about the detection of heavy metal content in the area, such as element type, concentration value, etc. Based on this, in the technical solution of the present application, a heavy metal content embedding matrix is used to perform high-dimensional embedding coding on the heavy metal content detection results of each of the multiple regional unit heavy metal content detection results to better capture the complex semantic relationships between the metals in the region, such as the distribution patterns in different regional units, and obtain semantic embedding coding vectors of the heavy metal content detection results of multiple regional units.

[0034] In step S140, query matching implicit optimization is performed on the semantic embedding encoding vector of the regional unit heavy metal content detection result corresponding to the first regional unit among the multiple regional units and the semantic embedding encoding vectors of the multiple regional unit heavy metal content detection results to obtain an implicitly optimized query matching representation of the first regional unit heavy metal content. Accordingly, considering that the semantic embedding encoding vector of the regional unit heavy metal content detection result corresponding to the first regional unit among the multiple regional units contains comprehensive information about the heavy metal content of the first regional unit, including not only the concentration information of the detected heavy metal elements but also the contextual information carried by this information after encoding, the semantic embedding encoding vectors of the multiple regional unit heavy metal content detection results are a sequence consisting of the semantic embedding encoding vectors of all regional units. Each vector in this sequence corresponds to a specific regional unit and contains a high-dimensional embedded representation of the heavy metal content detection result of that regional unit. Therefore, in order to more carefully analyze and evaluate the matching relationship between each regional unit and the entire region, the similarity or correlation between the first regional unit and other regional units is accurately identified and characterized, thereby assessing whether there is any abnormality in the soil heavy metal ecological damage of the first regional unit. In the technical solution of the present application, query matching implicit optimization is performed on a sequence of feature vectors composed of the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit among the multiple regional units and the semantic embedding coding vector of the multiple regional unit heavy metal content detection result to obtain a first regional unit heavy metal content query matching implicit optimization representation vector as the first regional unit heavy metal content query matching implicit optimization representation.

[0035] Specifically, first, the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit is embedded coded to extract the key semantics of the heavy metal content in the first regional unit, and obtain the semantic value feature vector of the first regional unit detection result. Then, considering that if the value vector, the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit (query vector) and the semantic embedding coding vectors of the heavy metal content detection results of multiple regional units (sequence of key vectors) are directly processed, the important local features in each key vector will be ignored, and the complex relationship between the data cannot be well captured. Therefore, in order to analyze the association within the local neighborhood of each region in a more fine-grained manner, the semantic embedding coding vectors of the heavy metal content detection results of the multiple regional units are further one-dimensionally convolutionally coded to capture the characteristic patterns in the local area and identify the change law of heavy metal content between adjacent regional units, and obtain a sequence of semantic association feature vectors of the local neighborhood semantics of the regional unit heavy metal content detection results, which provides support for subsequent more refined attention query matching. Then, the semantic embedding encoding vectors of the heavy metal content detection results of multiple regional units and the sequence of semantic association feature vectors of the local neighborhood of the query vector regional unit heavy metal content detection results are used as a sequence of key vectors. The semantic embedding encoding vector of the regional unit heavy metal content detection result corresponding to the first regional unit is used as the query vector, and the semantic value feature vector of the first regional unit detection result is used as the value vector. These are respectively input into the first and second multi-head attention query modules based on the converter structure to obtain a sequence of latent vectors matching the semantic granularity query of the regional unit heavy metal content detection results and a sequence of latent vectors matching the semantic neighborhood granularity query of the regional unit heavy metal content detection results. In other words, the first multi-head attention query module focuses on analyzing and extracting the features of the key vector itself, emphasizing the importance of the intrinsic properties of the key vector, thereby capturing the unique properties of each regional unit. The second multi-head attention query module focuses on the features of the key vector and its surrounding neighborhood. By considering the relationship between the key vector and the neighboring vectors, it emphasizes the importance of the mutual connection and influence between regional units, thereby better understanding the interactions and dependencies between regional units in a larger scope. Subsequently, the mean vector of the sequence of implicit vectors of semantic granularity query matching of regional unit heavy metal content detection results and the sequence of implicit vectors of semantic neighborhood granularity query matching of regional unit heavy metal content detection results are calculated, and the two mean vectors are fused to obtain a query matching representation that can fully reflect the query matching between the first area and the entire area, and the implicit optimized representation vector of the query matching of the first regional unit heavy metal content is obtained.

[0036] Specifically, Figure 3The present invention provides a flowchart of performing implicit optimization of query matching on the semantic embedding coding vector of the heavy metal content detection result of the first regional unit corresponding to the multiple regional units and the semantic embedding coding vector of the heavy metal content detection result of the multiple regional units in the method for identifying and assessing the ecological damage of heavy metals in soil based on the regional scale according to an embodiment of the present application to obtain an implicit optimization representation of the query matching of the heavy metal content of the first regional unit. Figure 3 As shown, query matching implicit optimization is performed on the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit among the multiple regional units and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results to obtain an implicitly optimized representation of the query matching of the first regional unit heavy metal content. The method includes: S141, processing the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit to obtain a semantic value feature vector of the first regional unit detection result; S142, using the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit as a query vector, the semantic value feature vector of the first regional unit detection result as a value vector, and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results as a key vector sequence, and performing a local domain-based attention gating analysis on the sequence of the query vector, the value vector, and the key vector to obtain an implicitly optimized representation of the query matching of the first regional unit heavy metal content.

[0037] More specifically, in an embodiment of the present application, the semantic embedding coding vector of the area unit heavy metal content detection result corresponding to the first area unit is processed to obtain the semantic value feature vector of the first area unit detection result, including: using the value embedding coding matrix to process the semantic embedding coding vector of the area unit heavy metal content detection result corresponding to the first area unit to obtain the semantic value feature vector of the first area unit detection result.

[0038] More specifically, Figure 4 In the method for identifying and assessing ecological damage of heavy metals in soil based on regional scale according to an embodiment of the present application, the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit is used as the query vector, the semantic value feature vector of the first regional unit detection result is used as the value vector and the semantic embedding coding vectors of the plurality of regional unit heavy metal content detection results are used as the sequence of key vectors, and the sequence of the query vector, the value vector and the key vector are subjected to attention gating analysis based on the local domain to obtain a flowchart of the implicit optimization representation of the query matching of the first regional unit heavy metal content. Figure 4As shown, the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit is used as a query vector, the semantic value feature vector of the first regional unit detection result is used as a value vector and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results are used as a sequence of key vectors, and the sequence of the query vector, the value vector and the key vector is subjected to attention gating analysis based on the local field to obtain the implicit optimization representation of the query matching of the first regional unit heavy metal content, including: S1421, one-dimensional convolution coding is performed on the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results to obtain a regional A sequence of local neighborhood semantic association feature vectors of unit heavy metal content detection results; S1422, using the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit as a query vector, the semantic value feature vector of the first regional unit detection result as a value vector and the semantic embedding coding vectors of the plurality of regional unit heavy metal content detection results as a sequence of key vectors, inputting the sequence of the query vector, the value vector and the key vector into a first multi-head attention query module based on a converter structure to obtain a sequence of semantic granularity query matching implicit vectors of the regional unit heavy metal content detection result; S1423, using the The semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit is used as a query vector, the semantic value feature vector of the first regional unit detection result is used as a value vector, and the sequence of the local neighborhood semantic association feature vectors of the regional unit heavy metal content detection result is used as a sequence of key vectors. The query vector, the value vector and the sequence of the key vector are input into the second multi-head attention query module based on the converter structure to obtain a sequence of implicit vectors for semantic neighborhood granularity query matching of the regional unit heavy metal content detection result; S1424, calculate the sequence of implicit vectors for semantic neighborhood granularity query matching of the regional unit heavy metal content detection result The column and the positional mean vector of the sequence of the semantic neighborhood granularity query matching implicit vectors of the regional unit heavy metal content detection results are obtained to obtain the regional unit heavy metal content detection result semantic granularity query matching feature vector and the regional unit heavy metal content detection result semantic neighborhood granularity query matching feature vector; S1425, fusing the regional unit heavy metal content detection result semantic granularity query matching feature vector and the regional unit heavy metal content detection result semantic neighborhood granularity query matching feature vector to obtain the first regional unit heavy metal content query matching implicit optimization representation vector as the first regional unit heavy metal content query matching implicit optimization representation.

[0039] More specifically, in an embodiment of the present application, the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit is used as a query vector, the semantic value feature vector of the first regional unit detection result is used as a value vector and the semantic embedding coding vectors of the plurality of regional unit heavy metal content detection results are used as a sequence of key vectors, and the sequence of the query vector, the value vector and the key vector are input into a first multi-head attention query module based on a converter structure to obtain a sequence of semantic granularity query matching implicit vectors of the regional unit heavy metal content detection result, including: calculating the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit and the semantic embedding coding vector of the regional unit heavy metal content detection result. The product of the transpose vectors of the vectors is performed, and the obtained semantic association matrix of the first regional unit heavy metal content detection result is divided by the square root of the length of the semantic embedding coding vector of the regional unit heavy metal content detection result by position to obtain the semantic association modulation matrix of the first regional unit heavy metal content detection result; the semantic association modulation matrix of the first regional unit heavy metal content detection result is input into the softmax function to obtain the semantic association activation matrix of the first regional unit heavy metal content detection result; the semantic association activation matrix of the first regional unit heavy metal content detection result is multiplied by the semantic value feature vector of the first regional unit detection result to obtain the semantic granularity query matching implicit vector of the regional unit heavy metal content detection result.

[0040] More specifically, in an embodiment of the present application, the semantic granularity query matching feature vector of the regional unit heavy metal content detection result and the semantic neighborhood granularity query matching feature vector of the regional unit heavy metal content detection result are fused to obtain a first regional unit heavy metal content query matching implicit optimization representation vector, including: respectively calculating the positional subtraction, positional dot product and positional addition between the semantic granularity query matching feature vector of the regional unit heavy metal content detection result and the semantic neighborhood granularity query matching feature vector of the regional unit heavy metal content detection result to obtain the regional unit heavy metal content detection result semantic granularity-neighborhood granularity difference vector, the regional unit heavy metal content detection result semantic granularity-neighborhood granularity dot product vector and the regional unit heavy metal content detection result semantic granularity-neighborhood granularity The method comprises the following steps: concatenating the semantic granularity-neighborhood granularity sum vector of the regional unit heavy metal content detection result, performing cascade processing on the semantic granularity-neighborhood granularity difference vector of the regional unit heavy metal content detection result, the semantic granularity-neighborhood granularity dot product vector of the regional unit heavy metal content detection result and the semantic granularity-neighborhood granularity sum vector of the regional unit heavy metal content detection result to obtain the semantic granularity-neighborhood granularity cascade vector of the regional unit heavy metal content detection result; performing one-dimensional convolution coding on the semantic granularity-neighborhood granularity cascade vector of the regional unit heavy metal content detection result, and performing local window-based maximum pooling processing on the obtained semantic granularity-neighborhood granularity convolution coding vector of the regional unit heavy metal content detection result to obtain the first regional unit heavy metal content query matching implicit optimization representation vector.

[0041] In the embodiment of the present application, specifically, query matching implicit optimization is performed on a sequence of feature vectors consisting of a semantic embedding coding vector of a regional unit heavy metal content detection result corresponding to a first regional unit among the multiple regional units and a semantic embedding coding vector of the multiple regional unit heavy metal content detection results, which can be expressed as:

[0042] v v =v q W q +b q

[0043] K={k1,k2,...,k i ,...,k n}

[0044] K′=Conv 1×l ({k1,k2,...,k i ,...,k n})

[0045] K′={k1′,k2′,...,k i ′,...,k m ′}

[0046]

[0047] Among them, v q is the semantic embedding coding vector of the heavy metal content detection result of the regional unit corresponding to the first regional unit, W q is the value embedding encoding matrix, b q is the value bias vector, v v is the semantic value feature vector of the detection result of the first area unit, K represents the semantic embedding coding vector of the heavy metal content detection results of the multiple area units, k1, k2, ..., k i ,...,k n is the semantic embedding coding vector of the first, second, ..., i-th, ..., n-th regional unit heavy metal content detection result in the semantic embedding coding vector of the multiple regional unit heavy metal content detection result, Conv 1×l (·) is a one-dimensional convolutional code, l is an integer multiple of the length of the semantic embedding coding vector of the heavy metal content detection result of the regional unit, K′ is a sequence of the local neighborhood semantic association feature vectors of the heavy metal content detection result of the regional unit, k1′, k2′, ..., k i ′,...,k m ′ is the first, second, ..., i-th, ..., m-th local neighborhood semantic association feature vector of the local neighborhood semantic association feature vector of the heavy metal content detection result of the regional unit, k i T and k i ' T k i and k i ′, d1 and d2 are the lengths of the semantic embedding encoding vector of the heavy metal content detection result of the regional unit and the local neighborhood semantic association feature vector of the heavy metal content detection result of the regional unit, softmax(·) is the softmax function, v ki is the semantic granularity query matching implicit vector of the i-th regional unit heavy metal content detection result in the sequence of semantic granularity query matching implicit vectors of the regional unit heavy metal content detection result, n is the number of feature vectors in the sequence of semantic granularity query matching implicit vectors of the regional unit heavy metal content detection result, v ni is the semantic neighborhood granularity query matching implicit vector of the i-th regional unit heavy metal content detection result in the sequence of semantic neighborhood granularity query matching implicit vectors of the regional unit heavy metal content detection result, m is the number of feature vectors in the sequence of semantic neighborhood granularity query matching implicit vectors of the regional unit heavy metal content detection result, v k is the semantic granularity query matching feature vector of the heavy metal content detection result of the regional unit, vn is the semantic neighborhood granularity query matching feature vector of the heavy metal content detection result of the regional unit, ⊙ and They are position-wise subtraction, position-wise multiplication, and position-wise addition, [·,·,·] is cascade processing, conv1D(·) is a one-dimensional convolutional encoding operation, MaxPool(·) is a maximum pooling operation, v p It is the implicit optimization representation vector of the query matching of the heavy metal content of the first regional unit.

[0048] In step S150, an assessment result is obtained based on the query matching implicitly optimized representation of the heavy metal content of the first regional unit. Specifically, in an embodiment of the present application, obtaining the assessment result based on the query matching implicitly optimized representation of the heavy metal content of the first regional unit includes: inputting the implicitly optimized representation vector of the query matching implicitly optimized representation of the heavy metal content of the first regional unit into a classifier-based ecological damage identification and assessment module to obtain an assessment result, wherein the assessment result indicates whether the heavy metal ecological damage of the soil in the first regional unit is abnormal. Specifically, the semantic embedding encoding vector of the regional unit heavy metal content detection result corresponding to the first regional unit and the semantic embedding encoding vectors of the heavy metal content detection results of the multiple regional units are used to perform query matching optimization, thereby automatically determining whether the heavy metal ecological damage of the soil in the first regional unit is abnormal. In this way, through automated sampling equipment and advanced detection technology, rapid on-site testing of soil samples can be achieved. Furthermore, by dividing the region into multiple regional units, subtle changes in heavy metal content in the soil can be more precisely analyzed, capturing the complex relationships between different heavy metal elements, thereby improving the accuracy of the assessment.

[0049] Here, in the technical solution of the present application, the semantic embedding coding vector of the heavy metal content detection result of the regional unit corresponding to the first regional unit among the multiple regional units is used as the query feature vector and the sequence of feature vectors composed of the semantic embedding coding vectors of the heavy metal content detection results of the multiple regional units is used as the sequence of imitation key feature vectors, and when the query feature vector and the sequence of the imitation key feature vector are implicitly optimized for query matching, it is considered that the query feature vector essentially belongs to the subset feature of the sequence of the imitation key feature vector, and there are relatively significant spatial distribution differences between the respective imitation key feature vectors in the sequence of the imitation key feature vector. Therefore, when performing attention gated query encoding, the query matching results based on the local correlation space topology difference and the feature set distribution redundancy will also have cross-domain dynamic matching differences. Therefore, it is expected to improve the detail semantic aggregation expression effect of the implicit optimization representation vector of the first regional unit heavy metal content query matching based on the cross-dynamic attention gate encoding difference.

[0050] In a preferred example, the first area unit heavy metal content query matching implicit optimization representation vector is passed through a classifier-based ecological damage identification and assessment module to obtain an assessment result, including: calculating the sum of the absolute values of each eigenvalue of the first area unit heavy metal content query matching implicit optimization representation vector to obtain the first area unit heavy metal content query matching implicit optimization and modulation value, and calculating the square root of the sum of the squares of each eigenvalue of the first area unit heavy metal content query matching implicit optimization representation vector to obtain the second area unit heavy metal content query matching implicit optimization and modulation value; performing dot-wise subtraction on the first area unit heavy metal content query matching implicit optimization representation vector and the second area unit heavy metal content query matching implicit optimization and modulation value, and then performing dot-wise multiplication with the number of eigenvalues of the first area unit heavy metal content query matching implicit optimization representation vector and the inverse of the first area unit heavy metal content query matching implicit optimization and modulation value, and taking the inverse of each eigenvalue to obtain the first area unit heavy metal content query matching implicit optimization representation vector. Metal content query matching implicit optimization phase conversion vector; after performing dot subtraction on the first area unit heavy metal content query matching implicit optimization representation vector and the first area unit heavy metal content query matching implicit optimization and modulation value, perform dot multiplication with the square root of the number of eigenvalues of the first area unit heavy metal content query matching implicit optimization representation vector and the reciprocal of the second area unit heavy metal content query matching implicit optimization and modulation value, and take the reciprocal of each eigenvalue to obtain the second area unit heavy metal content query matching implicit optimization phase conversion vector; subtract the dot product vector of the second area unit heavy metal content query matching implicit optimization phase conversion vector from the first area unit heavy metal content query matching implicit optimization phase conversion vector to obtain the optimized first area unit heavy metal content query matching implicit optimization representation vector; the optimized first area unit heavy metal content query matching implicit optimization representation vector is passed through the classifier-based ecological damage identification and assessment module to obtain the assessment result.

[0051] Here, the optimized representation of the implicit optimized representation vector of the query matching of the heavy metal content of the first regional unit is:

[0052]

[0053] v i ∈V∈R n

[0054] Wherein, V is the implicit optimization representation vector of the query matching of heavy metal content in the first regional unit, v irepresents the eigenvalue of the i-th position in the implicit optimization representation vector of the heavy metal content query matching of the first regional unit, n represents the number of eigenvalues of the implicit optimization representation vector of the heavy metal content query matching of the first regional unit, α represents the implicit optimization and modulation value of the heavy metal content query matching of the first regional unit, β represents the implicit optimization and modulation value of the heavy metal content query matching of the second regional unit, ⊙ represents the point product by position, Indicates point-wise subtraction, [·] ⊙-1 The inverse of each eigenvalue in the vector is taken, V1 represents the implicit optimization phase transition vector of the first regional unit heavy metal content query matching, V2 represents the implicit optimization phase transition vector of the second regional unit heavy metal content query matching, ω represents the weighted hyperparameter, and V′ represents the optimized first regional unit heavy metal content query matching implicit optimization representation vector.

[0055] In the preferred example, the difference in sum and modulation representation of the eigenvalues of the implicitly optimized representation vector for the first regional unit heavy metal content query match relative to the overall feature set of the first regional unit heavy metal content query match implicitly optimized representation vector is used as semantic change intensity information, and a phase-like transformation corresponding to position-based intensity modulation is performed through the different sum and modulation representation forms. By performing a spatial translation operation based on alternating stacking under the scale balance of the vector set of the implicitly optimized representation vector for the first regional unit heavy metal content query match, the aggregation enhancement of semantic change phase perception can improve the axial aggregation receptive field along the feature aggregation direction, thereby improving the perception effect of the aggregated semantics of the implicitly optimized representation vector for the first regional unit heavy metal content query match on detailed semantic changes, thereby improving the expression effect of the implicitly optimized representation vector for the first regional unit heavy metal content query match, and improving the accuracy of the assessment results obtained by the classifier-based ecological damage identification and assessment module. In this way, through automated sampling equipment and advanced detection technology, rapid on-site testing of soil samples can be achieved. Moreover, by dividing the region into multiple regional units, it is possible to more finely analyze the subtle changes in the heavy metal content in the soil and capture the complex relationship between different heavy metal elements, thereby improving the accuracy of the assessment.

[0056] In summary, a regional-scale soil heavy metal ecological damage identification and assessment method based on the embodiment of the present application is illustrated. The method divides the area to be assessed into multiple regional units, and uses ICP-MS detection technology to analyze the heavy metal content of each of the multiple regional units to obtain detection results for multiple regional units. The method also uses deep learning-based data analysis technology to embed and encode the detection results of the heavy metal content of each regional unit. In this way, based on the query matching representation between the semantic embedding features of the heavy metal content of the first regional unit and each region, it is automatically determined whether the soil heavy metal ecological damage of the first regional unit is abnormal. In this way, rapid on-site detection of soil samples can be achieved, and by more finely capturing subtle changes in the heavy metal content in the soil, it is beneficial to improve the accuracy of the assessment.

[0057] Figure 5 FIG. 1 is a system block diagram of a regional-scale soil heavy metal ecological damage identification and assessment system according to an embodiment of the present application. Figure 5As shown, according to the embodiment of the present application, the soil heavy metal ecological damage identification and assessment system 100 based on regional scale includes: an area division module 110 to be assessed, which is used to divide the area to be assessed into regions to obtain multiple regional units; a regional unit heavy metal content detection module 120, which is used to detect the heavy metal content of each regional unit in the multiple regional units to obtain multiple regional unit heavy metal content detection results; a regional unit heavy metal content detection data embedding coding module 130, which is used to perform high-dimensional embedding coding on each regional unit heavy metal content detection result in the multiple regional unit heavy metal content detection results to obtain multiple regional unit heavy metal content detection result semantic embedding coding vectors; a regional unit heavy metal content coding vector query matching optimization module 140, which is used to perform query matching implicit optimization on the regional unit heavy metal content detection result semantic embedding coding vector corresponding to the first regional unit in the multiple regional units and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results To obtain an implicitly optimized representation of the query match of the heavy metal content of the first regional unit, the method includes: an regional unit detection result semantic value feature generation unit, which is used to process the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit to obtain the first regional unit detection result semantic value feature vector; an regional unit heavy metal content query match implicitly optimized representation generation unit, which is used to use the regional unit heavy metal content detection result semantic embedding coding vector corresponding to the first regional unit as a query vector, the first regional unit detection result semantic value feature vector as a value vector and the plurality of regional unit heavy metal content detection result semantic embedding coding vectors as a sequence of key vectors, and perform a local domain-based attention gating analysis on the sequence of the query vector, the value vector and the key vector to obtain the implicitly optimized representation of the query match of the first regional unit heavy metal content; an evaluation result generation module 150, which is used to obtain an evaluation result based on the implicitly optimized representation of the query match of the first regional unit heavy metal content.

[0058] Here, those skilled in the art will appreciate that the specific functions and operations of the various units and modules in the regional-scale soil heavy metal ecological damage identification and assessment system 100 have been described in detail above. Figures 1 to 4 The description of the regional-scale soil heavy metal ecological damage identification and assessment method has been introduced in detail, and therefore, its repeated description will be omitted.

[0059] In summary, the regional-scale soil heavy metal ecological damage identification and assessment system 100 according to the embodiment of the present application is described. It divides the area to be assessed into multiple regional units, and uses ICP-MS detection technology to analyze the heavy metal content of each of the multiple regional units to obtain detection results for multiple regional units. It also uses deep learning-based data analysis technology to embed and encode the detection results of the heavy metal content of each regional unit. In this way, based on the query matching representation between the first regional unit and the semantic embedding features of the heavy metal content of each region, it is automatically determined whether the soil heavy metal ecological damage of the first regional unit is abnormal. In this way, rapid on-site detection of soil samples can be achieved, and by more finely capturing subtle changes in the heavy metal content in the soil, it is beneficial to improve the accuracy of the assessment.

[0060] As described above, the soil heavy metal ecological damage identification and assessment system 100 based on the regional scale according to the embodiment of the present application can be implemented in various wireless terminals, such as a server for soil heavy metal ecological damage identification and assessment based on the regional scale. In one example, the soil heavy metal ecological damage identification and assessment system 100 based on the regional scale according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the soil heavy metal ecological damage identification and assessment system 100 based on the regional scale 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 soil heavy metal ecological damage identification and assessment system 100 based on the regional scale can also be one of the many hardware modules of the wireless terminal.

[0061] Alternatively, in another example, the regional-scale soil heavy metal ecological damage identification and assessment system 100 and the wireless terminal may also be separate devices, and the regional-scale soil heavy metal ecological damage identification and assessment system 100 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.

Claims

1. A method for identifying and assessing soil heavy metal ecological damage based on a regional scale, characterized in that: include: Divide the area to be assessed into multiple regional units; Detecting the heavy metal content of each of the plurality of regional units to obtain heavy metal content detection results of the plurality of regional units; performing high-dimensional embedding coding on each of the plurality of regional unit heavy metal content detection results to obtain semantic embedding coding vectors of the plurality of regional unit heavy metal content detection results; The semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit among the multiple regional units and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results are implicitly optimized for query matching to obtain an implicitly optimized representation of the first regional unit heavy metal content query matching, including: processing the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit to obtain a semantic value feature vector of the first regional unit detection result; using the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit as a query vector, the semantic value feature vector of the first regional unit detection result as a value vector and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results as a key vector sequence, performing attention gating analysis based on local fields on the sequence of the query vector, the value vector and the key vector to obtain an implicitly optimized representation of the first regional unit heavy metal content query matching, including: performing one-dimensional convolution coding on the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results to obtain an implicitly optimized representation of the first regional unit heavy metal content query matching. A sequence of local neighborhood semantic association feature vectors of detection results; using the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit as a query vector, the semantic value feature vector of the first regional unit detection result as a value vector and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results as a sequence of key vectors, inputting the sequence of the query vector, the value vector and the key vector into a first multi-head attention query module based on a converter structure to obtain a sequence of semantic granularity query matching implicit vectors of regional unit heavy metal content detection results; using the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit as a query vector, the semantic value feature vector of the first regional unit detection result as a value vector and the sequence of local neighborhood semantic association feature vectors of the regional unit heavy metal content detection results as a sequence of key vectors, inputting the sequence of the query vector, the value vector and the key vector into a second multi-head attention query module based on a converter structure to obtain a sequence of semantic neighborhood granularity query matching implicit vectors of regional unit heavy metal content detection results; An evaluation result is obtained by querying and matching an implicit optimization representation based on the heavy metal content of the first regional unit.

2. The method for identifying and assessing soil heavy metal ecological damage based on regional scale according to claim 1, characterized in that: The heavy metal content of each of the multiple regional units is detected to obtain the multiple regional unit heavy metal content detection results, including: using ICP-MS detection technology to analyze the heavy metal content of each of the multiple regional units to obtain the multiple regional unit heavy metal content detection results.

3. The method for identifying and assessing soil heavy metal ecological damage based on regional scale according to claim 2, characterized in that: High-dimensional embedding coding is performed on each of the multiple regional unit heavy metal content detection results to obtain semantic embedding coding vectors of the multiple regional unit heavy metal content detection results, including: using a heavy metal content embedding matrix to perform high-dimensional embedding coding on each of the multiple regional unit heavy metal content detection results to obtain semantic embedding coding vectors of the multiple regional unit heavy metal content detection results.

4. The method for identifying and assessing soil heavy metal ecological damage based on regional scale according to claim 3, characterized in that: The semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit is processed to obtain the semantic value feature vector of the first regional unit detection result, including: using the value embedding coding matrix to process the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit to obtain the semantic value feature vector of the first regional unit detection result.

5. The method for identifying and assessing soil heavy metal ecological damage based on regional scale according to claim 4, characterized in that: Using the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit as a query vector, the semantic value feature vector of the first regional unit detection result as a value vector, and the semantic embedding coding vectors of the plurality of regional unit heavy metal content detection results as a sequence of key vectors, performing a local domain-based attention gating analysis on the sequence of the query vector, the value vector, and the key vector to obtain an implicitly optimized representation of the query matching of the first regional unit heavy metal content, including: Calculating the positional mean vector of the sequence of the semantic granularity query matching implicit vectors of the regional unit heavy metal content detection result and the sequence of the semantic neighborhood granularity query matching implicit vectors of the regional unit heavy metal content detection result to obtain the regional unit heavy metal content detection result semantic granularity query matching feature vector and the regional unit heavy metal content detection result semantic neighborhood granularity query matching feature vector; The semantic granularity query matching feature vector of the regional unit heavy metal content detection result and the semantic neighborhood granularity query matching feature vector of the regional unit heavy metal content detection result are fused to obtain a first regional unit heavy metal content query matching implicit optimization representation vector as the first regional unit heavy metal content query matching implicit optimization representation.

6. The method for identifying and assessing soil heavy metal ecological damage based on regional scale according to claim 5, characterized in that: The scale of the one-dimensional convolutional coding is equal to an integer multiple of the length of the semantic embedding coding vector of the heavy metal content detection result of the regional unit.

7. The method for identifying and assessing soil heavy metal ecological damage based on regional scale according to claim 6, characterized in that: The method uses the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit as a query vector, the semantic value feature vector of the first regional unit detection result as a value vector, and the semantic embedding coding vectors of the plurality of regional unit heavy metal content detection results as a sequence of key vectors, and inputs the sequence of the query vector, the value vector, and the key vector into a first multi-head attention query module based on a converter structure to obtain a sequence of semantic granularity query matching implicit vectors of the regional unit heavy metal content detection result, including: Calculating the product of the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit and the transposed vector of the semantic embedding coding vector of the regional unit heavy metal content detection result, and dividing the obtained semantic association matrix of the first regional unit heavy metal content detection result by the square root of the length of the semantic embedding coding vector of the regional unit heavy metal content detection result by position to obtain a semantic association modulation matrix of the first regional unit heavy metal content detection result; Inputting the semantic association modulation matrix of the heavy metal content detection results of the first regional unit into a softmax function to obtain a semantic association activation matrix of the heavy metal content detection results of the first regional unit; The semantic association activation matrix of the first area unit heavy metal content detection result is multiplied by the semantic value feature vector of the first area unit detection result to obtain the semantic granularity query matching implicit vector of the area unit heavy metal content detection result.

8. The method for identifying and assessing soil heavy metal ecological damage based on regional scale according to claim 7, characterized in that: The semantic granularity query matching feature vector of the regional unit heavy metal content detection result and the semantic neighborhood granularity query matching feature vector of the regional unit heavy metal content detection result are integrated to obtain a first regional unit heavy metal content query matching implicit optimization representation vector, including: Respectively calculating the positional subtraction, positional dot product, and positional addition between the semantic granularity query matching feature vector of the regional unit heavy metal content detection result and the semantic neighborhood granularity query matching feature vector of the regional unit heavy metal content detection result to obtain a regional unit heavy metal content detection result semantic granularity-neighborhood granularity difference vector, a regional unit heavy metal content detection result semantic granularity-neighborhood granularity dot product vector, and a regional unit heavy metal content detection result semantic granularity-neighborhood granularity sum vector; Cascade processing is performed on the difference vector of the regional unit heavy metal content detection result semantic granularity-neighborhood granularity, the dot product vector of the regional unit heavy metal content detection result semantic granularity-neighborhood granularity, and the regional unit heavy metal content detection result semantic granularity-neighborhood granularity sum vector to obtain a regional unit heavy metal content detection result semantic granularity-neighborhood granularity cascade vector; After performing one-dimensional convolution encoding on the semantic granularity-neighborhood granularity cascade vector of the heavy metal content detection result of the regional unit, the obtained semantic granularity-neighborhood granularity convolution encoding vector of the heavy metal content detection result of the regional unit is subjected to local window-based maximum pooling processing to obtain the implicit optimization representation vector of the first regional unit heavy metal content query matching.

9. The method for identifying and assessing soil heavy metal ecological damage based on regional scale according to claim 8, characterized in that: Based on the implicit optimization representation of the query matching of the heavy metal content in the first area unit, an evaluation result is obtained, including: inputting the implicit optimization representation vector of the query matching of the heavy metal content in the first area unit into the classifier-based ecological damage identification and evaluation module to obtain an evaluation result, and the evaluation result is used to indicate whether there is any abnormality in the soil heavy metal ecological damage situation in the first area unit.

10. A soil heavy metal ecological damage identification and assessment system based on regional scale, characterized by: include: An area division module for evaluation, used for dividing the area for evaluation into multiple area units; A regional unit heavy metal content detection module, configured to detect the heavy metal content of each of the plurality of regional units to obtain a plurality of regional unit heavy metal content detection results; A regional unit heavy metal content detection data embedding coding module is used to perform high-dimensional embedding coding on each of the multiple regional unit heavy metal content detection results to obtain multiple regional unit heavy metal content detection result semantic embedding coding vectors; The regional unit heavy metal content coding vector query matching optimization module is used to perform query matching implicit optimization on the regional unit heavy metal content detection result semantic embedding coding vector corresponding to the first regional unit among the multiple regional units and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results to obtain the first regional unit heavy metal content query matching implicit optimization representation, including: a regional unit detection result semantic value feature generation unit, used to process the regional unit heavy metal content detection result semantic embedding coding vector corresponding to the first regional unit to obtain the first regional unit detection result semantic value feature vector; a regional unit heavy metal content query matching implicit optimization representation generation unit, used to use the regional unit heavy metal content detection result semantic embedding coding vector corresponding to the first regional unit as the query vector, the first regional unit detection result semantic value feature vector as the value vector and the multiple regional unit heavy metal content detection result semantic embedding coding vectors as the key vector sequence, and perform attention gating analysis based on local domain on the sequence of the query vector, the value vector and the key vector to obtain the first regional unit heavy metal content query matching implicit optimization representation, including: The semantic embedding coding vector of the attribute content detection result is subjected to one-dimensional convolution coding to obtain a sequence of semantic association feature vectors of the local neighborhood of the regional unit heavy metal content detection result; the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit is used as a query vector, the semantic value feature vector of the first regional unit detection result is used as a value vector, and the semantic embedding coding vectors of the multiple regional unit heavy metal content detection results are used as a sequence of key vectors, and the sequence of the query vector, the value vector and the key vector is input into a first multi-head attention query module based on a converter structure to obtain a sequence of semantic granularity query matching implicit vectors of the regional unit heavy metal content detection result; the semantic embedding coding vector of the regional unit heavy metal content detection result corresponding to the first regional unit is used as a query vector, the semantic value feature vector of the first regional unit detection result is used as a value vector, and the sequence of semantic association feature vectors of the local neighborhood of the regional unit heavy metal content detection result is used as a sequence of key vectors, and the sequence of the query vector, the value vector and the key vector is input into a second multi-head attention query module based on a converter structure to obtain a sequence of semantic neighborhood granularity query matching implicit vectors of the regional unit heavy metal content detection result; An evaluation result generation module is used to obtain an evaluation result based on the query matching implicit optimization representation of the heavy metal content of the first regional unit.

Images