Intelligent analysis monitoring system and method based on marine environment sample

Abstract

The invention discloses an intelligent analysis monitoring system and method based on a marine environment sample, and belongs to the technical field of marine environment analysis. The method comprises the following steps: collecting marine environment samples, calling a deep learning model, outputting a marine environment sample separation scheme, and separating the samples; determining a class group code of each organism; inputting the class group codes into a preset conflict rule base, and screening potential conflict combinations; calculating a conflict coefficient; the sub-packaging operation is executed by taking conflict-free coexistence and the minimum sub-packaging number as targets; dividing the target monitoring sea area into sub-regions by adopting a grid method; matching the conflicting biological samples to the characteristic archives of the sub-regions; constructing a knowledge graph of the sea area, the biological group and the ecological characteristics; suspicious organisms in the conflicting organism samples are screened, ecological characteristics are extracted, and all sea area codes meeting the ecological characteristics in the knowledge graph are matched; and comparing the morphological characteristics of the organism with historical biological records of the candidate sea area, and determining the code of the native sea area.

Description

Technical Field

[0001] This invention relates to the field of marine environmental analysis technology, specifically to an intelligent analysis and monitoring system and method based on marine environmental samples. Background Technology

[0002] The ocean is a core component of Earth's ecosystem, playing crucial roles in regulating climate, providing biological resources, and purifying the environment. With accelerated industrialization and frequent coastal economic activities, marine pollution and ecological damage have become increasingly prominent issues. Red tide outbreaks, biodiversity loss, and transoceanic biological invasions are occurring frequently, seriously threatening marine ecological security and human life and production. Therefore, conducting precise and efficient marine environmental sample analysis and monitoring has become a core requirement for understanding marine environmental quality, tracing pollution sources, and protecting ecological balance.

[0003] Existing technologies often employ fixed parameters to separate seawater, sediment, and biological samples, failing to dynamically optimize these parameters based on marine characteristics and sample conditions. This can easily lead to residual sediment particles in seawater, damage to biological samples, or impurity residues. Traditional conflict assessment methods rely solely on rough classifications of biological groups, neglecting the actual concentration of conflicting substances and the inherent physiological tolerance thresholds of organisms. This can result in misjudgments or omissions of conflicts. Sample preparation methods often employ one-to-one individual preparation or random grouping, leading to either excessive sample quantities or, due to improper grouping, still present sample interference. Current technologies lack a systematic marine zoning and characteristic management system, and there is a lack of unified standards for verifying the marine attribution of collected samples. Cross-marine biological tracing often relies on manual comparison of historical data, resulting in low tracing efficiency. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent analysis and monitoring system and method based on marine environmental samples to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: Firstly, this application provides an intelligent analysis and monitoring method based on marine environmental samples, comprising the following steps: Marine environmental samples are collected and transferred to a temporary storage device; a deep learning model is invoked to output a marine environmental sample separation scheme, separating seawater samples, sediment samples, and biological samples. Rapid spectral detection of biological samples is performed, and combined with an AI image recognition model, the taxonomic code of each organism is determined; the taxonomic code is input into a preset conflict rule base to screen potential conflict combinations; the conflict coefficient is calculated based on the concentration of conflicting substances in potential conflict combinations and the physiological tolerance threshold of biological groups. Based on the conflict coefficient, a minimum packaging algorithm based on greedy clustering is invoked to perform packaging operations with the goal of non-conflict coexistence and the minimum number of packages. The target monitoring sea area is divided into sub-regions using a grid method, and each sub-region is assigned a code and a feature profile. The spatiotemporal data and corresponding sub-region codes when collecting conflict biological samples are extracted and matched with the feature profile of the sub-region. Construct a knowledge graph of marine areas, biological groups, and ecological characteristics; screen suspicious organisms in conflict biological samples based on the characteristics of seawater and sediment samples, extract the ecological characteristics of suspicious organisms, and match them with all marine area codes in the knowledge graph that meet the ecological characteristics; compare the morphological characteristics of the organism with the historical biological records of the candidate marine areas to determine the original marine area code.

[0006] In conjunction with the first aspect, in a first embodiment of the first aspect of this application, the step of calling a deep learning model to output a marine environmental sample separation scheme, separating seawater samples, sediment samples, and biological samples, includes: Extract basic characteristic data of marine environmental samples, including characteristic parameters of the sea area where the samples were collected, sample appearance characteristic data, and sample type prediction data; based on the basic characteristic data, supplement the separation parameters and separation effect feedback data corresponding to different sea areas and different sample combinations in historical monitoring, perform standardization processing, and form feature vectors; A pre-trained deep learning model for separating parameters is initiated. The model structure includes a feature extraction layer, a parameter prediction layer, and an effect feedback layer. Feature vectors are input into the model. The feature extraction layer identifies key influencing factors. Based on these key influencing factors and combined with the best separation cases from historical data, the parameter prediction layer outputs several sets of candidate separation parameter combinations, including seawater and sediment separation parameters and biological separation parameters. The seawater and sediment separation parameters include centrifugation speed, centrifugation time, and centrifugation temperature. The biological separation parameters include gradient sieve aperture sequence, rinsing pressure, and rinsing times. The effect feedback layer simulates the separation effect of the candidate separation parameter combinations, calculates the seawater purity compliance rate, biological survival rate, and sediment integrity score, eliminates combinations with scores below the threshold, adjusts the parameters of the remaining combinations, and outputs the optimal separation parameter set after several iterations. Based on the optimal separation parameter set, a marine environmental sample separation scheme is output, determining the operation sequence, specific parameters for each step, operation requirements, and quality judgment criteria. The scheme is executed to separate seawater samples, sediment samples, and biological samples.

[0007] In conjunction with the first aspect, in the second embodiment of the first aspect of this application, the step of performing rapid spectral detection on biological samples and combining it with an AI image recognition model to determine the taxonomic code for each organism includes: Biological samples are transferred into sterile transparent detection slides, and micro-dispersion technology is used to ensure uniform distribution of the biological individuals. Raman spectroscopy is used to focus chemical features, and fluorescence spectroscopy is used to capture specific fluorescent substances in the organism to obtain spectral data. Baseline correction and noise removal are performed on the spectral data, and chemical feature vectors are extracted. Bright-field and dark-field images of each biological individual are captured, and a pre-trained AI image recognition model is called. Through edge detection, contour extraction, and texture analysis algorithms, morphological feature parameters are extracted to form the morphological feature vector of each biological individual. Chemical and morphological feature vectors of the same biological individual are fused to form a comprehensive feature matrix, which is then input into a deep learning classification model. The model outputs the candidate class and matching confidence score corresponding to the biological individual through feature matching and probability calculation. A confidence score threshold is set. When the highest confidence score among the candidate classes meets the threshold, the model directly matches the preset class coding library and outputs the class code. When the highest confidence score does not meet the threshold, the candidate class is marked as a class to be confirmed, a temporary code is assigned, and a manual review process is triggered. After review, the code is updated to the official class code.

[0008] In conjunction with the first aspect, in the third embodiment of the first aspect of this application, the step of inputting the class group code into a preset conflict rule base and filtering potential conflict combinations includes: The system aggregates information on all biological samples with identified taxa codes to generate a taxa code set. It then calls a pre-defined conflict rule library, where each rule contains a rule ID, conflicting party A code, conflicting party B code, conflict type, and triggering condition. Rules are categorized and indexed according to conflict type. The system iterates through all possible coexisting combinations in the taxa code set. For each coexisting combination, it matches each rule in the conflict rule library, specifically: matching conflicting party A code with conflicting party B code to determine if the two taxa codes fall within the conflicting party range defined by the rule; when a code match is successful, the triggering condition is further verified. If the triggering condition is met, the combination is marked as a potential conflict combination.

[0009] In conjunction with the first aspect, in the fourth embodiment of the first aspect of this application, the step of calculating the conflict coefficient based on the concentration of conflicting substances in a potential conflict combination and the physiological tolerance threshold of biological groups includes: For potential conflicting combinations, the concentration values ​​C of conflicting substances obtained through rapid spectral detection are extracted. When multiple conflicting substances are present, their concentration values ​​are recorded separately. The set of concentration values ​​is represented as follows: , n represents the number of types of conflicting substances; from the marine biological ecological characteristics database, obtain the inherent physiological tolerance threshold T of conflicting parties A and B to each conflicting substance; For each conflicting substance, calculate its single-substance conflict coefficient K for the two types of organisms in the potential conflict combination, using the formula K=C / T; when there are n conflicting substances in the potential conflict combination, generate 2n single-substance conflict coefficients, and take the maximum value among the effective single-substance conflict coefficients as the conflict coefficient K of the potential conflict combination.

[0010] In conjunction with the first aspect, in the fifth embodiment of the first aspect of this application, the step of invoking a minimum packing algorithm based on greedy clustering based on the conflict coefficient, with the goal of minimizing the number of packings while ensuring no conflicts, includes: The number of conflicting objects and the maximum conflict coefficient of a sample are used as dual indicators. The more conflicting objects a sample has and the higher the conflict coefficient, the higher its priority. The samples are sorted from high to low priority to generate a sequence of samples to be packaged. Initialize an empty group set G and set the group numbering auto-increment rule; extract individual samples sequentially according to the sequence of samples to be packaged, start the greedy allocation process, and prioritize trying to integrate the current sample S into an existing group; iterate through each existing group in G. Verify the current sample S with The conflict relationship of all assigned samples within the sample, when S and All elements of the conflict relation matrix for all samples are 0, and If the sum of the total foundation volume of the internal sample and the foundation volume of S does not exceed the maximum bearing capacity of a single group, then the judgment is made. Compatible with S; if any sample conflicts with S or exceeds the maximum load capacity, the judgment is made. Incompatible with S, continue to verify the next existing packet; select the first compatible existing packet. Add S to this group and update. The sample list and total baseline quantity are recorded, along with group association information; when no compatible group exists, a new group is created. j is the total number of groups plus 1, and S is used as... The first sample, initialization The sample list and total basic quantity are added to the group set G; the grouping is repeated until all samples in the sample sequence to be packaged have been grouped.

[0011] In conjunction with the first aspect, in the sixth embodiment of the first aspect of this application, the extraction of spatiotemporal data and corresponding sub-region codes during the collection of conflict biological samples, and the matching of the characteristic profile of the sub-region, includes: Using sub-region codes as indexes, the feature profiles of conflict biological samples corresponding to their respective sub-regions are retrieved in batches. The core fields of the feature profiles include the sub-region code, geographic boundary, marine feature parameters, list of dominant biological communities, and background pollutant values. For conflict biological samples with complete data, the consistency between their spatiotemporal data and the corresponding sub-region code is verified. Specifically, this involves checking whether the sampling time and the time of collecting the feature profile are within the same time range threshold, whether the sampling latitude and longitude fall within the geographic boundary of the sub-region, and whether the sampling depth conforms to the water depth range of the sub-region. If there is a mismatch, the sub-region code is corrected. The sample's association identifier is then bound to the feature profile of the corresponding sub-region to generate an association table.

[0012] In conjunction with the first aspect, in the seventh embodiment of the first aspect of this application, the construction of a knowledge graph of marine areas, biological groups, and ecological characteristics includes: Marine area coding nodes are constructed, with sub-region codes as the core identifiers, and associated attributes including geographic boundaries, marine characteristic parameters, and marine area type labels. Biological group coding nodes are also constructed, with biological group codes as the core identifiers, and associated attributes including the Chinese name of the group, taxonomic level, core morphological features, and species distribution overview. Ecological feature nodes are constructed, split by feature type, with each feature node's associated attributes including feature name, quantification threshold, and feature description. The relationship type between marine areas and biological groups is typical habitat and occasional distribution, with associated attributes including data source. The relationship type between biological groups and ecological feature edges is adaptation, tolerance, and preference, with associated attributes including feature confidence and applicable life cycle. The relationship type between marine areas and ecological feature edges is inherent environmental characteristics, with associated attributes including feature mean and annual fluctuation range.

[0013] In conjunction with the first aspect, in the eighth embodiment of the first aspect of this application, the step of screening suspected organisms in conflict organism samples based on the characteristics of seawater and sediment samples, extracting the ecological characteristics of the suspected organisms, and matching all marine area codes in the knowledge graph that satisfy the ecological characteristics includes: For each conflict organism sample, it is determined whether the collected water temperature falls within the organism's adaptive water temperature range, whether the collected salinity falls within the organism's salinity tolerance range, whether the sediment type is consistent with the organism's substrate preference, and whether the sediment organic carbon content meets the organism's requirement threshold. If any one of these conditions is not met, the organism is marked as a suspected organism. The ecological characteristics of the suspected organism are extracted to form a complete ecological characteristic set F. The ecological characteristic nodes in the knowledge graph are traversed, and all ecological characteristic nodes that completely match the characteristics in F are selected. The marine area coding nodes associated with the matching ecological characteristic nodes are tracked, and all relevant marine area codes are collected to form a marine area coding list.

[0014] Secondly, this application provides an intelligent analysis and monitoring system based on marine environmental samples, comprising: The sample collection and temporary storage module collects marine environmental samples and transfers them to the temporary storage device. The intelligent sample separation module calls a deep learning model to output a marine environmental sample separation scheme, separating seawater samples, sediment samples and biological samples. The biological sample conflict analysis module performs rapid spectral detection on biological samples and, combined with an AI image recognition model, determines the taxonomic code for each organism. The taxonomic code is then input into a preset conflict rule base to screen potential conflict combinations. Based on the concentration of conflicting substances in the potential conflict combinations and the physiological tolerance threshold of the biological groups, the conflict coefficient is calculated. The intelligent packaging execution module uses the conflict coefficient to call the minimum packaging algorithm based on greedy clustering to perform packaging operations with the goal of avoiding conflict and minimizing the number of packages. The marine area zoning and characteristic management module uses a grid method to divide the target monitoring marine area into sub-regions, and assigns a code and characteristic file to each sub-region. The conflict sample association data processing module extracts the spatiotemporal data and corresponding sub-region codes when collecting conflict biological samples, and matches the characteristic profile of the sub-region. The knowledge graph construction module builds knowledge graphs of marine areas, biological groups, and ecological characteristics. The suspected organism tracing and marine area matching module screens suspected organisms in conflict organism samples based on the characteristics of seawater and sediment samples, extracts the ecological characteristics of suspected organisms, matches them with all marine area codes in the knowledge graph that meet the ecological characteristics, and compares the morphological characteristics of the organism with the historical biological records of the candidate marine areas to determine the original marine area code.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention uses a deep learning model to dynamically optimize separation parameters and outputs the optimal separation scheme by combining marine characteristics and sample status, so as to achieve accurate separation of seawater, sediment and biological samples. This ensures the purity of seawater, the integrity of sediment and the survival rate of organisms, and does not require human intervention, thus greatly improving separation efficiency and stability.

[0016] 2. This invention achieves rapid and accurate determination of biological group codes through spectral detection and AI image recognition; it calculates the conflict coefficient based on the concentration of conflicting substances and the biological physiological tolerance threshold, making the conflict judgment more scientific and avoiding misjudgment and omission; it adopts a greedy clustering minimum packing algorithm to minimize the number of packings while ensuring non-conflicting coexistence, thus balancing detection accuracy and resource utilization.

[0017] 3. This invention adopts a grid method to standardize the division of marine sub-regions and establish characteristic archives to achieve structured management of marine information; it constructs a ternary knowledge graph and combines biological morphological characteristics with historical records of candidate marine areas to form a dual traceability logic of ecological feature matching and morphological verification, which greatly improves the accuracy and efficiency of determining the native marine areas of cross-sea organisms. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the steps of an intelligent analysis and monitoring method based on marine environmental samples according to the present invention; Figure 2 This is a system structure diagram of an intelligent analysis and monitoring system based on marine environmental samples according to the present invention. Detailed Implementation

[0019] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example: Figures 1-2 As shown, the present invention provides a technical solution: like Figure 1 As shown, this application provides an intelligent analysis and monitoring method based on marine environmental samples, including the following steps: Step S100: Collect marine environmental samples and transfer them to a temporary storage device; call the deep learning model to output a marine environmental sample separation scheme, separating seawater samples, sediment samples and biological samples; Specifically, basic characteristic data of marine environmental samples are extracted, including characteristic parameters of the sea area where the samples were collected, sample appearance characteristics data, and sample type prediction data. Based on the basic characteristic data, separation parameters and separation effect feedback data corresponding to different sea areas and different sample combinations in historical monitoring are supplemented, and standardized processing is performed to form feature vectors. A pre-trained deep learning model for separating parameters is initiated. The model structure includes a feature extraction layer, a parameter prediction layer, and an effect feedback layer. Feature vectors are input into the model. The feature extraction layer identifies key influencing factors. Based on these key influencing factors and combined with the best separation cases from historical data, the parameter prediction layer outputs several sets of candidate separation parameter combinations, including seawater and sediment separation parameters and biological separation parameters. The seawater and sediment separation parameters include centrifugation speed, centrifugation time, and centrifugation temperature. The biological separation parameters include gradient sieve aperture sequence, rinsing pressure, and rinsing times. The effect feedback layer simulates the separation effect of the candidate separation parameter combinations, calculates the seawater purity compliance rate, biological survival rate, and sediment integrity score, eliminates combinations with scores below the threshold, adjusts the parameters of the remaining combinations, and outputs the optimal separation parameter set after several iterations. Based on the optimal separation parameter set, a marine environmental sample separation scheme is output, determining the operation sequence, specific parameters for each step, operation requirements, and quality judgment criteria. The scheme is executed to separate seawater samples, sediment samples, and biological samples.

[0021] In one specific embodiment, step S100 is executed at a monitoring point near the coast of the East China Sea (30.5°N, 122.3°E), and the specific process is as follows: The collected marine environmental samples were surface mixed samples (containing seawater, muddy sediments, phytoplankton, benthic shellfish, and other organisms). After being transferred to an anti-adsorption temporary storage device, basic characteristic data were extracted: the characteristic parameters of the collected sea area were water temperature 22℃, salinity 28‰, and low background pollutant values; the appearance characteristics of the samples were turbidity 15 NTU, light brown color, and a small amount of suspended particles; the sample type was predicted to be sediment particles with a size mostly <63μm, and organisms with a size concentrated between 50-200μm.

[0022] Supplement the historical monitoring data of this sea area (separation parameters of similar samples: centrifugation speed 3000 r / min, duration 10 min, gradient sieve aperture 200 μm → 50 μm, corresponding separation effect: seawater purity 98%, biological survival rate 80%), and after standardization, form feature vectors to input into the pre-trained deep learning model.

[0023] The model feature extraction layer identified key influencing factors as sediment particle size (<63μm) and biological size (50-200μm); the parameter prediction layer output three sets of candidate parameters: Group 1 (centrifugation at 3200r / min for 8min, gradient sieve from 200μm to 50μm, rinsing pressure 0.1MPa for 3 times), Group 2 (centrifugation at 3000r / min for 10min, same sieve and rinsing parameters as Group 1), and Group 3 (centrifugation at 2800r / min for 12min, gradient sieve from 150μm to 50μm).

[0024] The simulation results from the feedback layer showed that: Group 1 had a seawater purity of 99.2%, a biological survival rate of 88%, and a sediment integrity of 92%; Group 2 had corresponding values ​​of 98.5%, 85%, and 90%; and Group 3 had values ​​of 97%, 90%, and 88%. The thresholds were set as seawater purity ≥ 98%, biological survival rate ≥ 85%, and sediment integrity ≥ 88%, thus eliminating Group 3. The centrifugation time for Group 1 was fine-tuned to 9 minutes, and a second simulation yielded a seawater purity of 99.3%, a biological survival rate of 89%, and a sediment integrity of 93%, which was determined to be the optimal parameter set.

[0025] After the separation procedure was implemented, the test results were as follows: the turbidity of the seawater sample was 3.2 NTU (purity 99.3%), the survival rate of the biological sample was 89% (no obvious damage), and there was no seawater clumping in the sediment (integrity 93%), all of which met the quality judgment standards.

[0026] Step S200: Perform rapid spectral detection on biological samples, and combine with AI image recognition model to determine the taxonomic code of each organism; input the taxonomic code into a preset conflict rule library to screen potential conflict combinations; calculate the conflict coefficient based on the concentration of conflicting substances in potential conflict combinations and the physiological tolerance threshold of biological groups. Specifically, biological samples are transferred into sterile transparent detection slides, and micro-dispersion technology is used to ensure uniform distribution of the biological individuals. Raman spectroscopy is used to focus chemical features, and fluorescence spectroscopy is used to capture specific fluorescent substances in the organism to obtain spectral data. Baseline correction and noise removal are performed on the spectral data to extract chemical feature vectors. Bright-field and dark-field images of each biological individual are captured, and a pre-trained AI image recognition model is called to extract morphological feature parameters through edge detection, contour extraction, and texture analysis algorithms to form a morphological feature vector for each biological individual. Chemical and morphological feature vectors of the same biological individual are fused to form a comprehensive feature matrix, which is then input into a deep learning classification model. The model outputs the candidate class and matching confidence score corresponding to the biological individual through feature matching and probability calculation. A confidence score threshold is set. When the highest confidence score among the candidate classes meets the threshold, the model directly matches the preset class coding library and outputs the class code. When the highest confidence score does not meet the threshold, the candidate class is marked as a class to be confirmed, a temporary code is assigned, and a manual review process is triggered. After review, the code is updated to the official class code.

[0027] Furthermore, the information of all biological samples with identified taxa codes is aggregated to generate a taxa code set. A preset conflict rule library is invoked, where each rule contains a rule ID, conflicting party A code, conflicting party B code, conflict type, and triggering condition. The rules are categorized and indexed according to conflict type. All possible coexisting combinations in the taxa code set are traversed. For each coexisting combination, all rules in the conflict rule library are matched one by one. Specifically, conflicting party A code and conflicting party B code are matched to determine whether the two taxa codes in the combination fall within the conflicting party range defined by the rule. When the code match is successful, the triggering condition is further verified. If the triggering condition is met, the combination is marked as a potential conflict combination.

[0028] Furthermore, for potential conflicting combinations, the concentration values ​​C of conflicting substances obtained through rapid spectral detection are extracted. When multiple conflicting substances are present, their concentration values ​​are recorded separately, and the set of concentration values ​​is represented as follows: , n represents the number of types of conflicting substances; from the marine biological ecological characteristics database, obtain the inherent physiological tolerance threshold T of conflicting parties A and B to each conflicting substance; For each conflicting substance, calculate its single-substance conflict coefficient K for the two types of organisms in the potential conflict combination, using the formula K=C / T; when there are n conflicting substances in the potential conflict combination, generate 2n single-substance conflict coefficients, and take the maximum value among the effective single-substance conflict coefficients as the conflict coefficient K of the potential conflict combination.

[0029] In one specific embodiment, the separated biological sample contains three types of organisms: diatoms, oyster larvae, and copepods. The specific process is as follows: After the biological samples were transferred to sterile slides for dispersion, Raman spectroscopy was used to detect the characteristic peaks of the siliceous shells of diatoms and the characteristic peaks of the proteins of oyster larvae. Fluorescence spectroscopy captured the chlorophyll-specific fluorescence of diatoms. Chemical feature vectors were extracted after data processing. Simultaneously, bright-field / dark-field images were captured. The AI ​​model used edge detection to determine that the body length of diatoms was 20 μm, the body length of oyster larvae was 80 μm, and the body length of copepods was 50 μm. Morphological feature vectors such as contours and textures were extracted.

[0030] After dual feature fusion, the data is input into the classification model, and the output diatom candidate groups have a confidence level of 94%, oyster larvae 92%, and copepods 93% (all above the 85% threshold). These are matched with codes PP-D-003, PB-O-012, and PZ-C-005, respectively, to form a group coding set.

[0031] The code is input into the conflict rule base, and the combination {PP-D-003&PB-O-012} is obtained by matching the "metabolite interference" rule (the trigger condition is the simultaneous presence of two types of organisms), and is marked as a potential conflict combination. The concentration of the conflicting substance algal toxin in this combination is detected as C=0.4μg / L. The tolerance thresholds for diatoms TD=0.8μg / L and oyster larvae TO=0.5μg / L are retrieved, and the single-substance conflict coefficients KD=0.4 / 0.8=0.5 and KO=0.4 / 0.5=0.8 are calculated. The maximum value of 0.8 is taken as the conflict coefficient of this combination.

[0032] Step S300: Based on the conflict coefficient, call the minimum packaging algorithm based on greedy clustering to perform the packaging operation with the goal of non-conflict coexistence and the minimum number of packages; use the grid method to divide the target monitoring sea area into sub-regions, and assign a code and feature file to each sub-region; extract the spatiotemporal data and corresponding sub-region codes when collecting conflict biological samples, and match the feature file of the sub-region. Specifically, the number of conflicting objects and the maximum value of the conflict coefficient are used as dual indicators. The more conflicting objects and the higher the conflict coefficient, the higher the priority of the sample. The samples are sorted from high to low priority to generate a sequence of samples to be packaged. Initialize an empty group set G and set the group numbering auto-increment rule; extract individual samples sequentially according to the sequence of samples to be packaged, start the greedy allocation process, and prioritize trying to integrate the current sample S into an existing group; iterate through each existing group in G. Verify the current sample S with The conflict relationship of all assigned samples within the sample, when S and All elements of the conflict relation matrix for all samples are 0, and If the sum of the total foundation volume of the internal sample and the foundation volume of S does not exceed the maximum bearing capacity of a single group, then the judgment is made. Compatible with S; if any sample conflicts with S or exceeds the maximum load capacity, the judgment is made. Incompatible with S, continue to verify the next existing packet; select the first compatible existing packet. Add S to this group and update. The sample list and total baseline quantity are recorded, along with group association information; when no compatible group exists, a new group is created. j is the total number of groups plus 1, and S is used as... The first sample, initialization The sample list and total basic quantity are added to the group set G; the grouping is repeated until all samples in the sample sequence to be packaged have been grouped.

[0033] Furthermore, using sub-region codes as indexes, the characteristic profiles of the corresponding sub-regions for conflict biological samples are retrieved in batches. The core fields of the characteristic profiles include sub-region codes, geographical boundaries, marine characteristic parameters, a list of dominant biological communities, and background pollutant values. For conflict biological samples with complete data, the consistency between their spatiotemporal data and the corresponding sub-region codes is verified. Specifically, this involves checking whether the sampling time and the time of collecting the characteristic profiles are within the same time range threshold, whether the sampling latitude and longitude fall within the geographical boundaries of the sub-region, and whether the sampling depth conforms to the water depth range of the sub-region. If there is a mismatch, the sub-region codes are corrected. The sample's association identifier is then bound to the characteristic profile of the corresponding sub-region to generate an association table.

[0034] In one specific embodiment, the monitoring area near the coast of the East China Sea (30.3°-30.6°N, 122.1°-122.4°E) is used as the scenario. The conflict organism samples are diatoms (PP-D-003), oyster larvae (PB-O-012), and copepods (PZ-C-005) identified in step S200. The specific implementation process is as follows: First, the conflict relationships and conflict coefficients of each sample were clarified: Oyster larvae (PB-O-012) only conflicted with diatoms (PP-D-003), with 1 conflicting object and a conflict coefficient of 0.8; Diatoms (PP-D-003) only conflicted with oyster larvae, with 1 conflicting object and a conflict coefficient of 0.8; Copepods (PZ-C-005) had no conflict with the other two types, with 0 conflicting objects and a conflict coefficient of 0. Using "number of conflicting objects + maximum conflict coefficient" as dual indicators, the sorting priority was: oyster larvae (priority 1), diatoms (priority 1, followed by them in ascending order of coding), copepods (priority 2), generating the sample sequence to be packaged: [PB-O-012, PP-D-003, PZ-C-005].

[0035] Initialize an empty group set G={}, set the group numbering auto-increment rule (G1, G2, G3...), the maximum capacity of a single group is 30mL (the critical amount to meet the needs of subsequent detection), and the basic volume of the three types of biological samples is 10mL / sample.

[0036] Process the first sample PB-O-012: Set G is empty and there are no existing groups. Create a new group G1 and assign PB-O-012 to G1. The sample list of G1 is {PB-O-012}, with a total basic volume of 10 mL. Record the group association information as "no conflicting combination".

[0037] Process the second sample PP-D-003: Traverse the existing group G1 and check for conflict relationships—PP-D-003 conflicts with PB-O-012 in G1 (the conflict relationship matrix element is 1), and is determined to be incompatible; there are no other existing groups, create a new group G2, and assign PP-D-003 to G2. The sample list of G2 is {PP-D-003}, with a total basic volume of 10mL. Record the group association information "no conflict combination".

[0038] Processing the third sample PZ-C-005: First, verify the existing group G1 to confirm that PZ-C-003 and PB-O-012 in G1 have no conflict (the conflict relationship matrix element is 0), and the total basic volume of G1 is 10mL + 10mL of PZ-C-005 = 20mL ≤ 30mL, so it is considered compatible; PZ-C-005 is assigned to G1, and the sample list of G1 is updated to {PB-O-012, PZ-C-005}, with a total basic volume of 20mL, and the group association information "no conflict combination" is recorded.

[0039] All samples were grouped, resulting in two dispensing schemes: G1 (oyster larvae + copepods, 20 mL dispensing volume) and G2 (diatoms, 10 mL dispensing volume). The samples were transferred to anti-adsorption sterile containers according to the scheme, labeled with the relevant identifier and group number (e.g., "20240615-EA-N-028-B-001-G1"), and entered into the database simultaneously.

[0040] A 10km×10km grid method was used to divide the target monitoring sea area into 6 sub-regions, which were coded according to "sea area abbreviation-region type-serial number" (e.g., EA-N-028, EA=East China Sea, N=nearshore, 028=serial number). A characteristic profile was established for each sub-region. The core fields of the characteristic profile for the sub-region EA-N-028 corresponding to the conflict biological samples are: geographical boundary (30.4°-30.5°N, 122.2°-122.3°E), marine characteristic parameters (water temperature 21-23℃, salinity 27-29‰, water depth 15-20m, sediment type muddy), list of dominant biological communities (copepods, benthic mollusks, phytoplankton), and background pollutant values ​​(heavy metal content <0.01mg / L, organic pollutants <0.005mg / L).

[0041] The unique association identifiers of the three types of conflict organism samples were retrieved, and the spatiotemporal data of the collection were extracted: Oyster larvae were collected at 10:30 on June 15, 2024, at latitude 30.45°N, longitude 122.25°E, and at a sampling depth of 18m; Diatoms were collected at the same time as oyster larvae, at latitude 30.46°N, longitude 122.26°E, and at a sampling depth of 17m; Copepods were collected at the same time, at latitude 30.44°N, longitude 122.24°E, and at a sampling depth of 19m; The sub-region code corresponding to all three is EA-N-028.

[0042] The spatiotemporal data consistency was verified using the characteristic profiles—the sampling time (June) fell within the "summer monitoring time range (May-July)" recorded in the profiles; the sampling latitude and longitude were all within the geographical boundaries of EA-N-028; and the sampling depth (17-19m) conformed to the water depth range of 15-20m for the sub-region, confirming a complete match between the spatiotemporal data and the sub-region code. The association identifiers of the three types of samples were linked to the characteristic profiles of EA-N-028 to generate a "conflict organism sample-sub-region-characteristic profile" association table, clarifying the marine environmental characteristics and dominant biological communities of the sub-region to which each type of sample belongs, providing a basis for subsequent screening of suspected organisms.

[0043] Step S400: Construct a knowledge graph of marine areas, biological groups, and ecological characteristics; screen suspicious organisms in conflict biological samples based on the characteristics of seawater and sediment samples, extract the ecological characteristics of suspicious organisms, and match all marine area codes that meet the ecological characteristics in the knowledge graph; compare the morphological characteristics of the organism with the historical biological records of the candidate marine areas to determine the original marine area code.

[0044] Specifically, marine area coding nodes are constructed, with sub-region codes as the core identifiers, and associated attributes including geographic boundaries, marine characteristic parameters, and marine area type labels; biological group coding nodes are constructed, with biological group codes as the core identifiers, and associated attributes including the Chinese name of the group, taxonomic level, core morphological features, and species distribution overview; ecological feature nodes are constructed, split according to feature type, and each feature node's associated attributes include feature name, quantification threshold, and feature description; the relationship type between marine areas and biological groups is typical habitat marine areas and occasional distribution marine areas, and the associated attribute includes data source; the relationship type between biological groups and ecological feature edges is adaptation, tolerance, and preference, and the associated attributes include feature confidence and applicable life cycle; the relationship type between marine areas and ecological feature edges is inherent environmental characteristics, and the associated attributes include feature mean and annual fluctuation range.

[0045] Furthermore, for each conflict organism sample, it is determined whether the collected water temperature falls within the organism's adaptive water temperature range, whether the collected salinity falls within the organism's salinity tolerance range, whether the sediment type is consistent with the organism's substrate preference, and whether the sediment organic carbon content reaches the organism's required threshold. If any one of these conditions is not met, the organism is marked as a suspected organism. The ecological characteristics of the suspected organism are extracted to form a complete ecological characteristic set F. The ecological characteristic nodes in the knowledge graph are traversed, and all ecological characteristic nodes that completely match the characteristics in F are selected. The marine area coding nodes associated with the matching ecological characteristic nodes are tracked, and all relevant marine area codes are collected to form a marine area coding list.

[0046] In one specific embodiment, continuing the monitoring scenario of the East China Sea nearshore area, the conflict organism samples were previously identified diatoms (PP-D-003), oyster larvae (PB-O-012), and copepods (PZ-C-005), corresponding to the collection sub-region EA-N-028 (muddy area of ​​the East China Sea nearshore area). The specific implementation process is as follows: Marine area coding nodes: Two key marine areas were selected to construct nodes: ① Node ID: S001, core identifier EA-N-028, associated attributes are geographic boundary (30.4°-30.5°N, 122.2°-122.3°E), marine characteristic parameters (annual average water temperature 20-24℃, salinity 26-29‰, sediment type muddy, organic carbon content 0.8-1.2%), and marine area type label "East China Sea nearshore muddy area"; ② Node ID: S002, core identifier EA-O-035, associated attributes are geographic boundary (29.8°-30.0°N, 123.0°-123.2°E), marine characteristic parameters (annual average water temperature 21-25℃, salinity 30-33‰, sediment type sandy, organic carbon content 0.3-0.6%), and marine area type label "East China Sea offshore sandy area".

[0047] Biological group coding nodes: ① Node ID: B001, core identifier PP-D-003, associated attributes are: Chinese name of group "Skeletonema costatum", taxonomic level "Diatoms - Centroclass - Skeletona", core morphological characteristics "unicellular, body length 20-30μm, siliceous shell with radial texture", species distribution overview "common in the nearshore area of ​​East China Sea and central Yellow Sea"; ② Node ID: B002, core identifier PB-O-012, associated attributes are: Chinese name of group "Crassostrea iliata larva", taxonomic level "Malacus - Bivalvia - The species is classified as follows: ① Oyster (genus); ② Core morphological characteristics: "Shell length 50-100μm, shell surface with fine growth lines, larval stage is planktonic"; ③ Node ID: B003, core identifier PZ-C-005, associated attributes: Chinese name of group "Chinese water flea", taxonomic level "Arthropoda-Copepoda-Genus", core morphological characteristics: "Body length 40-60μm, distinct body segments, tail fork with bristles", ④ Species distribution: "Widely distributed in the nearshore and offshore areas of the East China Sea".

[0048] Ecological characteristic nodes: ① Node ID: E001, characteristic name "adapted water temperature", quantification threshold "18-25℃", characteristic description "suitable water temperature range for biological growth and reproduction"; ② Node ID: E002, characteristic name "salinity tolerance range", quantification threshold "25-32‰", characteristic description "critical salinity range for biological survival"; ③ Node ID: E003, characteristic name "sediment preference", quantification threshold "sandy", characteristic description "preferred sediment type for biological habitat"; ④ Node ID: E004, characteristic name "organic carbon requirement threshold", quantification threshold "≥0.5%", characteristic description "minimum content of organic carbon in sediments required for biological growth".

[0049] Marine Area-Biological Group Boundaries: S001 (EA-N-028) and B001 (PP-D-003) are "typical habitats", with the associated attribute "Data Source: 2020-2024 East China Sea nearshore monitoring records"; S001 and B003 (PZ-C-005) are "typical habitats", with the associated attribute "Data Source: 2020-2024 East China Sea nearshore monitoring records"; S002 (EA-O-035) and B002 (PB-O-012) are "typical habitats", with the associated attribute "Data Source: 2020-2024 East China Sea offshore monitoring records"; S001 and B002 are "occasional distribution areas", with the associated attribute "Data Source: 2019 East China Sea nearshore anomaly records".

[0050] Biological Groups - Ecological Characteristics Edges: B001 and E001 (adapted to water temperature) are "adapted", with the associated attribute "characteristic confidence level 95%, applicable life cycle: full life cycle"; B002 and E003 (substrate preference) are "preference", with the associated attribute "characteristic confidence level 98%, applicable life cycle: juvenile to adult"; B003 and E002 (salinity tolerance) are "tolerant", with the associated attribute "characteristic confidence level 92%, applicable life cycle: full life cycle".

[0051] Marine Area-Ecological Characteristics: S001 and E003 (substrate preference - sandy) are "inherent environmental characteristics", with the associated attribute "mean value of characteristics: muddy content 92%, annual fluctuation range ±3%"; S002 and E003 are "inherent environmental characteristics", with the associated attribute "mean value of characteristics: sandy content 95%, annual fluctuation range ±2%".

[0052] The measured seawater and sediment data of sub-region EA-N-028 were retrieved: the water temperature was 22℃ and the salinity was 28‰ (both of which are within the average range of the characteristic profile of this region); the sediment type was muddy and the organic carbon content was 1.0% (higher than the lower limit of the average range of 0.8-1.2% recorded in the profile).

[0053] The compatibility of the three types of organisms with the collection environment was verified one by one: Diatoms (PP-D-003): Adaptable water temperature 18-25℃ (sampling temperature 22℃ meets the criteria), salinity tolerance 25-32‰ (28℃ meets the criteria), preferred muddy substrate (sampling sediment is muddy meets the criteria), organic carbon requirement ≥0.5% (1.0% meets the criteria). If all four criteria are met, it is not marked as a suspicious organism.

[0054] Copepods (PZ-C-005): Adaptable water temperature 15-26℃ (22℃ meets the requirements), salinity tolerance 24-33‰ (28℃ meets the requirements), no restrictions on substrate preference (muddy substrate meets the requirements), organic carbon requirement ≥0.4% (1.0% meets the requirements). If all four conditions are met, it will not be marked as a suspected organism.

[0055] Oyster larva (PB-O-012): Adaptable water temperature 18-25℃ (22℃ meets the requirements), salinity tolerance 25-32‰ (28℃ meets the requirements), preferred sandy substrate (collected sediment was muddy, does not meet the requirements), organic carbon requirement ≥0.5% (1.0% meets the requirements). One key condition is not met, so it is marked as "suspected organism".

[0056] The complete ecological characteristic set F of oyster larvae (PB-O-012) was extracted: adapted water temperature 18-25℃, salinity tolerance 25-32‰, preference for sandy substrate, organic carbon requirement ≥0.5%, breeding water temperature ≥20℃, and filter feeding. The ecological characteristic nodes of the knowledge graph were traversed, and nodes (E001, E002, E003, E004) that perfectly matched F were selected. The associated marine area codes of these nodes were then traced. E003 (bottom sediment preference - sandy) is associated with sea area S002 (EA-O-035, sandy area in the outer East China Sea), with the associated attribute "95% sandy content"; E001 (adapted to water temperatures of 18-25℃) is associated with sea areas S002 (average annual water temperature of 21-25℃) and S005 (YS-N-019, nearshore sandy area of ​​the Yellow Sea, average annual water temperature of 17-24℃). Further verification of other feature matching: the average salinity of S005 is 24-26‰, which is close to the lower limit of salinity tolerance for oyster larvae of 25‰, and the annual fluctuation may be lower than 25‰; while the salinity of S002 of 28-31‰ completely covers the tolerance range. Finally, the candidate sea area coding list is selected as: [EA-O-035, YS-N-019].

[0057] Morphological characteristics of oyster larvae (PB-O-012) were extracted: shell length 80 μm, radial striations on the shell surface (2 μm spacing), 3 minute spines on the ventral margin, and the larval shell was transparent and pigment-free. Historical biological records from the candidate marine area were retrieved. Monitoring records from EA-O-035 (outer East China Sea) in 2023 showed that the morphology of juvenile oysters in this area was "shell length 75-85μm, radial striations spaced 2-3μm apart, 3-4 spines on the ventral margin, and transparent larval shell", which perfectly matched the morphological characteristics of the suspected organism, with a matching degree of 92%. The 2023 record for YS-N-019 (nearshore Yellow Sea) shows that the oyster larvae in this area have a shell length of 60-70 μm, radial striations spaced 1-2 μm apart, and two spines on the ventral margin, exhibiting significant morphological differences with a 75% match rate. Based on comprehensive analysis, EA-O-035 is determined to be the native marine area code for this oyster larvae. The association attribute of "PB-O-012-typical habitat-EA-O-035" in the knowledge graph is simultaneously updated to supplement the data source for this source tracing.

[0058] like Figure 2 As shown, this application provides an intelligent analysis and monitoring system based on marine environmental samples, including: The sample collection and temporary storage module collects marine environmental samples and transfers them to the temporary storage device. The intelligent sample separation module calls a deep learning model to output a marine environmental sample separation scheme, separating seawater samples, sediment samples and biological samples. The biological sample conflict analysis module performs rapid spectral detection on biological samples and, combined with an AI image recognition model, determines the taxonomic code for each organism. The taxonomic code is then input into a preset conflict rule base to screen potential conflict combinations. Based on the concentration of conflicting substances in the potential conflict combinations and the physiological tolerance threshold of the biological groups, the conflict coefficient is calculated. The intelligent packaging execution module uses the conflict coefficient to call the minimum packaging algorithm based on greedy clustering to perform packaging operations with the goal of avoiding conflict and minimizing the number of packages. The marine area zoning and characteristic management module uses a grid method to divide the target monitoring marine area into sub-regions, and assigns a code and characteristic file to each sub-region. The conflict sample association data processing module extracts the spatiotemporal data and corresponding sub-region codes when collecting conflict biological samples, and matches the characteristic profile of the sub-region. The knowledge graph construction module builds knowledge graphs of marine areas, biological groups, and ecological characteristics. The suspected organism tracing and marine area matching module screens suspected organisms in conflict organism samples based on the characteristics of seawater and sediment samples, extracts the ecological characteristics of suspected organisms, matches them with all marine area codes in the knowledge graph that meet the ecological characteristics, and compares the morphological characteristics of the organism with the historical biological records of the candidate marine areas to determine the original marine area code.

[0059] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A smart analysis and monitoring method based on marine environmental samples, characterized in that, Includes the following steps: Collect marine environmental samples and transfer them to a temporary storage device; The deep learning model is invoked to output a marine environmental sample separation scheme, which separates seawater samples, sediment samples, and biological samples. Rapid spectral detection is performed on biological samples, and combined with an AI image recognition model, the taxonomic code of each organism is determined; the taxonomic code is input into a preset conflict rule base to screen potential conflict combinations; The conflict coefficient is calculated based on the concentration of conflicting substances in potential conflict combinations and the physiological tolerance threshold of biological groups. Based on the conflict coefficient, the minimum packaging algorithm based on greedy clustering is invoked to perform the packaging operation with the goal of non-conflict coexistence and the minimum number of packages. The target monitoring area is divided into sub-regions using a grid method, and each sub-region is assigned a code and a characteristic profile; the spatiotemporal data and corresponding sub-region codes are extracted when collecting conflict biological samples, and the characteristic profiles of the sub-regions are matched; Construct a knowledge graph of marine areas, biological groups, and ecological characteristics; screen suspicious organisms in conflict biological samples based on the characteristics of seawater and sediment samples, extract the ecological characteristics of suspicious organisms, and match all marine area codes in the knowledge graph that meet the ecological characteristics. The morphological characteristics of the organism are compared with historical biological records of the candidate sea area to determine the original sea area code.

2. The intelligent analysis and monitoring method based on marine environmental samples according to claim 1, characterized in that, The process involves calling a deep learning model to output a marine environmental sample separation scheme, which separates seawater samples, sediment samples, and biological samples, including: Extract basic characteristic data of marine environmental samples, including characteristic parameters of the sea area where the samples were collected, sample appearance characteristic data, and sample type prediction data; based on the basic characteristic data, supplement the separation parameters and separation effect feedback data corresponding to different sea areas and different sample combinations in historical monitoring, perform standardization processing, and form feature vectors; A pre-trained deep learning model for separating parameters is initiated. The model structure includes a feature extraction layer, a parameter prediction layer, and an effect feedback layer. Feature vectors are input into the model. The feature extraction layer identifies key influencing factors. Based on these key influencing factors and combined with the best separation cases from historical data, the parameter prediction layer outputs several sets of candidate separation parameter combinations, including seawater and sediment separation parameters and biological separation parameters. The seawater and sediment separation parameters include centrifugation speed, centrifugation time, and centrifugation temperature. The biological separation parameters include gradient sieve aperture sequence, rinsing pressure, and rinsing times. The effect feedback layer simulates the separation effect of the candidate separation parameter combinations, calculates the seawater purity compliance rate, biological survival rate, and sediment integrity score, eliminates combinations with scores below the threshold, adjusts the parameters of the remaining combinations, and outputs the optimal separation parameter set after several iterations. Based on the optimal separation parameter set, a marine environmental sample separation scheme is output, determining the operation sequence, specific parameters for each step, operation requirements, and quality judgment criteria. The scheme is executed to separate seawater samples, sediment samples, and biological samples.

3. The intelligent analysis and monitoring method based on marine environmental samples according to claim 1, characterized in that, The process of performing rapid spectral detection on biological samples, combined with an AI image recognition model, to determine the taxonomic code for each organism includes: Biological samples are transferred into sterile transparent detection slides, and micro-dispersion technology is used to ensure uniform distribution of the biological individuals. Raman spectroscopy is used to focus chemical features, and fluorescence spectroscopy is used to capture specific fluorescent substances in the organism to obtain spectral data. Baseline correction and noise removal are performed on the spectral data, and chemical feature vectors are extracted. Bright-field and dark-field images of each biological individual are captured, and a pre-trained AI image recognition model is called. Through edge detection, contour extraction, and texture analysis algorithms, morphological feature parameters are extracted to form the morphological feature vector of each biological individual. Chemical and morphological feature vectors of the same biological individual are fused to form a comprehensive feature matrix, which is then input into a deep learning classification model. The model outputs the candidate class and matching confidence score corresponding to the biological individual through feature matching and probability calculation. A confidence score threshold is set. When the highest confidence score among the candidate classes meets the threshold, the model directly matches the preset class coding library and outputs the class code. When the highest confidence score does not meet the threshold, the candidate class is marked as a class to be confirmed, a temporary code is assigned, and a manual review process is triggered. After review, the code is updated to the official class code.

4. The intelligent analysis and monitoring method based on marine environmental samples according to claim 1, characterized in that, The step of inputting the group code into a preset conflict rule base and filtering potential conflict combinations includes: The system aggregates information on all biological samples with identified taxa codes to generate a taxa code set. It then calls a pre-defined conflict rule library, where each rule contains a rule ID, conflicting party A code, conflicting party B code, conflict type, and triggering condition. Rules are categorized and indexed according to conflict type. The system iterates through all possible coexisting combinations in the taxa code set. For each coexisting combination, it matches each rule in the conflict rule library, specifically: matching conflicting party A code with conflicting party B code to determine if the two taxa codes fall within the conflicting party range defined by the rule; when a code match is successful, the triggering condition is further verified. If the triggering condition is met, the combination is marked as a potential conflict combination.

5. The intelligent analysis and monitoring method based on marine environmental samples according to claim 1, characterized in that, The calculation of the conflict coefficient based on the concentration of conflicting substances in a potential conflict combination and the physiological tolerance threshold of biological groups includes: For potential conflicting combinations, the concentration values ​​C of conflicting substances obtained through rapid spectral detection are extracted. When multiple conflicting substances are present, their concentration values ​​are recorded separately. The set of concentration values ​​is represented as follows: , n represents the number of types of conflicting substances; from the marine biological ecological characteristics database, obtain the inherent physiological tolerance threshold T of conflicting parties A and B to each conflicting substance; For each conflicting substance, calculate its single-substance conflict coefficient K for the two types of organisms in the potential conflict combination, using the formula K=C / T; when there are n conflicting substances in the potential conflict combination, generate 2n single-substance conflict coefficients, and take the maximum value among the effective single-substance conflict coefficients as the conflict coefficient K of the potential conflict combination.

6. The intelligent analysis and monitoring method based on marine environmental samples according to claim 1, characterized in that, The process involves invoking a greedy clustering-based minimum packaging algorithm based on the conflict coefficient. The goal is to achieve conflict-free coexistence with the minimum number of packages. The packaging operation includes: The number of conflicting objects and the maximum conflict coefficient of a sample are used as dual indicators. The more conflicting objects a sample has and the higher the conflict coefficient, the higher its priority. The samples are sorted from high to low priority to generate a sequence of samples to be packaged. Initialize an empty group set G and set the group numbering auto-increment rule; extract individual samples sequentially according to the sequence of samples to be packaged, start the greedy allocation process, and prioritize trying to integrate the current sample S into an existing group; iterate through each existing group in G. Verify the current sample S with The conflict relationship of all assigned samples within the sample, when S and All elements of the conflict relation matrix for all samples are 0, and If the sum of the total foundation volume of the internal sample and the foundation volume of S does not exceed the maximum bearing capacity of a single group, then the judgment is made. Compatible with S; if any sample conflicts with S or exceeds the maximum load capacity, the judgment is made. Incompatible with S, continue to verify the next existing packet; select the first compatible existing packet. Add S to this group and update. The sample list and total baseline quantity are recorded, along with group association information; when no compatible group exists, a new group is created. j is the current total number of groups plus 1, and S is used as... The first sample, initialization The sample list and total basic quantity are added to the group set G; the grouping is repeated until all samples in the sample sequence to be packaged have been grouped.

7. The intelligent analysis and monitoring method based on marine environmental samples according to claim 1, characterized in that, The extraction of spatiotemporal data and corresponding sub-region codes from conflict biological samples, and matching the characteristic profile of that sub-region, includes: Using sub-region codes as indexes, the feature profiles of conflict biological samples corresponding to their respective sub-regions are retrieved in batches. The core fields of the feature profiles include the sub-region code, geographic boundary, marine feature parameters, list of dominant biological communities, and background pollutant values. For conflict biological samples with complete data, the consistency between their spatiotemporal data and the corresponding sub-region code is verified. Specifically, this involves checking whether the sampling time and the time of collecting the feature profile are within the same time range threshold, whether the sampling latitude and longitude fall within the geographic boundary of the sub-region, and whether the sampling depth conforms to the water depth range of the sub-region. If there is a mismatch, the sub-region code is corrected. The sample's association identifier is then bound to the feature profile of the corresponding sub-region to generate an association table.

8. The intelligent analysis and monitoring method based on marine environmental samples according to claim 1, characterized in that, The construction of the knowledge graph of marine areas, biological groups, and ecological characteristics includes: Marine area coding nodes are constructed, with sub-region codes as the core identifiers, and associated attributes including geographic boundaries, marine characteristic parameters, and marine area type labels. Biological group coding nodes are also constructed, with biological group codes as the core identifiers, and associated attributes including the Chinese name of the group, taxonomic level, core morphological features, and species distribution overview. Ecological feature nodes are constructed, split by feature type, with each feature node's associated attributes including feature name, quantification threshold, and feature description. The relationship type between marine areas and biological groups is typical habitat and occasional distribution, with associated attributes including data source. The relationship type between biological groups and ecological feature edges is adaptation, tolerance, and preference, with associated attributes including feature confidence and applicable life cycle. The relationship type between marine areas and ecological feature edges is inherent environmental characteristics, with associated attributes including feature mean and annual fluctuation range.

9. The intelligent analysis and monitoring method based on marine environmental samples according to claim 1, characterized in that, The process involves screening for suspicious organisms in conflict organism samples based on the characteristics of seawater and sediment samples, extracting the ecological features of the suspicious organisms, and matching them with all marine area codes in the knowledge graph that satisfy those ecological features, including: For each conflict organism sample, it is determined whether the collected water temperature falls within the organism's adaptive water temperature range, whether the collected salinity falls within the organism's salinity tolerance range, whether the sediment type is consistent with the organism's substrate preference, and whether the sediment organic carbon content meets the organism's requirement threshold. If any one of these conditions is not met, the organism is marked as a suspected organism. The ecological characteristics of the suspected organism are extracted to form a complete ecological characteristic set F. The ecological characteristic nodes in the knowledge graph are traversed, and all ecological characteristic nodes that completely match the characteristics in F are selected. The marine area coding nodes associated with the matching ecological characteristic nodes are tracked, and all relevant marine area codes are collected to form a marine area coding list.

10. An intelligent analysis and monitoring system based on marine environmental samples, using the intelligent analysis and monitoring method based on marine environmental samples according to any one of claims 1-9, characterized in that, include: The sample collection and temporary storage module collects marine environmental samples and transfers them to the temporary storage device. The intelligent sample separation module calls a deep learning model to output a marine environmental sample separation scheme, separating seawater samples, sediment samples and biological samples. The biological sample conflict analysis module performs rapid spectral detection on biological samples and, combined with an AI image recognition model, determines the taxonomic code for each organism. The taxonomic code is then input into a preset conflict rule base to screen potential conflict combinations. Based on the concentration of conflicting substances in the potential conflict combinations and the physiological tolerance threshold of the biological groups, the conflict coefficient is calculated. The intelligent packaging execution module uses the conflict coefficient to call the minimum packaging algorithm based on greedy clustering to perform packaging operations with the goal of avoiding conflict and minimizing the number of packages. The marine area zoning and characteristic management module uses a grid method to divide the target monitoring marine area into sub-regions, and assigns a code and characteristic file to each sub-region. The conflict sample association data processing module extracts the spatiotemporal data and corresponding sub-region codes when collecting conflict biological samples, and matches the characteristic profile of the sub-region. The knowledge graph construction module builds knowledge graphs of marine areas, biological groups, and ecological characteristics. The suspected organism tracing and marine area matching module screens suspected organisms in conflict organism samples based on the characteristics of seawater and sediment samples, extracts the ecological characteristics of suspected organisms, and matches them with all marine area codes in the knowledge graph that meet the ecological characteristics. The morphological characteristics of the organism are compared with historical biological records of the candidate sea area to determine the original sea area code.

Images