Microplastic particle automatic detection system and method based on microscopic infrared imaging

Abstract

The invention relates to the technical field of micro-plastic detection, and discloses a micro-plastic particle automatic detection system and method based on microscopic infrared imaging. The method includes creating and maintaining a dynamically updated microscopic infrared data map. And performing structural defect scanning and problem area positioning on the map. According to a scanning result, various virtual intervention path sequences are simulated and generated in the data atlas. And the system receives an external intervention instruction, and screens and fuses the virtual path sequence according to the instruction. And converting the fused intervention path sequence into a physical control instruction, driving microscopic infrared hardware to execute detection operation, and reconstructing internal connection of the microscopic infrared data atlas according to an execution result. According to the method, dynamic optimization and man-machine intelligent cooperation of the detection process are realized, and the automation level and the analysis precision of micro-plastic detection are improved.

Description

Technical Field

[0001] This invention relates to the field of microplastic detection technology, specifically to an automated detection system and method for microplastic particles based on microscopic infrared imaging. Background Technology

[0002] In the field of microplastic detection, microscopic infrared imaging technology is an important means of identifying and characterizing micron-sized plastic particles. Current techniques typically rely on single or a limited number of infrared spectral scans of a sample, resulting in static, isolated datasets or images for analysis. This approach treats each detection as an independent event, lacking organic connections between the data and failing to form a coherent understanding of the sample's state or allow for historical tracing. When it is necessary to re-examine specific areas of the same sample, monitor changes, or conduct in-depth analysis, static data cannot provide effective guidance.

[0003] Current automated testing processes largely rely on preset, fixed scanning paths or manual operation based entirely on visual positioning. Preset programs lack dynamic responsiveness to the actual condition of the sample and cannot intelligently adjust testing strategies based on real-time analysis results. Manual intervention, on the other hand, is inefficient, highly subjective, and difficult to coordinate effectively with automated systems. This results in insufficient flexibility in the testing process, making it difficult to optimize and accurately allocate testing resources when dealing with complex sample areas or areas requiring special attention. Summary of the Invention

[0004] The purpose of this invention is to provide an automated detection system and method for microplastic particles based on microscopic infrared imaging, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an automated detection method for microplastic particles based on microscopic infrared imaging, the method comprising: Create and maintain a dynamically updated microscopic infrared data spectrum; Structural defect scanning and problem area localization are performed on the aforementioned microscopic infrared data spectrum; Based on the defect scanning results, multiple virtual intervention path sequences are simulated and generated within the microscopic infrared data spectrum; Receive external intervention instructions, and filter and fuse the virtual intervention path sequence according to the external intervention instructions; The fused intervention path sequence is converted into physical control commands to drive the microscopic infrared hardware to perform detection operations, and the internal connections of the microscopic infrared data spectrum are reconstructed based on the execution results.

[0006] Preferably, the creation and maintenance of a dynamically updated microscopic infrared data spectrum includes the following specific processes: Import multiple batches of microscopic infrared imaging records from a historical sample database. The microscopic infrared imaging records include particle spectral sequences, spatial coordinate trajectories, and background noise patterns. Layered deconstruction processing is performed on the multiple batches of microscopic infrared imaging records to generate the bottom-level data units of the spectrum; the layered deconstruction processing includes: Feature locking is performed on the particle spectral sequence to extract discriminative peak frequency combinations and absorption intensity profiles, thus forming a spectral fingerprint; Perform trajectory analysis on the spatial coordinate trajectory, encode the physical displacement path of the particle into a spatial topological chain, and record the distance vector between the path node and the adjacent particle; Perform pattern stripping on the background noise pattern, calculate and store the background fluctuation baseline and noise spectrum characteristics of each imaging region; Based on the spectral fingerprint, the spatial topological chain, and the background fluctuation baseline, a multi-layer graph structure is instantiated in memory as the microscopic infrared data spectrum; In the multi-layer graph structure, each data unit is defined as a particle entity, and each particle entity is assigned a set of state parameters, including the particle entity's spectral fingerprint encoding, spatial topological chain identifier, background fluctuation baseline value, and a reliability index generated by an internal evaluation mechanism. The internal evaluation mechanism operates in real time, dynamically refreshing the reliability index of each particle entity by periodically comparing newly input microscopic infrared imaging data with the spectral fingerprint code and spatial topological chain identifier stored in the multi-layer graph structure.

[0007] Preferably, the structural defect scanning and problem area localization of the microscopic infrared data spectrum is specifically implemented through a spectrum self-inspection process, which includes the following steps: Initiate a graph integrity check, traverse all particle entities in the multi-layer graph structure, check whether there is data loss in the spectral fingerprint code of each particle entity, whether the spectral fingerprint code has an unacceptable deviation from the standard spectral library, and check whether the spatial topology chain identifier is completely closed. Based on the verification results, calculate the integrity score for each particle entity; Initiate association consistency verification, analyze the interconnected particle entities in the multi-layer graph structure, check whether the spectral fingerprint encoding similarity of the particle entities at both ends of the connection is lower than the connection strength threshold, and check whether the spatial relationship represented by the connection contradicts the background fluctuation baseline value. Based on the verification results, calculate the contradiction score for each connection between entities; Based on the completeness score and the contradiction score, region clustering is performed in the multi-layer graph structure; From the clustered regions, extract the regions where the integrity score of all particle entities is lower than the integrity threshold and the contradiction score of all connections is higher than the contradiction threshold, and mark these regions as structurally fragile regions. From the clustered regions, further extract regions containing at least one particle entity with an extremely low integrity score or an extremely high contradiction score, and whose average reliability index is lower than the reliability threshold. Mark these regions as data fuzzy areas. Generate a defect report that details the spatial coordinate range of all structurally vulnerable areas and the reliability index distribution of all data ambiguity areas.

[0008] Preferably, the step of simulating and generating multiple virtual intervention path sequences within the microscopic infrared data spectrum based on the defect scanning results specifically includes the following simulation steps: Read the defect report and initialize different simulation environments for the structurally vulnerable area and the data ambiguity area, respectively; For the structurally vulnerable region, the goal of the simulation environment is to repair data integrity by simulating and generating a virtual path sequence for spectral re-measurement; the process is as follows: In the multi-layer graph structure, starting from the particle entities in the structurally fragile area, the search extends outward to find adjacent particle entities with a reliability index higher than the preset standard. Simulate the transmission of spectral data stream from a high-reliability particle entity to a target particle entity within a structurally vulnerable region, and calculate the improvement in the integrity score of the target particle entity's spectral fingerprint encoding for each simulated transmission; Based on the magnitude of the boost value and the complexity of the transmission path, multiple virtual data supplementation paths are planned. Each path includes a specific source particle entity, a target particle entity, the data type of the transmission, and the expected completeness score after the simulation is executed. The set of these paths is denoted as the spectral retest virtual path sequence. For the aforementioned data ambiguity area, the goal of the simulation environment is to clarify data contradictions and simulate the generation of associated reconstructed virtual path sequences; the process is as follows: In the multi-layer graph structure, all inter-entity connections within the data ambiguity region that are considered to have excessively high contradiction scores are suspended; Outside the data ambiguity region, search for particle entity pairs with potentially similar spectral fingerprint codes but currently not connected; Simulate the establishment of connections between these new particle entity pairs and evaluate the impact of the new connections on the average inconsistency score of the entire data fuzzy region. Based on the degree of decrease in the contradiction score, the optimal combination of new connections is selected, and a list of virtual operations containing the steps of disconnecting old connections and establishing new connections is generated. This list is recorded as the virtual path sequence of association reconstruction.

[0009] Preferably, the method further includes the steps of prioritizing and detecting conflicts in the spectral re-measurement virtual path sequence and the associated reconstruction virtual path sequence, specifically: A comprehensive utility value is calculated for each of the aforementioned spectral retest virtual path sequences. The comprehensive utility value is jointly determined by the total integrity score that the spectral retest virtual path sequence can improve, the estimated resource consumption required to execute the path sequence, and the current reliability index of the target particle entity. A comprehensive clarification value is calculated for each of the aforementioned associated reconstructed virtual path sequences. The comprehensive clarification value is jointly determined by the total contradiction score that the associated reconstructed virtual path sequence can reduce, the impact of the path sequence execution on the overall connectivity of the graph, and the background fluctuation baseline value involving particle entities. Sort all the virtual path sequences for spectral remeasurement in descending order of their overall utility value to generate a remeasurement sequence queue; Sort all the associated reconstructed virtual path sequences in descending order of their comprehensive clarification values ​​to generate a reconstructed sequence queue; Perform virtual conflict detection: Check whether the top-ranked spectral remeasurement virtual path sequences and associated reconstructed virtual path sequences are competing for the same micro-infrared hardware resources or the same spectral data region; If a virtual conflict is detected, the sorting position of the conflict path sequence is adjusted, or it is broken down into multiple conflict-free subsequences, and the retest sequence queue and the reconstruction sequence queue are updated. The final output consists of a conflict-free retest sequence queue and a reconstruction sequence queue, which serve as a set of virtual intervention path sequences to be selected.

[0010] Preferably, the process of receiving external intervention instructions and filtering and fusing the virtual intervention path sequence based on the external intervention instructions is accomplished through the following interaction and decision-making mechanism: The system provides an interactive interface for receiving external intervention commands, which include commands to specify a key detection area, a target particle type, or a detection efficiency mode. When a command to designate a key detection area is received, all path sequences whose starting point, ending point, or path coverage overlaps with the designated key detection area are selected from the set of virtual intervention path sequences. When a command for a specified target particle type is received, all path sequences whose spectral fingerprint codes match the characteristics of the specified target particle type are selected from the set of virtual intervention path sequences. When a specified detection efficiency mode instruction is received, the sorting weights of the retest sequence queue and the reconstruction sequence queue are dynamically adjusted according to the requirements of the efficiency mode, and path sequences with high comprehensive utility values ​​or short execution times are selected first. The selected multiple path sequences are merged. The fusion process includes: Identify common operation steps between different path sequences and merge these common operation steps into a single operation. Examine the operational order dependencies between different path sequences, and generate a linear, acyclic total sequence of operational steps based on the dependencies; In the overall sequence of operation steps, necessary status checkpoints and data synchronization instructions are inserted to ensure that the output data of the preceding steps can be correctly read by the subsequent steps. Finally, a unified and executable fusion intervention path is generated. The fusion intervention path is a data structure that includes specific microscopic infrared hardware operation commands, target data addresses, and expected intermediate states.

[0011] Preferably, the process of converting the fused intervention path sequence into physical control commands to drive the microscopic infrared hardware to perform detection operations is implemented through an instruction interpretation and execution engine, which operates according to the following steps: The fusion intervention path is analyzed, and each virtual operation step is mapped to one or more underlying physical control commands; The physical control commands include, but are not limited to: commands to control the microscope stage to move to the target coordinates, commands to control the infrared spectrometer to focus and acquire spectra within a specified wavenumber range, and commands to control the image sensor to adjust the integration time to adapt to background noise. Establish an instruction execution queue and send the physical control instructions to the corresponding microscopic infrared hardware controller in sequence; After each physical control command is executed, the hardware status feedback data and the initially collected data fragments are read in real time through sensors; The state feedback data is compared with the predefined expected intermediate states in the fusion intervention path; If the comparison result is within the allowable error range, then continue to execute the next physical control command; If the comparison result exceeds the allowable error range, an adaptive adjustment subprocess is triggered: The adaptive adjustment subprocess selects a compensation strategy from a preset strategy library based on the type and magnitude of the deviation between the current state and the expected state. The compensation strategy may include resending the previous physical control command, fine-tuning the parameters of the next physical control command, or inserting a new calibration command. After executing the selected compensation strategy, the state comparison is performed again until the conditions for continuing execution are met or the maximum number of retries is reached.

[0012] Preferably, the reconstructing of the internal connections of the microscopic infrared data spectrum based on the execution result specifically includes the following spectrum update steps: After all the physical control commands have been executed, all the final result data returned by the hardware is collected. The final result data includes newly acquired infrared spectra, updated particle space images, and environmental parameter readings. The final result data is preprocessed, including spectral denoising, image registration, and data format standardization, to generate standardized new data blocks. In the multi-layer graph structure, locate the target particle entity or target connection relationship corresponding to the new data block; The original state parameters of the target particle entity are replaced or supplemented with the standardized new data blocks. Specifically, the new infrared spectral data is used to generate a new spectral fingerprint code after feature locking, which is then used to update the spectral fingerprint code of the particle entity. After trajectory analysis, the new particle space image generates a new spatial topology chain, which is used to update the spatial topology chain identifier of the particle entity. New environmental parameter readings are used to update the background fluctuation baseline values; Recalculate the reliability index of all affected particle entities, as well as the contradiction score of all affected connections; Based on the updated reliability index and contradiction score, the local or global connectivity of the multi-layer graph structure is optimized; The optimization includes: deleting isolated particle entities with a reliability index below the deletion threshold and no important connections; establishing new connections between particle entities with highly similar spectral fingerprint codes and spatial topological chain identifiers indicating proximity; weakening or breaking low-quality connections with inconsistency scores consistently above the break threshold; After optimization, the microscopic infrared data spectrum enters a stable, updated state, awaiting the next detection cycle.

[0013] Preferably, the reliability index generated by the internal evaluation mechanism is dynamically generated through a multi-factor iterative calculation model, which performs the following steps: The basic factors of the particle entity are obtained, including the matching degree between the spectral fingerprint encoding and the standard spectral library, the continuity and rationality of the spatial topological chain identification, and the stability of the background fluctuation baseline value. Obtain the derivation factors of the particle entity, including the number of connections of the particle entity in the multi-layer graph structure, the average reliability index of the adjacent particle entities connected to it, and the frequency of the particle entity's historical state parameters being successfully updated. A dynamic weighting coefficient is assigned to each basic factor and derived factor, and the dynamic weighting coefficient is periodically adjusted according to the problem area type marked in the defect report; When the number of structurally vulnerable areas increases in the defect reports, the weighting coefficient of the continuity and rationality factors of the spatial topology chain identifier is increased; When the number of ambiguous areas in the defect report increases, the weighting coefficient of the matching factor between the spectral fingerprint code and the standard spectral library is increased. A raw reliability score is obtained by multiplying all basic and derived factors by their corresponding dynamic weight coefficients and summing the results. The original reliability score is input into a standardization function and mapped to a numerical range between zero and one. The mapped value is the current reliability index of the particle entity. Each time the microscopic infrared data spectrum is updated, the reliability index of the relevant particle entity is recalculated, thereby realizing the dynamic refresh of the reliability index.

[0014] Preferably, when the processor executes the computer program, it implements the steps of the automated detection method for microplastic particles based on microscopic infrared imaging as described in any of the above-mentioned methods.

[0015] Compared with the prior art, the beneficial effects of the present invention are: Unlike conventional methods that construct static datasets, this approach creates and maintains a dynamically updated microscopic infrared data spectrum that integrates spatial, spectral, and historical information. As a continuously evolving digital model, it comprehensively records the state changes of the sample and the correlation data of each detection. The analysis process can be based on a comprehensive data foundation with temporal dimensions and causal relationships, enabling defect identification and localization to move beyond the instantaneous information of a single scan. Instead, it allows for trend judgment and correlation analysis based on historical data, improving the comprehensiveness of defect detection and the reliability of analytical conclusions.

[0016] To address the limitations of fixed detection paths, multiple virtual intervention path sequences targeting identified problem areas are simulated within the data map. External commands are then introduced to filter and fuse these virtual sequences. This combines the computational power of automated algorithms with the experience and judgment of operators. Based on real-time defect scanning results, the system can quickly propose multiple feasible automated detection solutions. Operators can then optimize these solutions or inject specific targets using external commands, according to actual needs. The final fused physical control commands possess the efficiency of automatic planning while incorporating human decision-making, driving hardware to perform efficient and targeted detection operations. This achieves intelligent automation processes and deep collaboration between humans and machines. Attached Figure Description

[0017] Figure 1 This is a schematic diagram illustrating the working principle of the automated detection method for microplastic particles based on microscopic infrared imaging as described in this invention. Figure 2 A flowchart for creating and maintaining dynamically updated microscopic infrared data spectra; Figure 3 A flowchart for implementing defect scanning and problem area localization in the map self-inspection process; Figure 4 The effect curve of instruction execution status monitoring and compensation strategy; Figure 5 This is a heatmap showing the distribution of spectral fingerprint matching degree and peak number. Detailed Implementation

[0018] 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.

[0019] Please see Figure 1 This invention provides an automated detection method for microplastic particles based on microscopic infrared imaging. The method includes: creating a dynamically updated microscopic infrared data spectrum as the core data model. This spectrum is continuously maintained and updated, and undergoes structural defect scanning to locate problem areas. Based on the defect scanning results, multiple virtual intervention path sequences are simulated and generated within the spectrum. The system receives external intervention commands and filters and fuses the simulated virtual intervention path sequences according to the commands. The fused intervention path sequences are converted into specific physical control commands to drive the microscopic infrared hardware system to perform the actual detection operation. After the hardware execution is completed, the system reconstructs the internal connections of the microscopic infrared data spectrum based on the execution results, thereby completing a complete detection and spectrum update cycle.

[0020] In one embodiment of the present invention, see [reference] Figure 2A dynamically updated microscopic infrared data spectrum is created and maintained, and the processing procedure is as follows: Multiple batches of microscopic infrared imaging records are imported from a historical sample library. These records contain particle spectral sequences, spatial coordinate trajectories, and background noise patterns. Layered deconstruction processing is performed on these batches of microscopic infrared imaging records to generate the bottom-level data units of the spectrum. Layered deconstruction processing includes feature locking of the particle spectral sequences, extracting discriminative peak frequency combinations and absorption intensity profiles to construct spectral fingerprints. Trajectory parsing is performed on the spatial coordinate trajectories, encoding the physical displacement paths of the particles into spatial topological chains, and recording the distance vectors between path nodes and adjacent particles. Pattern stripping is performed on the background noise patterns, calculating and storing the background fluctuation baseline and noise spectrum features of each imaging region. Based on the spectral fingerprints, spatial topological chains, and background fluctuation baselines, a multi-layer graph structure is instantiated in memory as the microscopic infrared data spectrum. In the multi-layer graph structure, each data unit is defined as a particle entity, and each particle entity is assigned a set of state parameters, including the particle entity's spectral fingerprint code, spatial topological chain identifier, background fluctuation baseline value, and a reliability index generated by an internal evaluation mechanism. The internal evaluation mechanism operates in real time, dynamically refreshing the reliability index of each particle entity by periodically comparing the newly input microscopic infrared imaging data with the spectral fingerprint code and spatial topological chain identifier stored in the multi-layer graph structure.

[0021] The reliability index generated by the internal evaluation mechanism is dynamically generated through a multi-factor iterative calculation model. The multi-factor iterative calculation model performs the following steps: First, it obtains the basic factors of the particle entity, including the matching degree between the spectral fingerprint code and the standard spectral library, the continuity and rationality of the spatial topological chain identifier, and the stability of the background fluctuation baseline value. Second, it obtains the derived factors of the particle entity, including the number of connections of the particle entity in the multi-layer graph structure, the average reliability index of its connected neighboring particle entities, and the frequency of successful updates of the particle entity's historical state parameters. Third, it assigns a dynamic weight coefficient to each basic and derived factor, which is periodically adjusted according to the type of problem area marked in the defect report. When the number of structurally vulnerable areas in the defect report increases, the weight coefficient of the continuity and rationality factor of the spatial topological chain identifier is increased. When the number of data ambiguity areas in the defect report increases, the weight coefficient of the matching degree factor between the spectral fingerprint code and the standard spectral library is increased. Fourth, it multiplies all basic and derived factors by their corresponding dynamic weight coefficients and sums them to obtain an original reliability score. Fifth, it inputs the original reliability score into a standardization function, mapping it to a numerical range between zero and one; the mapped value is the current reliability index of the particle entity. Each time the microscopic infrared data spectrum is updated, the reliability index of the relevant particle entity is recalculated, thereby realizing the dynamic refresh of the reliability index.

[0022] In practice, a dynamically updated microscopic infrared data spectrum is created and maintained. The process begins with importing multiple batches of microscopic infrared imaging records from a historical sample library. These records contain particle spectral sequences captured in previous detections, spatial coordinate trajectories of particles on the sample stage, and background noise patterns in the imaging area. In some embodiments, the particle spectral sequences contain thousands of data points, the spatial coordinate trajectories consist of a series of continuous two-dimensional or three-dimensional coordinate points, and the background noise patterns record the infrared background signal characteristics under specific conditions. A hierarchical deconstruction process is then performed on the imported batches of microscopic infrared imaging records to generate the structured underlying data units required for constructing the spectrum.

[0023] The hierarchical deconstruction process comprises three parallel analysis steps. The first step involves feature locking of the particle spectral sequence. The feature locking algorithm scans each spectral segment, identifying and extracting distinctive peak frequency combinations and their corresponding absorption intensity profiles. These features collectively constitute a unique spectral fingerprint identifying the particle. The second step involves trajectory analysis of the spatial coordinate trajectory. The trajectory analysis algorithm encodes the recorded physical displacement path of the particle, transforming it into a spatial topological chain describing positional relationships. This chain records not only key nodes along the path but also the distance vector information between each node and its neighboring particle nodes. In practice, the trajectory analysis algorithm involves in-depth processing of the particle's spatial coordinate trajectory. It first reads particle physical displacement path data imported from a historical sample database, which consists of a series of continuous spatial coordinate points. The algorithm identifies key nodes along the path by scanning the coordinate sequence. These key nodes typically correspond to locations where the displacement direction changes significantly or where velocity reaches extreme values, thus capturing characteristic turning points in particle motion. For each identified key node, the trajectory analysis algorithm calculates its Euclidean distance vector with neighboring particle nodes. This distance vector information includes magnitude and direction, used to quantify the spatial relationship between nodes. Next, the algorithm encodes key nodes and their associated distance vectors into a linear chain structure, namely a spatial topological chain, in path order. Each node entry in the spatial topological chain stores its own coordinates and vector pointers to adjacent nodes. The resulting spatial topological chain, as the core spatial attribute of the particle entity, is integrated into a multi-layer graph structure, providing a foundation for subsequent defect scanning and map optimization. The third process is to perform pattern stripping on the background noise patterns. The pattern stripping operation calculates and stores the background fluctuation baseline value for each specific imaging region, while simultaneously analyzing and storing the spectral characteristics of the noise signal in that region.

[0024] In practical implementation, within the instantiated multi-layer graph structure, each data unit is explicitly defined as a particle entity. Each particle entity is assigned a set of state parameters, which constitute a complete digital profile of the particle entity. These state parameters specifically include the particle entity's spectral fingerprint encoding, spatial topological chain identifier, background fluctuation baseline value of the region where the particle entity is located, and a reliability index calculated in real-time by an internal evaluation mechanism. This internal evaluation mechanism runs continuously in the system background. It works by periodically comparing the newly input microscopic infrared imaging data with the spectral fingerprint encoding and spatial topological chain identifier stored in the multi-layer graph structure, dynamically refreshing the reliability index value of each particle entity, thus keeping the microscopic infrared data spectrum in a dynamically updated state.

[0025] In some embodiments, the reliability index generated by the internal evaluation mechanism is dynamically generated through a multi-factor iterative calculation model. The execution steps of the multi-factor iterative calculation model are as follows: First, the basic factors of the particle entity are obtained. These basic factors include the matching degree between the particle entity's spectral fingerprint encoding and the standard spectral library, the continuity and rationality of the particle entity's spatial topological chain identification, and the stability of the background fluctuation baseline value of the region where the particle entity is located. Second, the derived factors of the particle entity are obtained. These derived factors include the number of connections of the particle entity in the multi-layer graph structure, the average reliability index of adjacent particle entities connected to the particle entity, and the frequency with which the particle entity's historical state parameters have been successfully updated. Third, a dynamic weight coefficient is assigned to each basic factor and each derived factor. This dynamic weight coefficient is not fixed but is periodically adjusted according to the problem area type marked in the defect report.

[0026] It is understandable that as the number of structurally vulnerable areas marked in defect reports increases, the multi-factor iterative calculation model will increase the weight coefficient of the fundamental factor—the continuity and rationality of spatial topological chain identification. Conversely, as the number of data ambiguity areas marked in defect reports increases, the multi-factor iterative calculation model will increase the weight coefficient of the fundamental factor—the matching degree between spectral fingerprint coding and the standard spectral library. The multi-factor iterative calculation model multiplies all fundamental factors and all derived factors by their corresponding dynamic weight coefficients and then sums the results to obtain an initial reliability score. The formula for calculating the initial reliability score is: in: This represents the original reliability score. This represents the total number of fundamental factors. Indicates the first The dynamic weighting coefficients of the basic factors Indicates the first The quantified values ​​of the basic factors, This represents the total number of derived factors. Indicates the first The dynamic weighting coefficients of each derived factor Indicates the first The quantified value of each derived factor.

[0027] In practical implementation, the multi-factor iterative calculation model inputs the calculated raw reliability score into a predefined normalization function. The normalization function maps the raw reliability score to a closed numerical range between zero and one, and the resulting value is the current reliability index of the particle entity. Optionally, each time the microscopic infrared data spectrum is updated due to new data entry or changes in connection relationships, the system triggers a recalculation process of the reliability index of the relevant particle entity, thereby achieving dynamic refreshing of the reliability index.

[0028] In one embodiment of the present invention, see [reference] Figure 3 This study performs structural defect scanning and problem area localization on microscopic infrared data maps through a map self-inspection process. The map self-inspection process includes the following steps: First, it initiates a map integrity check, traversing all particle entities in the multi-layer graph structure. It checks for missing data in the spectral fingerprint encoding of each particle entity, unacceptable deviations from the standard spectral library, and the complete closure of spatial topological chain identifiers. Based on the check results, it calculates the integrity score for each particle entity. Second, it initiates an association consistency check, analyzing interconnected particle entities in the multi-layer graph structure. It checks whether the similarity of the spectral fingerprint encodings of the two connected particle entities is below the connection strength threshold, and whether the spatial relationship represented by the connection contradicts the background fluctuation baseline value. Based on the check results, it calculates the contradiction score for each connection between entities. Third, it performs region clustering in the multi-layer graph structure based on the integrity and contradiction scores. From the clustered regions, it extracts regions where all particle entity integrity scores are below the integrity threshold and all connection contradiction scores are above the contradiction threshold, marking these regions as structurally vulnerable areas. From the clustered regions, further extract regions containing at least one particle entity with an extremely low integrity score or an extremely high inconsistency score, and whose average reliability index is below the reliability threshold. Mark these regions as data fuzzy areas. Generate a defect report detailing the spatial coordinate range of all structurally vulnerable areas and the reliability index distribution of all data fuzzy areas.

[0029] In practice, initiating the spectrum self-check process signifies the system's commencement of a systematic inspection of the multi-layered graph structure. This process comprises a series of interconnected verification and labeling steps. The first step is initiating spectrum integrity verification. This verification involves traversing all particle entities within the multi-layered graph structure, checking for missing data fields in the spectral fingerprint code of each entity, verifying unacceptable mismatches between the spectral fingerprint code and the standard spectra of the corresponding substances in the standard spectral library, and confirming the complete closure of the spatial topological chain identifier for each entity. A complete spatial topological chain identifier signifies the ability to describe the continuous spatial relationship of a particle from its starting point to its ending point. Based on the verification results, the system calculates a completeness score for each inspected particle entity; this score is a quantified numerical indicator.

[0030] In practice, initiating association consistency verification is a parallel step in the graph self-inspection process. Association consistency verification focuses on analyzing interconnected particle entity pairs in a multi-layered graph structure. It checks whether the similarity between the spectral fingerprint codes of the particle entities at both ends of the connection is lower than a preset connection strength threshold, and whether the spatial relationship represented by the connection logically contradicts the background fluctuation baseline value of the region where the particle is located. Based on the verification results, the system calculates a contradiction score for each inter-entity connection. This contradiction score quantifies the credibility of the relationship represented by the connection.

[0031] In some embodiments, the system performs extraction and labeling operations from the regions divided by the region clustering algorithm. The extraction operation identifies regions from all clusters where the integrity score of all particle entities is below a preset integrity threshold and the connectivity inconsistency score is above a preset inconsistency threshold; these regions are then uniformly labeled as structurally fragile regions. A further extraction operation identifies regions from all clusters that contain at least one particle entity with an extremely low integrity score or an extremely high inconsistency score, and whose average reliability index is below a preset reliability threshold; these regions are then uniformly labeled as data ambiguity regions. The inconsistency score can be calculated using the following formula: in: This represents the contradiction score of a certain connection. This indicates the total number of feature comparison terms used to evaluate the connection. Indicates the first The similarity normalization value of each feature term. Indicates the first The normalized value of the logical contradiction degree of each feature term. and These are the weighting coefficients for similarity and contradiction, respectively.

[0032] In practice, after scanning and marking all areas, the map self-check process generates a structured defect report. This defect report is a document containing specific coordinates and values, detailing the spatial coordinate ranges of all marked structurally vulnerable areas and the reliability index distribution of all marked data-ambiguous areas. In some embodiments, the defect report is presented as a list, listing the boundary coordinates of each structurally vulnerable area and the reliability index of each particle entity within each data-ambiguous area. The execution of the map self-check process provides a clear target for subsequent simulation and generation of virtual intervention paths, and the output of the defect report is a crucial data interface connecting defect scanning and intervention planning.

[0033] In one embodiment of the invention, a defect report is read, and different simulation environments are initialized for structurally vulnerable areas and data-ambiguous areas, respectively. For structurally vulnerable areas, the goal of the simulation environment is to repair data integrity and simulate the generation of a virtual path sequence for spectral re-measurement. The process involves, in a multi-layer graph structure, starting with particle entities within the structurally vulnerable area, searching outwards for adjacent particle entities with reliability indices higher than a preset standard. The spectral data stream transmission from high-reliability particle entities to target particle entities within the structurally vulnerable area is simulated, and the improvement value of each simulated transmission on the spectral fingerprint encoding integrity score of the target particle entity is calculated. Based on the magnitude of the improvement value and the complexity of the transmission path, multiple virtual data supplementation paths are planned. Each path includes a specific source particle entity, target particle entity, data type being transmitted, and the expected integrity score after simulation execution. The set of these paths is denoted as the spectral re-measurement virtual path sequence. For data-ambiguous areas, the goal of the simulation environment is to clarify data contradictions and simulate the generation of a virtual path sequence for correlation reconstruction. The process involves, in a multi-layer graph structure, suspending all inter-entity connections within the data-ambiguous area that are considered to have excessively high contradiction scores. Outside the data-ambiguous area, particle entity pairs with potentially similar spectral fingerprint encodings but currently without established connections are searched. The simulation establishes connections between these new particle entity pairs and evaluates the impact of these new connections on the average inconsistency score of the entire data ambiguity region. Based on the reduction in inconsistency score, the optimal combination of new connections is selected, and a list of virtual operations containing steps to disconnect old connections and establish new connections is generated. This list is denoted as the association reconstruction virtual path sequence.

[0034] For each spectral re-measurement virtual path sequence, a comprehensive utility value is calculated. This value is determined by the overall integrity score improved by the spectral re-measurement virtual path sequence, the estimated resource consumption required for path sequence execution, and the current reliability index of the target particle entity. For each associated reconstruction virtual path sequence, a comprehensive clarification value is calculated. This value is determined by the overall inconsistency score reduced by the associated reconstruction virtual path sequence, the impact of path sequence execution on the overall connectivity of the spectrum, and the background fluctuation baseline value of the involved particle entities. All spectral re-measurement virtual path sequences are sorted in descending order of their comprehensive utility values ​​to generate a re-measurement sequence queue. All associated reconstruction virtual path sequences are sorted in descending order of their comprehensive clarification values ​​to generate a reconstruction sequence queue. Virtual conflict detection is performed to check whether the top-ranked spectral re-measurement virtual path sequences and associated reconstruction virtual path sequences are competing for the same micro-infrared hardware resources or the same spectral data region. If a virtual conflict is detected, the ranking of the conflicting path sequences is adjusted, or they are broken down into multiple conflict-free subsequences, and the re-measurement sequence queue and reconstruction sequence queue are updated. The final output consists of a conflict-free retest sequence queue and a reconstruction sequence queue, which serve as a set of virtual intervention path sequences to be selected.

[0035] In practical implementation, the system reads the defect report generated by the graph self-inspection process. The defect report details the location and characteristics of structurally vulnerable areas and data ambiguity areas. The system initializes two different simulation environments for the structurally vulnerable areas and data ambiguity areas clearly listed in the defect report. For structurally vulnerable areas, the goal of the simulation environment is to repair data integrity. Its core task is to simulate and generate a virtual path sequence for spectral retesting. The process of simulating and generating the virtual path sequence for spectral retesting involves using one or more particle entities within the structurally vulnerable area as the simulation starting point in a multi-layer graph structure. It searches outward for high-reliability neighboring particle entities with a reliability index higher than the preset standard. It simulates the transmission of spectral data from high-reliability particle entities to target particle entities within the structurally vulnerable area. The system calculates the improvement value of each simulated transmission on the spectral fingerprint encoding integrity score of the target particle entity. Based on the magnitude of the calculated improvement value and the topological complexity of the simulated transmission path, the planning algorithm plans multiple virtual data supplementation paths. Each virtual data supplementation path includes a specific source particle entity identifier, a target particle entity identifier, the data type of transmission, and the expected integrity score after simulation execution. The set of these paths is recorded as the virtual path sequence for spectral retesting.

[0036] In some embodiments, for data ambiguity regions, the goal of the simulation environment is to clarify data inconsistencies. The core task is to simulate and generate a sequence of virtual paths for reconstructing related data. The process of simulating and generating the sequence of virtual paths for reconstructing related data involves pausing all connections between entities marked as having excessively high inconsistency scores within the data ambiguity region in a multi-layer graph structure. Outside the data ambiguity region, the system searches for particle entity pairs with potentially similar spectral fingerprint codes but currently not connected. It simulates the establishment of virtual connections between these newly discovered potentially similar particle entity pairs and evaluates the impact of establishing each new virtual connection on the average inconsistency score of the entire data ambiguity region. Based on the magnitude of the reduction in inconsistency score, the system selects the new connection combinations that maximize the reduction in inconsistency score and generates a list of virtual operations containing steps to disconnect old connections and establish new connections. This list is recorded as the sequence of virtual paths for reconstructing related data.

[0037] In specific implementation, the steps for prioritizing and conflict detection of spectral remeasurement virtual path sequences and associated reconstruction virtual path sequences are as follows: The system performs sorting and detection operations, calculates a comprehensive utility value for each spectral remeasurement virtual path sequence, which is jointly determined by the overall integrity score improvement after execution, the estimated hardware resource consumption required for execution, and the current reliability index of the target particle entity. A comprehensive clarification value is calculated for each associated reconstruction virtual path sequence, which is jointly determined by the overall inconsistency score reduction after execution, the impact of execution on the overall connectivity of the spectrum, and the normalized result of the background fluctuation baseline values ​​of the particle entities involved in the associated reconstruction virtual path sequence. The comprehensive clarification value of the associated reconstruction virtual path sequence... Calculated using the following formula: in: This represents the comprehensive clarification value of the associated reconstructed virtual path sequence. This indicates the expected reduction in the total contradiction score after the associated reconstruction of the virtual path sequence. This represents a quantified value indicating the negative impact on the overall connectivity of the graph. This represents the normalized result of the average background fluctuation baseline value of the particle entities involved in the path. , , represents the weighting coefficients for each item.

[0038] The process involves sorting all virtual path sequences for spectral re-measurement in descending order of their overall utility value, generating an ordered queue of re-measurement sequences, and sorting all associated reconstruction virtual path sequences in descending order of their overall clarification value, generating an ordered queue of reconstruction sequences. Virtual conflict detection is a crucial follow-up step. This check examines whether the top-ranked virtual path sequences for spectral re-measurement and associated reconstruction sequences are competing for the same micro-infrared hardware resources or the same spectral data region. If a virtual conflict is detected, the system adjusts the ranking of the conflicting path sequences in their respective queues, or breaks down the competing long path sequences into multiple temporally and spatially conflict-free subsequences. After the adjustment or decomposition, the system updates the re-measurement sequence queue and the reconstruction sequence queue, ultimately outputting a conflict-free re-measurement sequence queue and a reconstruction sequence queue as the set of virtual intervention path sequences to be selected.

[0039] In one embodiment of the present invention, receiving external intervention instructions and filtering and fusing virtual intervention path sequences based on these instructions is accomplished through the following interaction and decision-making mechanism. The system provides an interactive interface for receiving external intervention instructions, which may include instructions to specify a key detection area, a target particle type, or a detection efficiency mode. When an instruction to specify a key detection area is received, all path sequences whose start point, end point, or path coverage overlaps spatially with the specified key detection area are selected from the set of virtual intervention path sequences. When an instruction to specify a target particle type is received, all path sequences whose spectral fingerprint codes match the specified target particle type features are selected from the set of virtual intervention path sequences. When an instruction to specify a detection efficiency mode is received, the ranking weights of the retest sequence queue and the reconstruction sequence queue are dynamically adjusted according to the efficiency mode requirements, prioritizing path sequences with high overall utility values ​​or short execution times. The selected multiple path sequences are then fused. The fusion process includes identifying common operation steps between different path sequences and merging these common operation steps into a single operation. The operation sequence dependencies between different path sequences are examined, and a linear, acyclic overall sequence of operation steps is generated based on these dependencies. Necessary state checkpoints and data synchronization instructions are inserted into this overall sequence to ensure that the output data of preceding steps can be correctly read by subsequent steps. Finally, a unified, executable fusion intervention path is generated. This fusion intervention path is a data structure containing specific microscopic infrared hardware operation commands, target data addresses, and expected intermediate states.

[0040] The fused intervention path sequence is transformed into physical control commands to drive the microscopic infrared hardware to perform detection operations. This is achieved through a command interpretation and execution engine. The command interpretation and execution engine works as follows: it parses the fused intervention path, mapping each virtual operation step into one or more underlying physical control commands. These physical control commands include commands to move the microscope stage to the target coordinates, commands to focus the infrared spectrometer and acquire spectra within a specified wavenumber range, and commands to adjust the integration time of the image sensor to adapt to background noise. A command execution queue is established, and physical control commands are sent sequentially to the corresponding microscopic infrared hardware controllers. After each physical control command is executed, the hardware's status feedback data and the initially acquired data fragments are read in real time through the sensors. The status feedback data is compared with the predefined expected intermediate states in the fused intervention path. If the comparison result is within the allowable error range, the next physical control command is executed. If the comparison result exceeds the allowable error range, an adaptive adjustment sub-process is triggered. The adaptive adjustment sub-process selects a compensation strategy from a preset strategy library based on the type and magnitude of the deviation between the current state and the expected state. Compensation strategies include resending the previous physical control command, fine-tuning the parameters of the next physical control command, or inserting a new calibration command. After executing the selected compensation strategy, a state comparison is performed again until the conditions for continuing execution are met or the maximum number of retries is reached.

[0041] In practical implementation, the system provides an interactive interface to receive external intervention commands. These commands include commands from operators or higher-level systems specifying key detection areas, target particle types, or detection efficiency modes. When the system receives a command specifying a key detection area, the filtering logic selects all path sequences from the virtual intervention path sequence set whose start point, end point, or path coverage overlaps with the key detection area specified in the command. When the system receives a command specifying a target particle type, the filtering logic selects all path sequences from the virtual intervention path sequence set whose spectral fingerprint codes match the target particle type characteristics specified in the command. When the system receives a command specifying a detection efficiency mode, the system dynamically adjusts the sorting weights of the retest sequence queue and the reconstruction sequence queue according to the specific requirements of the efficiency mode, prioritizing path sequences with high comprehensive utility values ​​or short estimated execution times. Refer to Table 1 for the correspondence between external command types and filtering rules.

[0042] Table 1: Correspondence between External Intervention Instruction Types and Screening Rules

[0043] In practical implementation, multiple path sequences selected based on external intervention commands are fused. The fusion process involves a series of structured operations. First, the fusion process identifies common operational steps among different path sequences and merges these common steps into a single operation. For example, if multiple path sequences require moving the microscope stage to the same coordinate region, this movement command will only appear once in the fused intervention path. The fusion process checks the operational sequence dependencies between different path sequences and generates a linear, acyclic total sequence of operational steps based on these dependencies. In the generated total sequence of operational steps, the system inserts necessary state checkpoints and data synchronization commands. State checkpoints confirm that preceding hardware operations have been completed, and data synchronization commands ensure that the output data of preceding steps can be correctly read by subsequent steps. Finally, a unified, executable fused intervention path is generated. The fused intervention path is a data structure containing specific microscopic infrared hardware operation commands, target data addresses, and expected intermediate states. The generation efficiency of the fused intervention path can be evaluated through the following relationship: in: This represents the fusion efficiency coefficient. This represents the number of unique operations in the total sequence of operation steps after fusion. Indicates the first The estimated time for a single operation. This represents the total number of operation steps in all the original path sequences selected before fusion. Indicates the first The estimated time for each original operation step, This represents the coordination and synchronization overhead introduced during the fusion process.

[0044] In practice, the fused intervention path sequence is transformed into physical control commands to drive the microscopic infrared hardware to perform detection operations. This is achieved through an command interpretation and execution engine. The command interpretation and execution engine works as follows: it parses the fused intervention path, mapping each virtual operation step into one or more low-level physical control commands. These physical control commands include, but are not limited to, commands to move the microscope stage to the target coordinates, commands to focus the infrared spectrometer and acquire spectra within a specified wavenumber range, and commands to adjust the integration time of the image sensor to adapt to background noise. The command interpretation and execution engine establishes a command execution queue and sends physical control commands sequentially to the corresponding microscopic infrared hardware controller. After each physical control command is executed, the command interpretation and execution engine reads the hardware's status feedback data and the initially acquired data fragments in real time through the sensors. The command interpretation and execution engine compares the read status feedback data with the predefined expected intermediate states in the fused intervention path.

[0045] In some embodiments, if the state comparison result is within the system's allowed error range, the instruction interpretation and execution engine continues to execute the next physical control instruction in the instruction execution queue. If the state comparison result exceeds the system's allowed error range, the instruction interpretation and execution engine triggers an adaptive adjustment sub-process. The adaptive adjustment sub-process selects a compensation strategy from a preset strategy library based on the type and magnitude of the deviation between the current state and the expected state. Compensation strategies include resending the previous physical control instruction, fine-tuning the parameters of the next physical control instruction, or inserting a new calibration instruction. After executing the selected compensation strategy, the instruction interpretation and execution engine re-performs the state comparison until the conditions for continued execution are met or the maximum number of retries set by the system is reached.

[0046] See Figure 4 In the state control of the execution of microscopic infrared hardware commands, the dynamic changes of the expected state and the actual state, as well as the effect of the compensation strategy, are presented simultaneously. Specifically, the figure uses "execution steps" as the horizontal axis, the left vertical axis corresponds to the state values ​​(including the expected state and the actual state), and the right vertical axis corresponds to the error reduction value after compensation. Red dots mark the error exceeding the limit, and orange bars represent the error reduction magnitude after the implementation of the compensation strategy. During actual execution, the expected state remains stable with 100 as the baseline, while the actual state fluctuates with the execution steps: when the actual state deviates from the expected state (such as at steps 5.0, 15.0, etc.) and triggers the error exceeding the limit, the system activates the compensation strategy. This can be observed through the error reduction value corresponding to the orange bars. After compensation, the actual state converges to the expected state (e.g., at step 5.0, the actual state rises from 90, and the error reduction value after compensation reaches more than 2.5). In addition, the large fluctuation of the actual state after step 15.0 is accompanied by a high-amplitude error reduction bar, reflecting the rapid correction capability of the compensation strategy for deviations; while the horizontal dashed line of the expected state serves as the baseline, clearly showing the regression effect of the actual state after compensation.

[0047] In one embodiment of the present invention, after all physical control commands have been executed, all final result data returned by the hardware is collected. This final result data includes newly acquired infrared spectra, updated particle space images, and environmental parameter readings. The final result data is preprocessed, including spectral denoising, image registration, and data format standardization, generating standardized new data blocks. In the multi-layer graph structure, the target particle entities or target connections corresponding to the new data blocks are located. The original state parameters of the target particle entities are replaced or supplemented using the standardized new data blocks. The new infrared spectral data, after feature locking, generates new spectral fingerprint codes to update the spectral fingerprint codes of the particle entities. The new particle space images, after trajectory analysis, generate new spatial topology chains to update the spatial topology chain identifiers of the particle entities. New environmental parameter readings are used to update the background fluctuation baseline values. The reliability index of all affected particle entities and the inconsistency score of all affected connections are recalculated. Based on the updated reliability index and inconsistency score, the local or global connections of the multi-layer graph structure are optimized. Optimization includes deleting isolated particle entities with reliability indices below the deletion threshold and no important connections. New connections are established between particle entities whose spectral fingerprints are highly similar and whose spatial topological chain identifiers indicate proximity. Low-quality connections with inconsistency scores consistently above the disconnection threshold are weakened or broken. After optimization, the microscopic infrared data spectrum enters a stable, updated state, awaiting the next detection cycle.

[0048] In practice, after all physical control commands have been executed, the system collects all final result data returned by the microscopic infrared hardware. This final result data includes newly acquired particle infrared spectra, updated particle space images, and relevant environmental parameter readings. Preprocessing this final result data is a necessary step. Preprocessing includes denoising the newly acquired infrared spectra, registering the updated particle space images, and standardizing the format of all data. After preprocessing, standardized new data blocks are generated.

[0049] In practical implementation, the target particle entity or target connection relationship corresponding to the standardized new data block is located in the multi-layer graph structure. The location operation is based on the spatial coordinate identifier or unique particle entity code carried in the standardized new data block. The original state parameters of the target particle entity are replaced or supplemented with the standardized new data block. The specific update operation involves multiple aspects. After the new infrared spectral data undergoes feature locking processing, a new spectral fingerprint code is generated, which is used to update the original spectral fingerprint code of the target particle entity. After the new particle space image undergoes trajectory analysis processing, a new spatial topology chain is generated, which is used to update the original spatial topology chain identifier of the target particle entity. The new environmental parameter readings are directly used to update the background fluctuation baseline value of the area where the target particle entity is located.

[0050] In some embodiments, after updating the state parameters, the system needs to recalculate the reliability index of all affected particle entities and the inconsistency score of all affected connections. Based on the updated reliability index and inconsistency score, the system performs optimization operations on the local or global connections of the multi-layer graph structure. Optimization operations include deleting isolated particle entities with a reliability index below a preset deletion threshold and no significant connections. Optimization operations include establishing new connections between particle entities with highly similar spectral fingerprint encodings and spatial topological chain identifiers indicating spatial proximity. Optimization operations include weakening or disconnecting low-quality connections with inconsistency scores consistently above a preset disconnection threshold. The updated connection quality can be internally evaluated based on the following formula: in: This represents the connectivity quality assessment value of the local map. This represents the average reliability index involving particles. This represents the maximum degree of contradiction score for connections within the region. This represents the sparsity coefficient of the connected network. , , These are the adjustment factors for each item.

[0051] Understandably, after completing the above optimization operations, the microscopic infrared data spectrum enters a stable, updated state, awaiting the start of the next detection cycle. In some embodiments, optionally, the system generates a spectrum update log, recording the specific particle entities involved in this reconstruction, connection changes, and snapshots of updated key parameters.

[0052] See Figure 5In the feature analysis stage after spectral data preprocessing, the distribution relationship between spectral fingerprint coding matching degree and the number of spectral fingerprint peaks was presented in the form of a heatmap, with color depth corresponding to the number of particles (darker colors represent a larger number of particles). Specifically, when the number of peaks was 4, a local high value region of particle number appeared in the matching degree range of 0.65-0.75 (corresponding to the dark blue block in the figure); while when the number of peaks was 4, 6, and 7, a certain number of particles were distributed in the matching degree range of 0.75-0.95. In terms of analysis dimensions, the number of peaks, as one of the core feature parameters of spectral fingerprints, and its correlation distribution with the matching degree can help screen spectral fingerprints with high discriminative power: for example, samples with a peak number of 4 and a matching degree of 0.65-0.75 can be used as potential feature identifiers of "structurally vulnerable areas" in subsequent spectral self-checks; while samples with a peak number of 6-7 and a matching degree of 0.80 or higher can be used as spectral feature references for high-reliability particle entities. During the parameter mapping process, the matching degree interval is divided according to the 0.05 gradient, the number of peaks is discretely distributed with integer scales, and the thermal mapping range of the number of particles is 0-3.0.

[0053] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0054] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An automated detection method for microplastic particles based on microscopic infrared imaging, characterized in that, The method includes: Create and maintain a dynamically updated microscopic infrared data spectrum; Structural defect scanning and problem area localization are performed on the aforementioned microscopic infrared data spectrum; Based on the defect scanning results, multiple virtual intervention path sequences are simulated and generated within the microscopic infrared data spectrum; Receive external intervention instructions, and filter and fuse the virtual intervention path sequence according to the external intervention instructions; The fused intervention path sequence is converted into physical control commands to drive the microscopic infrared hardware to perform detection operations, and the internal connections of the microscopic infrared data spectrum are reconstructed based on the execution results.

2. The automated detection method for microplastic particles based on microscopic infrared imaging according to claim 1, characterized in that, The creation and maintenance of a dynamically updated microscopic infrared data spectrum includes the following specific processes: Import multiple batches of microscopic infrared imaging records from a historical sample database. The microscopic infrared imaging records include particle spectral sequences, spatial coordinate trajectories, and background noise patterns. The multiple batches of microscopic infrared imaging records are subjected to layered deconstruction processing to generate the bottom-level data units of the spectrum; The hierarchical deconstruction process includes: Feature locking is performed on the particle spectral sequence to extract discriminative peak frequency combinations and absorption intensity profiles, thus forming a spectral fingerprint; Perform trajectory analysis on the spatial coordinate trajectory, encode the physical displacement path of the particle into a spatial topological chain, and record the distance vector between the path node and the adjacent particle; Perform pattern stripping on the background noise pattern, calculate and store the background fluctuation baseline and noise spectrum characteristics of each imaging region; Based on the spectral fingerprint, the spatial topological chain, and the background fluctuation baseline, a multi-layer graph structure is instantiated in memory as the microscopic infrared data spectrum; In the multi-layer graph structure, each data unit is defined as a particle entity, and each particle entity is assigned a set of state parameters, including the particle entity's spectral fingerprint encoding, spatial topological chain identifier, background fluctuation baseline value, and a reliability index generated by an internal evaluation mechanism. The internal evaluation mechanism operates in real time, dynamically refreshing the reliability index of each particle entity by periodically comparing newly input microscopic infrared imaging data with the spectral fingerprint code and spatial topological chain identifier stored in the multi-layer graph structure.

3. The automated detection method for microplastic particles based on microscopic infrared imaging according to claim 2, characterized in that, The structural defect scanning and problem area localization of the microscopic infrared data spectrum are specifically achieved through a spectrum self-checking process, which includes the following steps: Initiate a graph integrity check, traverse all particle entities in the multi-layer graph structure, check whether there is data loss in the spectral fingerprint code of each particle entity, whether the spectral fingerprint code has an unacceptable deviation from the standard spectral library, and check whether the spatial topology chain identifier is completely closed. Based on the verification results, calculate the integrity score for each particle entity; Initiate association consistency verification, analyze the interconnected particle entities in the multi-layer graph structure, check whether the spectral fingerprint encoding similarity of the particle entities at both ends of the connection is lower than the connection strength threshold, and check whether the spatial relationship represented by the connection contradicts the background fluctuation baseline value. Based on the verification results, calculate the contradiction score for each connection between entities; Based on the completeness score and the contradiction score, region clustering is performed in the multi-layer graph structure; From the clustered regions, extract the regions where the integrity score of all particle entities is lower than the integrity threshold and the contradiction score of all connections is higher than the contradiction threshold, and mark these regions as structurally fragile regions. From the clustered regions, further extract regions containing at least one particle entity with an extremely low integrity score or an extremely high contradiction score, and whose average reliability index is lower than the reliability threshold. Mark these regions as data fuzzy areas. Generate a defect report that details the spatial coordinate range of all structurally vulnerable areas and the reliability index distribution of all data ambiguity areas.

4. The automated detection method for microplastic particles based on microscopic infrared imaging according to claim 3, characterized in that, Based on the defect scanning results, multiple virtual intervention path sequences are simulated and generated within the microscopic infrared data spectrum. This process specifically includes the following simulation steps: Read the defect report and initialize different simulation environments for the structurally vulnerable area and the data ambiguity area, respectively; For the structurally vulnerable region, the goal of the simulation environment is to repair data integrity by simulating and generating a virtual path sequence for spectral re-measurement; the process is as follows: In the multi-layer graph structure, starting from the particle entities in the structurally fragile area, the search extends outward to find adjacent particle entities with a reliability index higher than the preset standard. Simulate the transmission of spectral data stream from a high-reliability particle entity to a target particle entity within a structurally vulnerable region, and calculate the improvement in the integrity score of the target particle entity's spectral fingerprint encoding for each simulated transmission; Based on the magnitude of the boost value and the complexity of the transmission path, multiple virtual data supplementation paths are planned. Each path includes a specific source particle entity, a target particle entity, the data type of the transmission, and the expected completeness score after the simulation is executed. The set of these paths is denoted as the spectral retest virtual path sequence. For the aforementioned data ambiguity area, the goal of the simulation environment is to clarify data contradictions and simulate the generation of associated reconstructed virtual path sequences; the process is as follows: In the multi-layer graph structure, all inter-entity connections within the data ambiguity region that are considered to have excessively high contradiction scores are suspended; Outside the data ambiguity region, search for particle entity pairs with potentially similar spectral fingerprint codes but currently not connected; Simulate the establishment of connections between these new particle entity pairs and evaluate the impact of the new connections on the average inconsistency score of the entire data fuzzy region. Based on the degree of decrease in the contradiction score, the optimal combination of new connections is selected, and a list of virtual operations containing the steps of disconnecting old connections and establishing new connections is generated. This list is recorded as the virtual path sequence of association reconstruction.

5. The automated detection method for microplastic particles based on microscopic infrared imaging according to claim 4, characterized in that, The method further includes the steps of prioritizing and detecting conflicts in the spectral re-measurement virtual path sequence and the associated reconstructed virtual path sequence, specifically: A comprehensive utility value is calculated for each of the aforementioned spectral retest virtual path sequences. The comprehensive utility value is jointly determined by the total integrity score that the spectral retest virtual path sequence can improve, the estimated resource consumption required to execute the path sequence, and the current reliability index of the target particle entity. A comprehensive clarification value is calculated for each of the aforementioned associated reconstructed virtual path sequences. The comprehensive clarification value is jointly determined by the total contradiction score that the associated reconstructed virtual path sequence can reduce, the impact of the path sequence execution on the overall connectivity of the graph, and the background fluctuation baseline value involving particle entities. Sort all the virtual path sequences for spectral remeasurement in descending order of their overall utility value to generate a remeasurement sequence queue; Sort all the associated reconstructed virtual path sequences in descending order of their comprehensive clarification values ​​to generate a reconstructed sequence queue; Perform virtual conflict detection: Check whether the top-ranked spectral remeasurement virtual path sequences and associated reconstructed virtual path sequences are competing for the same micro-infrared hardware resources or the same spectral data region; If a virtual conflict is detected, the sorting position of the conflict path sequence is adjusted, or it is broken down into multiple conflict-free subsequences, and the retest sequence queue and the reconstruction sequence queue are updated. The final output consists of a conflict-free retest sequence queue and a reconstruction sequence queue, which serve as a set of virtual intervention path sequences to be selected.

6. The automated detection method for microplastic particles based on microscopic infrared imaging according to claim 5, characterized in that, The process of receiving external intervention instructions and filtering and fusing the virtual intervention path sequence based on those instructions is accomplished through the following interaction and decision-making mechanisms: The system provides an interactive interface for receiving external intervention commands, which include commands to specify a key detection area, a target particle type, or a detection efficiency mode. When a command to designate a key detection area is received, all path sequences whose starting point, ending point, or path coverage overlaps with the designated key detection area are selected from the set of virtual intervention path sequences. When a command for a specified target particle type is received, all path sequences whose spectral fingerprint codes match the characteristics of the specified target particle type are selected from the set of virtual intervention path sequences. When a specified detection efficiency mode instruction is received, the sorting weights of the retest sequence queue and the reconstruction sequence queue are dynamically adjusted according to the requirements of the efficiency mode, and path sequences with high comprehensive utility values ​​or short execution times are selected first. The selected multiple path sequences are merged. The fusion process includes: Identify common operation steps between different path sequences and merge these common operation steps into a single operation. Examine the operational order dependencies between different path sequences, and generate a linear, acyclic total sequence of operational steps based on the dependencies; In the overall sequence of operation steps, necessary status checkpoints and data synchronization instructions are inserted to ensure that the output data of the preceding steps can be correctly read by the subsequent steps. Ultimately, a unified and executable fusion intervention path is generated. The fusion intervention path is a data structure that includes specific microscopic infrared hardware operation commands, target data addresses, and expected intermediate states.

7. The automated detection method for microplastic particles based on microscopic infrared imaging according to claim 6, characterized in that, The process of converting the fused intervention path sequence into physical control commands to drive the microscopic infrared hardware to perform detection operations is achieved through an instruction interpretation and execution engine, which operates according to the following steps: The fusion intervention path is analyzed, and each virtual operation step is mapped to one or more underlying physical control commands; The physical control commands include, but are not limited to: commands to control the microscope stage to move to the target coordinates, commands to control the infrared spectrometer to focus and acquire spectra within a specified wavenumber range, and commands to control the image sensor to adjust the integration time to adapt to background noise. Establish an instruction execution queue and send the physical control instructions to the corresponding microscopic infrared hardware controller in sequence; After each physical control command is executed, the hardware status feedback data and the initially collected data fragments are read in real time through sensors; The state feedback data is compared with the predefined expected intermediate states in the fusion intervention path; If the comparison result is within the allowable error range, then continue to execute the next physical control command; If the comparison result exceeds the allowable error range, an adaptive adjustment subprocess is triggered: The adaptive adjustment subprocess selects a compensation strategy from a preset strategy library based on the type and magnitude of the deviation between the current state and the expected state. The compensation strategy may include resending the previous physical control command, fine-tuning the parameters of the next physical control command, or inserting a new calibration command. After executing the selected compensation strategy, the state comparison is performed again until the conditions for continuing execution are met or the maximum number of retries is reached.

8. The automated detection method for microplastic particles based on microscopic infrared imaging according to claim 7, characterized in that, The process of reconstructing the internal connections of the microscopic infrared data spectrum based on the execution results specifically includes the following spectrum update steps: After all the physical control commands have been executed, all the final result data returned by the hardware is collected. The final result data includes newly acquired infrared spectra, updated particle space images, and environmental parameter readings. The final result data is preprocessed, including spectral denoising, image registration, and data format standardization, to generate standardized new data blocks. In the multi-layer graph structure, locate the target particle entity or target connection relationship corresponding to the new data block; The original state parameters of the target particle entity are replaced or supplemented with the standardized new data blocks. Specifically, the new infrared spectral data is used to generate a new spectral fingerprint code after feature locking, which is then used to update the spectral fingerprint code of the particle entity. After trajectory analysis, the new particle space image generates a new spatial topology chain, which is used to update the spatial topology chain identifier of the particle entity. New environmental parameter readings are used to update the background fluctuation baseline values; Recalculate the reliability index of all affected particle entities, as well as the contradiction score of all affected connections; Based on the updated reliability index and contradiction score, the local or global connectivity of the multi-layer graph structure is optimized; The optimization includes: deleting isolated particle entities with a reliability index below the deletion threshold and no important connections; establishing new connections between particle entities with highly similar spectral fingerprint codes and spatial topological chain identifiers indicating proximity; weakening or breaking low-quality connections with inconsistency scores consistently above the break threshold; After optimization, the microscopic infrared data spectrum enters a stable, updated state, awaiting the next detection cycle.

9. The automated detection method for microplastic particles based on microscopic infrared imaging according to claim 2, characterized in that, The reliability index generated by the internal evaluation mechanism is dynamically generated through a multi-factor iterative calculation model, which performs the following steps: The basic factors of the particle entity are obtained, including the matching degree between the spectral fingerprint encoding and the standard spectral library, the continuity and rationality of the spatial topological chain identification, and the stability of the background fluctuation baseline value. Obtain the derivation factors of the particle entity, including the number of connections of the particle entity in the multi-layer graph structure, the average reliability index of the adjacent particle entities connected to it, and the frequency of the particle entity's historical state parameters being successfully updated. A dynamic weighting coefficient is assigned to each basic factor and derived factor, and the dynamic weighting coefficient is periodically adjusted according to the problem area type marked in the defect report; When the number of structurally vulnerable areas increases in the defect reports, the weighting coefficient of the continuity and rationality factors of the spatial topology chain identifier is increased. When the number of ambiguous areas in the defect report increases, the weighting coefficient of the matching factor between the spectral fingerprint code and the standard spectral library is increased. A raw reliability score is obtained by multiplying all basic and derived factors by their corresponding dynamic weight coefficients and summing the results. The original reliability score is input into a standardization function and mapped to a numerical range between zero and one. The mapped value is the current reliability index of the particle entity. Each time the microscopic infrared data spectrum is updated, the reliability index of the relevant particle entity is recalculated, thereby realizing the dynamic refresh of the reliability index.

10. An automated detection system for microplastic particles based on microscopic infrared imaging, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the automated detection method for microplastic particles based on microscopic infrared imaging as described in any one of claims 1 to 9.

Images