Aerosol optical tweezer formant trajectory identification method and system based on iterative clustering

Abstract

The invention relates to an aerosol optical tweezer formant trajectory identification method and system based on iterative clustering, belongs to the technical field of spectrum formant identification, and solves the problem that in the prior art, when aerosol spectrum data with a very long time sequence or extremely high data point density is analyzed, the spectrum data is not accurate. And the problems of difficult trajectory tracking and large computing resource consumption caused by noise and baseline drift are solved. The method comprises the following steps: acquiring multi-frame aerosol spectrum data, and constructing a time-peak position data set; according to the time-peak data set, based on a segmented anchoring rule, determining at least one anchoring trajectory and a corresponding time-peak data subset by adopting a dynamic trajectory tracking algorithm; based on the anchoring trajectory, calculating a relative wavelength difference value, carrying out clustering analysis, and determining a corresponding segmented trajectory; a formant trajectory is determined based on the anchoring trajectory and the segmentation trajectory. The method achieves the automatic and high-precision recognition of a plurality of formant tracks of long-sequence spectral data based on the stable physical constraint of the relative distance between formants.

Description

Technical Field

[0001] This invention relates to a method and system for identifying aerosol optical tweezers resonance peak trajectories based on iterative clustering, belonging to the field of spectral resonance peak identification technology. Background Technology

[0002] The physicochemical properties of aerosol particles, such as particle size, refractive index, and their dynamic changes, are of paramount importance in research fields such as materials science and biomedicine. Combining spectral analysis with Mie scattering theory for data analysis, fitting, and inversion can yield particle dynamic information with extremely high precision. However, in actual experimental measurements, the acquired spectral data is often not ideal but rather complex, noisy, and data-dense. Identifying resonance peaks and tracing the true resonance peak trajectory from temporal and resonance peak information containing numerous stray points presents significant challenges.

[0003] Existing peak finding algorithms are prone to misjudgment, inaccurate positioning, or loss of accuracy, directly affecting the accuracy of subsequent inversion results. Furthermore, even if peak position information is obtained, traditional tracking methods such as simple linear connections or Kalman filtering are difficult to effectively distinguish between real signals and noise, easily resulting in incorrect connections, trajectory interruptions, or confusion of different trajectories. They also rely on human experience, which is extremely inefficient and lacks objectivity and repeatability due to experience and subjective judgment.

[0004] Existing technologies face challenges in analyzing aerosol spectral data with exceptionally long time series or extremely high data point density. These challenges include difficulty in tracing the trajectory due to noise and baseline drift, as well as high computational resource consumption. There is an urgent need for an automated, high-precision, and highly complete method to accurately locate peak positions and extract multiple resonance peak trajectories. Summary of the Invention

[0005] In view of the above analysis, the present invention aims to provide a method and system for identifying aerosol optical tweezers resonance peak trajectories based on iterative clustering, in order to solve the problem that in existing aerosol optical tweezers spectral analysis, due to noise, baseline drift and signal discontinuity, particularly long time series or extremely high data point density, it is difficult to accurately, completely and efficiently identify and track multiple resonance peak trajectories, which in turn affects the accuracy of subsequent physical parameter inversion.

[0006] On one hand, embodiments of the present invention provide a method for identifying aerosol optical tweezers resonance peak trajectories based on iterative clustering, including: Acquire multiple frames of aerosol spectral data to construct a time-peak dataset; wherein, the time-peak dataset includes time frames and multiple candidate resonance peaks corresponding to each time frame; Based on the time-peak dataset, and using a dynamic trajectory tracking algorithm based on segmented anchoring rules, at least one anchoring trajectory and its corresponding time-peak data subset are determined. Based on the anchoring trajectory, the relative wavelength difference of each candidate resonance peak in the corresponding time-peak data subset is calculated; Cluster analysis is performed on all relative wavelength differences corresponding to the time-peak data subset to determine the corresponding segmented trajectories; The resonance peak trajectory is determined based on the anchored trajectory and the segmented trajectory corresponding to all the time-peak data subsets.

[0007] Furthermore, based on the aforementioned time-peak dataset, and using a dynamic trajectory tracking algorithm based on segmented anchoring rules, at least one anchoring trajectory and its corresponding time-peak data subset are determined, including: Based on the time-peak dataset, segmented anchoring rules are executed to determine the current time frame and the initial formant. Based on the initial resonance peak, starting from the next time frame of the current time frame, the dynamic trajectory tracking algorithm is used to search for and locate the anchor resonance peak frame by frame in the time-peak dataset in chronological order. Based on the initial resonance peak and all the anchoring resonance peaks, the anchoring trajectory is generated, and the current time period covered by the anchoring trajectory is determined. Based on the current time period, determine the corresponding time-peak data subset.

[0008] Furthermore, the segmented anchoring rule includes a first segmentation rule and a first anchoring rule; Based on the time-peak dataset, segmented anchoring rules are executed to determine the current time frame and the initial resonance peak, including: Determine if this is the first time the segmented anchoring rule has been executed; If this is the first time the segmented anchoring rule is executed, the current time frame is determined from the time-peak dataset based on the first segmentation rule. From all the candidate resonance peaks corresponding to the current time frame, select the initial resonance peak according to the first anchoring rule.

[0009] Furthermore, the segmented anchoring rule also includes a second segmentation rule and a second anchoring rule; Based on the time-peak dataset, segmented anchoring rules are executed to determine the current time frame and the initial formant, including: If this is not the first time the segmented anchoring rule is executed, then the current time period determined after the previous execution is obtained, and the current time frame is determined from the current time period based on the current time period and the second segmentation rule; From all the candidate resonance peaks corresponding to the current time frame, an initial resonance peak is selected according to the second anchoring rule.

[0010] Further, from all the candidate formants corresponding to the current time frame, an initial formant is selected according to the second anchoring rule, including: Obtain the anchored resonance peak corresponding to the current time frame, and compare it with the difference of each candidate resonance peak corresponding to the current time frame; Based on the aforementioned differences, an initial resonance peak is selected according to the second anchoring rule.

[0011] Further, based on the anchoring trajectory, the relative wavelength difference of each candidate resonance peak in the corresponding time-peak data subset is calculated, including: The initial resonance peak and the wavelength values ​​of each anchoring resonance peak in the anchoring trajectory are extracted as anchoring wavelength values; Select one frame from the corresponding time-peak data subset as the target time frame, extract the wavelength value of each candidate resonance peak corresponding to the target time frame, and calculate the difference between the wavelength value and the anchor wavelength value corresponding to the target time frame to obtain the relative wavelength difference of each candidate resonance peak corresponding to the target time frame. By traversing each frame in the time-peak data subset, the relative wavelength difference of each candidate resonance peak in the time-peak data subset is obtained.

[0012] Furthermore, cluster analysis is performed on all relative wavelength differences corresponding to the time-peak data subset to determine the corresponding segmented trajectories, including: The relative wavelength differences are clustered to obtain at least one segmented cluster; wherein, the segmented cluster includes multiple relative wavelength differences; Obtain the candidate resonance peak corresponding to each relative wavelength difference in the segmented cluster, and use it as a segmented resonance peak cluster; The segmented trajectory is determined based on the segmented resonance peak cluster.

[0013] Further, based on the segmented resonance peak cluster, determining the segmented trajectory includes: Detect the temporal continuity between the candidate resonance peaks in the segmented resonance peak cluster; For two candidate resonance peaks that do not conform to time continuity, the first interpolation method is used to determine the complete resonance peak; The completed resonance peak is added to the segmented resonance peak cluster to form a calibration segmented resonance peak cluster; The segmented trajectory is determined based on the calibrated segmented resonance peak cluster.

[0014] Further, determining the resonance peak trajectory based on the segmented trajectories corresponding to all the time-peak data subsets includes: For each anchored trajectory and each segmented trajectory, determine its continuity with each existing trajectory in the trajectory set; If the anchored trajectory or the segmented trajectory meets the trajectory continuity requirement with the existing trajectory, then the existing trajectory is updated based on the anchored trajectory or the segmented trajectory; If the anchored trajectory or the segmented trajectory does not conform to trajectory continuity with all the existing trajectories, then the anchored trajectory or the segmented trajectory is regarded as a new existing trajectory, and the trajectory set is updated. Traverse all the anchored trajectories and segmented trajectories, and determine the existing trajectories in the updated trajectory set as the formant trajectories.

[0015] On the other hand, embodiments of the present invention provide an aerosol optical tweezers resonance peak trajectory recognition system based on iterative clustering, comprising: The dataset generation module acquires multiple frames of aerosol spectral data and constructs a time-peak dataset; wherein, the time-peak dataset includes time frames and multiple candidate resonance peaks corresponding to each time frame; The anchor trajectory determination module, based on the time-peak dataset and a segmented anchoring rule, uses a dynamic trajectory tracking algorithm to determine at least one anchor trajectory and its corresponding time-peak data subset. The wavelength difference calculation module is used to calculate the relative wavelength difference of each candidate resonance peak in the corresponding time-peak data subset based on the anchoring trajectory. The segmented trajectory determination module is used to perform cluster analysis on all relative wavelength differences corresponding to the time-peak data subset to determine the corresponding segmented trajectory. The trajectory determination module determines the resonance peak trajectory based on the anchored trajectory and the segmented trajectory corresponding to all the time-peak data subsets.

[0016] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects: 1. By using segmented anchoring rules, segmented processing of spectral data with particularly long time series or extremely high data point density is achieved, i.e., time-peak data subsets. For each subset, the relative wavelength difference is calculated based on the anchoring trajectory. Further cluster analysis is performed to determine the segmented trajectory, and then the resonance peak trajectory is determined. Using the physical constraint of stable relative distance between resonance peaks, clustering denoising and trajectory reconstruction are performed to achieve automated and high-precision identification of multiple resonance peak trajectories in complex noisy spectral data. The original accuracy of peak position information is preserved, improving data processing efficiency and the accuracy, objectivity and repeatability of subsequent aerosol physical parameter inversion.

[0017] 2. By using segmented anchoring rules and dynamic trajectory tracking algorithms, the anchoring trajectory and the corresponding time-peak data subset are determined, ensuring that the anchoring trajectory within each subset is effective, stable, and highly continuous. This effectively addresses the complexities in actual spectral data and serves as a benchmark for subsequent relative wavelength differences, thereby achieving accurate segmented clustering analysis and segmented trajectory identification and separation.

[0018] 3. By clustering relative wavelength differences to obtain clusters, similar and dense points can be identified and aggregated, thereby determining multiple segmented trajectories. This can effectively address complex situations commonly found in actual spectral data, such as transient signal loss, a large number of stray points, and multiple physical trajectories occurring concurrently, close to, or even intersecting. It has achieved accurate and automated identification of multiple resonance peak trajectories from noisy and complex spectral data.

[0019] 4. For segmented trajectories, the integrity of the segmented trajectories is ensured by judging time continuity and interpolation completion. The trajectory is iteratively updated by judging trajectory continuity, which enables the extension and tracking of segments. While effectively controlling computing resources, the accuracy and efficiency of data processing are guaranteed, the quality of the final result is ensured, and the needs of large-scale data analysis are met.

[0020] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0021] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Figure 1 This is a flowchart illustrating a method for identifying aerosol optical tweezers resonance peak trajectories based on iterative clustering in an embodiment of the present invention. Figure 2 This is a schematic diagram of the main modules of an aerosol optical tweezers resonance peak trajectory recognition system based on iterative clustering in an embodiment of the present invention; Figure 3 This is an example of a single-frame aerosol optical tweezers WGM spectral data in an embodiment of the present invention; Figure 4 This is an example scatter plot showing the evolution of candidate resonance peak positions over time in an embodiment of the present invention; Figure 5 This is an example diagram of multiple resonance peak trajectories extracted by the method of this embodiment of the invention. Detailed Implementation

[0022] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0023] A specific embodiment of the present invention discloses a method for identifying aerosol optical tweezers resonance peak trajectories based on iterative clustering, such as... Figure 1 As shown, it includes: Step S1: Acquire multiple frames of aerosol spectral data and construct a time-peak dataset; wherein the time-peak dataset includes time frames and multiple candidate resonance peaks corresponding to each time frame.

[0024] Multi-frame aerosol spectral data can come from continuous spectral detection. This data contains the whispering-gallery mode formant signals to be analyzed, typically exhibiting a low signal-to-noise ratio and superimposed on a changing background baseline. Therefore, baseline correction and candidate formant extraction are required for the multi-frame aerosol spectral data to construct a time-peak dataset, specifically including: Step S11: Acquire multiple frames of aerosol spectral data.

[0025] The physicochemical properties of aerosol particles are of paramount importance in their applied research. To achieve precise, non-contact, and in-situ dynamic characterization of individual aerosol particles, combining optical trapping techniques (especially aerosol optical tweezers) with microcavity whispering gallery mode (WGM) spectroscopy (i.e., aerosol spectral data) analysis has become a cutting-edge and powerful research method.

[0026] Due to the instability of the light source, detector noise, ambient stray light, and the complexity of the particle's own scattering characteristics, raw spectral data typically has a low signal-to-noise ratio and is superimposed on a changing background baseline. Furthermore, resonance peaks themselves may shift position depending on the particle state (e.g., hygroscopic growth, chemical reactions, phase transitions, etc.), and multiple sets of resonance peaks of different modes (e.g., TE, TM modes) or different orders may appear and evolve simultaneously. These peak clusters coexist and evolve according to their own rules, further increasing the complexity of data interpretation. These various factors result in complex and noisy raw spectral data, necessitating processing of the time-series spectral data to extract candidate resonance peaks.

[0027] Time-series spectral data can originate from continuous spectral detection of individual droplets captured by an aerosol optical tweezers system. The data includes the whispering-gallery mode resonance signals to be analyzed. Each frame records the distribution of light signal intensity scattered by the captured single particle as a function of wavelength (or the pixel coordinates of the spectrometer detector) at a specific time point. Therefore, each frame of aerosol spectral data includes wavelength and intensity information for that frame's spectrum (e.g.,...). Figure 3 (This is an example image of a single frame of WGM aerosol optical tweezers spectral data).

[0028] It should be noted that if the raw data output by the aerosol optical tweezers system or other spectral detection equipment adopts a specially defined binary or special text format, it needs to be converted into a standard two-dimensional text format in advance using a data conversion tool. For example, each line represents a wavelength channel and its corresponding light intensity reading, which is easy for subsequent programs to parse. Data format conversion can be performed using existing technologies, which will not be elaborated in this embodiment.

[0029] Step S12: Perform baseline correction on the aerosol spectral data to obtain corrected spectral data.

[0030] Specifically, this embodiment employs the adaptive iterative reweighted penalized least squares method (airPLS algorithm) to perform baseline correction on each frame of aerosol spectral data in order to accurately deduct the background baseline component superimposed on the resonance peak signal and obtain the corresponding corrected spectral data.

[0031] In this embodiment, the airPLS algorithm is adjusted by preset calibration parameters. These preset calibration parameters include the smoothing factor (λ), the difference order (porder), and the maximum number of iterations (N). max At least one of the following: The smoothing factor can be used to control the baseline smoothness to prevent it from overfitting noise or signal peaks, and is typically set between 1000 and 100000, for example, λ=10000; the difference order can be used to set the baseline penalty term, which can be set to 1 or 2, for example, porder=2; the maximum number of iterations can be used to determine the maximum number of iterations for algorithm convergence, and is typically set between 10 and 50, for example, N. max =20, and stop iterating after reaching the maximum number of iterations.

[0032] By presetting correction parameters and using the airPLS algorithm for baseline correction, it can flexibly adapt to various nonlinear and time-varying complex baseline shapes. While effectively removing background interference, it minimizes the disturbance to the inherent line shape and precise position of the resonance peak signal itself.

[0033] Step S13: Extract the candidate resonance peak from the corrected spectral data based on a preset height threshold and a preset local window threshold.

[0034] From the calibrated spectral data, low-intensity noise fluctuations are filtered out based on a preset height threshold, and the peak search neighborhood is limited based on a preset local window threshold to help distinguish spatially (e.g.) Figure 3 (On the wavelength axis) adjacent to the resonance peak, identify and extract all candidate resonance peaks that meet the conditions, and record the precise position information of each candidate resonance peak (including the corresponding precise position information before correction, i.e., wavelength value information).

[0035] More specifically, the preset height threshold can be set to 0.1-1, and the preset local window threshold can be set to 30-100nm. This is merely an example and not intended to limit the implementation of the invention. Figure 3 Taking a single frame of data as an example, the preset height threshold is set to 0.5, and the preset local window threshold is set to 5nm. That is, with a wavelength window interval of 5nm, peaks with a height exceeding 0.5 in the search and correction spectral data are identified as candidate resonance peaks and recorded for each wavelength window. Figure 5 The time point and the precise location information (wavelength value information) of the candidate resonance peaks found.

[0036] Iterate through each frame of the calibrated spectral data, find all local maximum points of peak signals that satisfy the preset height threshold and preset local window threshold, and record the precise wavelength position information of these candidate resonance peaks.

[0037] In another feasible implementation method, multiple peaks can be determined from time-series spectral data using multi-peak Gaussian fitting and centroid peak finding algorithms, which are then used as candidate resonance peaks.

[0038] Step S14: Arrange the extracted candidate resonance peaks in chronological order to construct a time-peak dataset.

[0039] Based on the time information and corresponding candidate resonances of each frame of corrected spectral data, a time-peak dataset is constructed, arranged chronologically. This dataset includes time frames and multiple candidate resonances corresponding to each time frame. It is understood that a single time point may correspond to several candidate resonances, and the initial dataset typically contains a mixture of real resonance signal points and a large amount of noise and stray data points. When all points in this dataset are visualized as a scatter plot, it presents the following... Figure 4 Example: A state containing a large number of diffusely distributed gray stray points.

[0040] Step S2: Based on the time-peak dataset, and using the segmented anchoring rule, a dynamic trajectory tracking algorithm is employed to determine at least one anchoring trajectory and its corresponding time-peak data subset.

[0041] This embodiment employs a dynamic trajectory tracking algorithm with intermittent processing capabilities to analyze the time-peak dataset generated in step S1. Based on segmented anchoring rules, it determines at least one anchoring trajectory with good temporal continuity and relatively stable signal, i.e., the corresponding time-peak data subset, as the data basis for subsequent iterative analysis. Specifically, this includes steps S21-S24.

[0042] Step S21: Based on the time-peak dataset, execute the segmented anchoring rule to determine the current time frame and the initial resonance peak.

[0043] The segmented anchoring rules include the first segmentation rule and the first anchoring rule. Based on the time-peak dataset, the segmented anchoring rules are executed to determine the current time frame and the initial formant, including steps S211-S213.

[0044] Step S211: Determine whether this is the first time the segmented anchoring rule has been executed.

[0045] Specifically, each time the segmented anchoring rule is executed, the system or device executing the method of this embodiment can be queried to see if the segmented anchoring rule has been executed within a specified time, such as within 30 minutes. If not, it is determined to be the first execution of the segmented anchoring rule. Alternatively, an execution record table of the segmented anchoring rule can be established in the system or device executing the method of this embodiment, and the number of executions recorded in the execution record table can be used for judgment. There is no limitation here.

[0046] Step S212: If this is the first time the segmented anchoring rule is executed, then the current time frame is determined from the time-peak dataset based on the first segmentation rule.

[0047] Specifically, if this is the first time the segmented anchoring rule is executed, then the first time frame of the time-peak dataset is determined as the current time frame according to the first segmentation rule.

[0048] Step S213: Select an initial resonance peak from all the candidate resonance peaks corresponding to the current time frame according to the first anchoring rule.

[0049] Specifically, from multiple candidate resonants corresponding to the current time frame, according to the first anchoring rule, the candidate resonant with the middle wavelength value is selected as the initial resonant. The "middle" can be either a candidate resonant in the center of its position or a candidate resonant whose wavelength value is closest to the average wavelength value. For example, if the wavelength values ​​of the five candidate resonants corresponding to the current time frame are 2, 3, 4, 7, and 14, the candidate resonant with a wavelength value of 4 can be selected as the initial resonant. Alternatively, the average wavelength value of the five candidate resonants can be calculated, which is 6, and the candidate resonant with a wavelength value of 7 that is closest to this average can be selected as the initial resonant.

[0050] Furthermore, the segmented anchoring rule also includes a second segmented rule and a second anchoring rule. Based on the time-peak dataset, the segmented anchoring rule is executed to determine the current time frame and the initial formant. The rule also includes steps S214-S215, which are used to determine the current time frame and the initial formant when the segmented anchoring rule is not being executed for the first time.

[0051] Step S214: If this is not the first time the segmented anchoring rule is executed, then obtain the current time period determined after the previous execution, and determine the current time frame from the current time period according to the current time period and the second segmentation rule.

[0052] Specifically, if this is not the first time the segmented anchoring rule is executed, the current time period determined in the previous execution is obtained. Based on the time frame range of the current time period, and according to the time overlap requirement set in the second segmentation rule, a time frame is selected from the current time period as the current time frame. The time overlap requirement ensures that there is some overlap between the current time frames determined by each execution of the segmented anchoring rule, which is beneficial for judging the continuity and effectiveness of subsequent segmented trajectories. For example, the time overlap requirement can be a fixed number of overlap frames, such as k overlap frames, in which case the kth time frame from the end of the current time period is selected as the current time frame; the time overlap requirement can also be a time frame overlap rate, for example, if the current time period has 100 time frames and the time frame overlap rate is 5%, then the 95th time frame is selected as the current time frame.

[0053] By requiring time overlap, it is ensured that data comparison and connection between time-peak data subsets are easily performed in subsequent processing. This ensures accurate and efficient data processing while effectively controlling computing resources, guaranteeing the quality of the final results and adapting to the needs of large-scale data analysis.

[0054] Step S215: Select an initial resonance peak from all the candidate resonance peaks corresponding to the current time frame according to the second anchoring rule.

[0055] Specifically, based on the current time frame determined in step S215, all corresponding candidate resonance peaks are obtained from the time-peak dataset, and a new initial resonance peak is selected according to the second anchoring rule. The second anchoring rule can be the same as the first anchoring rule, that is, selecting the candidate resonance peak with the middle wavelength value as the initial resonance peak.

[0056] Furthermore, the second anchoring rule may also include a resonance peak difference judgment, and a new initial resonance peak is selected based on the judgment result, specifically including steps A-B.

[0057] Step A: Obtain the anchored resonance peak corresponding to the current time frame and compare its difference with that of each candidate resonance peak corresponding to the current time frame.

[0058] Specifically, from the anchoring trajectory determined in the previous execution of the segmented anchoring rule, the anchoring resonance peak corresponding to the current time frame determined in step S215 is compared with the difference between the current time frame and each candidate resonance peak in the time-peak dataset. For example, the difference is judged based on the wavelength value difference; the larger the difference, the greater the difference.

[0059] Step B: Based on the difference, select the initial resonance peak according to the second anchoring rule.

[0060] Specifically, based on the differences determined in step A, the candidate resonance peak with the largest difference is selected as the new initial resonance peak according to the second anchoring rule.

[0061] The selection of initial resonance peaks is achieved by judging differences, thereby enabling iterative updates of the anchor trajectory and exploring regions in the time-peak dataset that have not yet been divided.

[0062] Step S22: Based on the initial resonance peak, starting from the next time frame of the current time frame, the dynamic trajectory tracking algorithm is used to search for and anchor the resonance peak frame by frame in the time-peak dataset in chronological order.

[0063] Based on the initial resonance peak and the current time frame, a dynamic trajectory tracking algorithm is used to search for anchor resonance peaks frame by frame in chronological order, finding a trajectory with good effectiveness and stability as the benchmark for subsequent relative wavelength differences, thereby achieving accurate clustering analysis and trajectory identification and separation. Specifically, this includes steps S221-S225.

[0064] Step S221: Use the initial resonance peak as the current resonance peak.

[0065] The initial resonance peak is used as the current resonance peak, serving as the starting point for frame-by-frame search by the dynamic trajectory tracking algorithm.

[0066] Step S222: Based on the wavelength jump threshold, determine the connection validity between each candidate resonance peak corresponding to the next time frame and the current resonance peak.

[0067] Specifically, the dynamic trajectory tracking algorithm uses a wavelength jump threshold to determine the connection validity, obtains all candidate resonance peaks corresponding to the next time frame, and judges the connection validity between these candidate resonance peaks and the initial resonance peak one by one, that is, judges whether the wavelength difference of these candidate resonance peaks is greater than the wavelength jump threshold.

[0068] The wavelength jump threshold is the maximum jump in wavelength value that occurs during the drift or evolution of the resonant peak position. By setting a preset wavelength jump threshold, it is determined whether candidate resonants in adjacent time frames belong to the same trajectory, thereby improving the accuracy of the anchoring trajectory. For example, the wavelength jump threshold can be set to 0.01-0.5.

[0069] Step S223: Update the current time frame and the current resonance peak based on the validity judgment result.

[0070] Specifically, if among all candidate formants corresponding to the next time frame, there exists a candidate formant that satisfies the connection validity requirement with the current formant, then the anchor formant for the next time frame is determined based on the candidate formant that satisfies the connection validity requirement. Simultaneously, the current time frame is updated to the next time frame, and the current formant is updated to the anchor formant corresponding to the next time frame. Specifically, when there is only one candidate formant that satisfies the validity requirement, that candidate formant that satisfies the continuity validity requirement is used as the anchor formant for the next time frame; when there are multiple candidate formants that satisfy the validity requirement, the candidate formant with the smallest wavelength difference from the current formant is selected as the anchor formant for the next time frame.

[0071] If all candidate resonants in the next time frame do not meet the connection validity requirement with the current resonant, then the next time frame is treated as an intermittent time frame. The search for anchor resonants is then performed on the next-next time frame. After finding a corresponding anchor resonant, the current time frame is updated to the next-next time frame, and the current resonant is updated to the anchor resonant corresponding to the next-next time frame. Specifically, when searching for anchor resonants in the next-next time frame, the connection validity between each candidate resonant and the current resonant is determined based on the wavelength transition threshold. Candidate resonants that satisfy the connection validity requirement with the current resonant are used as the anchor resonants for the next-next time frame.

[0072] Furthermore, the next time frame is treated as an intermittent time frame, and its state is temporarily cached (forming an intermittent buffer sequence). If all candidate formants in the next-next time frame do not meet the connectivity validity requirement with the current formant, then the next-next time frame becomes another new intermittent time frame, recorded as a continuous intermittent frame. The number of consecutive intermittent time frames is 2. The search continues in chronological order. After finding the corresponding anchored formant, the current time frame is updated to the next-next time frame, and the current formant is updated to the anchored formant corresponding to the next-next time frame. The principle is the same as the search method for the next-next time frame, and will not be repeated here.

[0073] Step S224: Determine whether the search stopping condition has been met. If the search condition has not been met, continue to search for anchored resonant peaks frame by frame in the time-peak dataset in chronological order. When the search stopping condition is met, output the initial resonant peak and the anchored resonant peak.

[0074] In this embodiment, the search stopping conditions include time-based stopping conditions and discontinuity-based stopping conditions. Time-based stopping conditions include the current time frame being the last time frame in the time-peak dataset, and discontinuous time frames being the last time frames in the time-peak dataset. Discontinuity-based stopping conditions include the number of consecutive discontinuous time frames reaching a missing frame threshold. Meeting either condition constitutes a search stopping condition. The missing frame threshold is a preset maximum number of frames allowed for continuous interpolation, which can be set to 10-100,000. The larger the data volume in the time-peak dataset, the larger the missing frame threshold. For example, setting 1 / 100 of the data volume in the time-peak dataset as the missing frame threshold.

[0075] After updating the current time frame and the current formant, it is determined whether the search stopping condition has been met. If the search condition has not been met, the search for anchor formants continues frame by frame in the time-peak dataset, starting from the next time frame after the current time frame, in chronological order. When the time-based stopping condition is met, the initial formant and all anchor formants are output as the basis for anchor trajectory generation, and a segmented anchoring rule execution is recorded as complete. When the discontinuity-based stopping condition is met, the initial formant and all anchor formants before the last continuous discontinuity (excluding the current formant) are output as the basis for anchor trajectory generation, and a segmented anchoring rule execution is also recorded as complete.

[0076] Step S23: Based on the initial resonance peak and all the anchoring resonance peaks, generate the anchoring trajectory and determine the current time period covered by the anchoring trajectory.

[0077] Once the search stops, the initial resonance peak and all anchor resonance peaks output in step S22 are used to generate the anchor trajectory, and the current time period covered by the anchor trajectory is determined, specifically including steps S231-S234.

[0078] Step S231: Determine the temporal continuity of the initial resonance peak and all anchored resonance peaks.

[0079] Specifically, the temporal continuity between the initial formant and all anchored formants can be determined by checking whether the number of frames within the time range between the time frame corresponding to the initial formant and the time frame corresponding to the last anchored formant is the same as the number of frames within the corresponding time range in the time-peak dataset, or whether the time frames correspond one-to-one. Alternatively, it can be determined by checking whether there are discontinuous time frames within the time range covered by the initial formant and the last anchored formant during the search process.

[0080] Step S232: When the initial resonance peak and all anchored resonance peaks meet the time continuity, generate the anchoring trajectory based on the initial resonance peak and all anchored resonance peaks.

[0081] When the initial resonance peak and all anchor resonance peaks are time-continuous, the initial resonance peak and all anchor resonance peaks are connected in chronological order to form the anchor trajectory. It should be noted that the connection mentioned in this embodiment can be either an actual connection of points as a line or a continuous storage of points as a data sequence in chronological order; no limitation is made here.

[0082] Step S233: When the initial resonance peak and all anchored resonance peaks do not conform to time continuity, the second interpolation method is used to determine the supplementary anchored peaks. Based on the initial resonance peak, all anchored resonance peaks and the supplementary anchored peaks, the anchoring trajectory is generated.

[0083] In this embodiment, the determination of whether there is an intermittent time frame is used as an example. When the initial resonance peak and all anchor resonance peaks do not have time continuity, that is, there is an intermittent time frame, the second interpolation method is used to determine the completion of the anchor peak, and then the anchor trajectory is determined to improve the integrity and continuity of the anchor trajectory. Specifically, it includes steps C to F.

[0084] Step C: Determine the evolution trend of the anchored resonant peaks based on the anchored resonant peaks, and calculate the estimated wavelength values ​​of the resonant peaks corresponding to the missing time frames. Specifically, based on the anchored resonant peaks that have been searched, the evolution trend of the resonant peaks is determined through linear fitting or polynomial fitting, and then the estimated wavelength values ​​of the resonant peaks corresponding to the intermittent time frames are calculated. Specifically, the anchored resonant peaks of adjacent time frames can be selected for fitting based on the intermittent time frames, or all anchored resonant peaks found in this search can be selected for fitting. The specific fitting method can use existing techniques, which will not be elaborated here.

[0085] Step D: Extract wavelengths from each frame of aerosol spectral data to construct a corresponding set of effective wavelengths. Specifically, for each frame of aerosol spectral data, read the original wavelength axis data to obtain multiple wavelength values, construct a corresponding set of effective wavelengths, and use this set as a reference for determining the anchoring and filling resonance peaks of intermittent time frames. This set includes each abscissa point in the aerosol spectral data (e.g., ...). Figure 3 The original wavelength axis data corresponding to the data shown constitute a stable, discrete set of effective wavelengths. Since the detector pixels or sampling channels are fixed, all spectral data for the same single aerosol particle experiment are based on this fixed set of wavelength values ​​(i.e., wavelength values). This set of wavelength values ​​is defined as the "effective wavelength set". It should be noted that this set may differ for different droplet experiments due to instrument calibration or setting fine-tuning; therefore, different droplet experiments should be recorded independently.

[0086] Step E: Construct a time-effective wavelength dataset based on the effective wavelength set for each frame. Specifically, obtain the effective wavelength set of each frame of aerosol spectral data, arrange them in chronological order, and construct a time-effective wavelength dataset.

[0087] Step F: Determine the anchor peak for completion based on the time-effective wavelength dataset and the estimated wavelength of the formant. Specifically, considering that the true wavelength value of the formant should only fall on a fixed sampling channel, i.e., the effective wavelength set, the effective wavelength set corresponding to the intermittent time frames is obtained from the time-effective wavelength dataset and compared. The original wavelength axis data (wavelength value) with the smallest difference from the estimated wavelength of the formant is selected as the wavelength value for completing the anchor peak, thus determining the anchor peak for completion. It can be understood that the position of the formant is determined by its wavelength value; therefore, determining the wavelength value for completing the anchor peak is equivalent to determining the anchor peak for completion.

[0088] Furthermore, the difference between the original wavelength axis data and the anchored resonant peak of the previous time frame is not greater than the wavelength jump threshold can be used as the completion constraint. The original wavelength axis data that meets the completion constraint is used as the final anchored completion resonant peak wavelength value. For the original wavelength axis data that does not meet the completion constraint, the wavelength value is corrected. For example, another original wavelength axis data that meets the wavelength jump threshold and is closest to the current original wavelength axis data is selected as the anchored completion resonant peak wavelength value.

[0089] The initial resonance peak, all anchored resonance peaks, and the completed anchored peaks are connected in chronological order to form the anchoring trajectory.

[0090] The anchored and supplemented resonant peaks were determined by combining the time-effective wavelength dataset with the evolution trend of the anchored resonant peaks, ensuring that the resonant peaks determined by the interpolation method strictly originated from the wavelength values ​​that the experimental instruments could actually resolve and record. This not only eliminated non-physical data generated by mathematical interpolation but also guaranteed a high degree of consistency between the reconstructed trajectory and the original experimental observations in terms of data structure, avoiding systematic errors in subsequent physical inversions (such as particle size calculations). In other embodiments, the time-effective wavelength dataset can also be used as the basis for baseline correction and resonant peak localization.

[0091] Step S234: Determine the current time period covered by the anchoring trajectory.

[0092] The time frame corresponding to the initial resonance peak is taken as the start, and the time frame corresponding to the last anchor resonance point is taken as the end. The time range covered is the current time period.

[0093] Step S24: Based on the current time period, determine the corresponding time-peak data subset.

[0094] Based on the current time period determined in step S23, the data of the corresponding time period in the time-peak dataset is taken as the corresponding time-peak data subset.

[0095] It is understandable that when the discontinuity-type stopping condition is met in step S224, it indicates that the time-peak dataset has not been completely searched. For the unsearched parts of the time-peak dataset, the current time frame and initial formant are determined according to step S21, and steps S22-S24 are executed sequentially until the time-type stopping condition is met, that is, the search of each frame of the time-peak dataset is completed, and the time-peak dataset is divided into multiple time-peak data subsets without omission. At least one anchor trajectory and the corresponding time-peak data subset are determined. Each subset has a stable, continuous, and effective anchor trajectory. According to the constraints of the segmented anchoring rules, there is a certain overlap between the subsets, which facilitates the tracking and updating of subsequent segmented trajectories.

[0096] Step S3: Based on the anchoring trajectory, calculate the relative wavelength difference of each candidate resonance peak in the corresponding time-peak data subset.

[0097] First, the wavelength values ​​of the initial resonance peak and each of the anchoring resonance peaks in the anchoring trajectory are extracted as anchoring wavelength values. Using the wavelength values ​​of the initial resonance peak and each of the anchoring resonance peaks in the anchoring trajectory as anchoring wavelength values ​​serves as the benchmark for subsequent calculations of the relative wavelength difference.

[0098] Next, a frame is randomly selected from the corresponding time-peak data subset as the target time frame. Simultaneously, the wavelength value of each candidate resonance peak corresponding to the target time frame is extracted, and the difference between this wavelength and the anchored wavelength value corresponding to the target time frame is calculated to obtain the relative wavelength difference for each candidate resonance peak corresponding to the target time frame. For example, for any candidate resonance peak in the target time frame, its wavelength is denoted as the candidate wavelength value (time). The anchored wavelength value of the anchored trajectory in the target time frame is found, and its wavelength is denoted as the anchored wavelength value (time). The relative wavelength difference between the two is calculated: Difference (time) = Candidate wavelength value (time) - Anchored wavelength value (time).

[0099] Finally, each frame in the time-peak data subset is traversed to obtain the relative wavelength difference of each candidate resonance peak in the time-peak data subset. Using the anchor trajectory as a reference, all candidate resonance peaks are traversed, and the relative wavelength difference corresponding to each candidate resonance peak is calculated.

[0100] Furthermore, the relative wavelength differences corresponding to each time-peak data subset are mapped to a relative wavelength difference space, which facilitates data acquisition during subsequent cluster analysis. The relative wavelength difference spaces corresponding to different time-peak data subsets can be independent of each other, or they can be merged into a total relative wavelength difference space.

[0101] Understandably, all candidate resonants in the time-peak dataset subset, including anchor resonants, have a relative wavelength difference of 0. Before calculating the relative wavelength difference, all candidate resonants selected as anchor resonants can be removed from the time-peak dataset subset. Alternatively, duplicate data with a relative wavelength difference of 0 can be removed, or no removal can be performed. Segmented trajectories can be generated subsequently, and finally, judgment and processing can be performed when determining the resonant trajectory.

[0102] Step S4: Perform cluster analysis on all relative wavelength differences corresponding to the time-peak data subset to determine the corresponding segmented trajectory.

[0103] By leveraging the physical constraint of stable relative distances between resonance peaks, cluster analysis is performed on all relative wavelength differences corresponding to subsets. This identifies and clusters candidate peak sites with similar relative wavelength differences into the same stable difference cluster, effectively distinguishing trajectory clusters representing different trajectories from discrete points representing noise. Through analysis of each cluster, multiple segmented trajectories are identified, effectively addressing the complex situations commonly found in real-world spectral data, such as transient signal loss, numerous stray points, and multiple concurrent, close, or even intersecting physical trajectories. Specifically, this includes: Step S41: Cluster the relative wavelength differences to obtain at least one segmented cluster; wherein the segmented cluster includes multiple relative wavelength differences.

[0104] In this embodiment, the clustering analysis employs the density-based clustering algorithm (DBSCAN algorithm). Points with similar and sufficiently dense relative wavelength differences are identified as the same cluster (i.e., a cluster). Clustering all relative wavelength differences yields at least one cluster, and each cluster includes multiple relative wavelength differences. Each identified cluster is expected to correspond to a segmented trajectory that maintains a stable relative distance from the anchoring trajectory. Discrete points that do not belong to any cluster (which can be marked as noise or outliers by the clustering algorithm) are identified as stray signals and discarded.

[0105] The DBSCAN algorithm is constrained by preset clustering parameters. In this embodiment, the preset clustering parameters include the cluster radius (eps) and the minimum number of samples (min_samples), which are used to limit the judgment of similarity and density of relative wavelength differences, respectively. The cluster radius, also known as the distance threshold, is used to define the distance threshold parameter of the neighborhood range, corresponding to the allowable fluctuation range of the relative difference. It can be the same as the wavelength jump threshold, different, or generated based on the DBSCAN algorithm. The minimum number of samples, also known as the minimum number of neighborhood samples, is used to define the number of core points, corresponding to the minimum length or density requirement of the effective cluster. It can be set to the number of frames in the time-peak data subset, or 70% of the number of frames, to ensure that the amount of data within the same cluster closely matches the actual number of resonant peaks, avoiding the identification of scattered noise and occasional relative wavelength differences as effective clusters.

[0106] Step S42: Obtain the candidate resonance peak corresponding to each relative wavelength difference in the segmented cluster, as a segmented resonance peak cluster.

[0107] Based on the identified segmented clusters, the candidate resonance peaks corresponding to each relative wavelength difference in each segmented cluster are obtained, thus obtaining the segmented resonance peak clusters corresponding to each segmented cluster.

[0108] Step S43: Determine the segmented trajectory based on the segmented resonance peak cluster.

[0109] For each segmented resonance peak cluster, connect all candidate resonance peaks within the cluster in chronological order to form a segmented trajectory. By traversing all segmented resonance peak clusters, you can obtain all segmented trajectories corresponding to the time-peak position data subset.

[0110] Furthermore, for all candidate resonance peaks in each segmented resonance peak cluster, the validity is judged using the wavelength jump threshold parameter, and valid candidate resonance peaks are retained to form a segmented trajectory.

[0111] Furthermore, to improve the continuity and integrity of the segmented trajectories, the segmented trajectories are determined based on the segmented resonant clusters, including: Step S431: Detect the temporal continuity between the candidate resonance peaks in the segmented resonance peak cluster.

[0112] Specifically, the temporal continuity between candidate resonants in a segmented resonant cluster can be determined by comparing the number of frames within the time range between the time frame corresponding to the first candidate resonant and the time frame corresponding to the last candidate resonant in the cluster with the number of frames within the corresponding time range in the time-peak data subset, or with the number of frames within the corresponding time range in the effective wavelength data set, or by ensuring a one-to-one correspondence between the time frames. If there are uncorresponding time frames, it is considered that all candidate resonants in the segmented resonant cluster do not exhibit temporal continuity, and the corresponding time frames are considered discontinuous time frames in the segmented resonant cluster that do not exhibit temporal continuity.

[0113] Step S432: For the two candidate resonance peaks that do not conform to time continuity, use the first interpolation method to determine the complete resonance peak.

[0114] For the two candidate resonants that do not conform to time continuity, namely the last candidate resonant before the time frame is discontinuous in the segmented resonant cluster and the first candidate resonant after the time frame is discontinuous, the corresponding filler resonant is determined for the discontinuous time frame using the first interpolation method.

[0115] First, based on all candidate resonants of the segmented resonant cluster, the evolution trend of the segmented trajectory is determined, and the wavelength estimates of the resonants corresponding to the intermittent time frames are calculated. Specifically, based on all candidate resonants of the segmented resonant cluster, the evolution trend of the segmented trajectory is determined through linear fitting or polynomial fitting, and the wavelength estimates of the resonants corresponding to the intermittent time frames are calculated. This can be referred to step C above, and will not be repeated here.

[0116] Secondly, the missing resonant peaks are determined based on the time-effective wavelength dataset and the estimated resonant peak wavelengths. The time-effective wavelength dataset can be obtained through the aforementioned steps D and E, and then the missing resonant peaks can be determined. The specific principle is the same as that of step F, and will not be repeated here.

[0117] Furthermore, for the segmented trajectories of each time-peak data subset, the physical constraint of stable relative distance between resonance peaks can be used to re-verify the completion of resonance peaks.

[0118] Furthermore, actual resonance peak trajectories may disappear or reappear over time. To avoid over-filling such trajectories, after determining the interrupted time frame, the continuity of the time frame and the start and end time frames can be judged. Specifically, if there are consecutive time frames within the interrupted time frame, it is determined whether it starts with the first time frame of the time-peak data subset or ends with the last time frame of the time-peak data subset. If so, this consecutive time frame is not considered an exception, and the determination to fill in the resonance peak is not performed. Alternatively, interrupted time frames with more than a missing frame threshold can also be considered exceptions.

[0119] Step S433: Add the completed resonance peak to the segmented resonance peak cluster to form a calibrated segmented resonance peak cluster.

[0120] The resonance peaks will be supplemented and added to the segmented resonance peak clusters in chronological order to obtain the calibrated segmented resonance peak clusters.

[0121] Step S434: Determine the segmented trajectory based on the calibration segmented resonance peak cluster.

[0122] For a calibrated segmented resonance peak cluster, connect all candidate resonance peaks and the supplementary resonance peaks within the cluster in chronological order to form a segmented trajectory.

[0123] Step S5: Determine the resonance peak trajectory based on the segmented trajectories corresponding to all the time-peak data subsets.

[0124] After determining a subset of time-peak data in step S22, steps S3-S5 can be performed directly on that subset, or steps S3-S5 can be performed separately after all subsets of time-peak data have been confirmed; there is no restriction here. This embodiment will illustrate the process by performing steps S3-S5 after each confirmed subset of time-peak data.

[0125] Step S51: For each anchored trajectory and segmented trajectory, determine the trajectory continuity with each existing trajectory in the trajectory set.

[0126] For each execution of the anchored trajectory and segmented trajectory determined in steps S2-S5, a trajectory continuity check is performed against the existing trajectories in the trajectory set. It is understandable that for the first execution of the anchored trajectory and segmented trajectory determined in steps S2-S5, the trajectory set is empty; therefore, the anchored trajectory and segmented trajectory are directly treated as existing trajectories, and the trajectory set is updated.

[0127] Specifically, based on the overlapping time intervals determined by the time overlap requirements in the segmented anchoring rules, two adjacent time-peak data subsets including this overlapping time interval are identified. For the anchored trajectories of these two adjacent time-peak data subsets and all segmented trajectories, trajectory continuity is assessed. This process of identifying and assessing the continuity of adjacent time-peak data subsets is performed for all overlapping time intervals to achieve comprehensive coverage of the time-peak dataset and ensure the integrity and continuity of the formant trajectory. For example, if each time-peak data subset is processed sequentially according to time order, the anchored and segmented trajectories corresponding to the earlier time-peak data subset in two adjacent time-peak data subsets are already in the trajectory set, i.e., they are existing trajectories. The anchored and segmented trajectories corresponding to the later time-peak data subset are then compared with all segmented trajectories in the trajectory set for trajectory continuity assessment.

[0128] Specifically, trajectory continuity is determined by checking whether the candidate resonance peaks of the anchored trajectory, segmented trajectory, and each existing trajectory overlap within the overlapping time interval. For example, for any segmented trajectory, candidate resonance peaks within the overlapping time interval are extracted. Similarly, for each existing trajectory, candidate resonance peaks within the overlapping time interval are extracted. The wavelength values ​​of the candidate resonance peaks in the segmented trajectory are checked frame by frame in chronological order to see if they are the same as the wavelength values ​​of the candidate peaks in the existing trajectory. In this embodiment, if the wavelength difference is less than the similarity threshold, they are considered the same. From the existing trajectories where the number of candidate resonance peaks reaches the quantity threshold, the existing trajectory with the highest number of identical peaks is selected and considered to be consistent with the segmented trajectory, meaning the two trajectories are different parts of the same physical trajectory. The similarity threshold and quantity threshold can be set based on actual conditions and are not restricted here. If no existing trajectory has a number of candidate resonance peaks reaching the quantity threshold, then no existing trajectory is considered to be consistent with the anchored trajectory or the segmented trajectory.

[0129] In other embodiments, candidate resonance peaks in overlapping time intervals can be extracted for any segmented trajectory or anchored trajectory, and the wavelength mean can be calculated. For each existing trajectory, candidate resonance peaks in overlapping time intervals can also be extracted, and the wavelength mean can be calculated respectively. The continuity of the trajectory can be judged by comparing the wavelength mean corresponding to the segmented trajectory or anchored trajectory with the wavelength mean corresponding to each existing trajectory.

[0130] Step S52: If the anchored trajectory or the segmented trajectory meets the trajectory continuity requirement with the existing trajectory, then the existing trajectory is updated based on the anchored trajectory or the segmented trajectory.

[0131] For an existing trajectory, if it is determined to have trajectory continuity with a certain anchored trajectory or segmented trajectory, then the time frame of that anchored trajectory or segmented trajectory and its corresponding candidate formants are added to the existing trajectory, and the existing trajectory is updated. For overlapping time intervals, either the candidate formants of the segmented trajectory or anchored trajectory or the candidate formants of the existing trajectory are retained; there is no restriction here.

[0132] After dividing the time-peak data subsets, the segmented trajectories that were divided due to the data division are updated by judging the continuity of the anchored trajectory, segmented trajectory and existing trajectory, so as to extend the trajectory on the time line. By updating and iterating the segmented trajectories corresponding to all time-peak data subsets, trajectory tracking is completed to support the analysis of large-scale data.

[0133] Step S53: If the anchored trajectory or the segmented trajectory does not conform to trajectory continuity with all the existing trajectories, then the anchored trajectory or the segmented trajectory is used as a new existing trajectory, and the trajectory set is updated.

[0134] If no existing trajectory matches the anchor trajectory or segmented trajectory in terms of trajectory continuity, then the anchor trajectory or segmented trajectory will be treated as a new existing trajectory, and the trajectory set will be updated.

[0135] Step S54: Traverse all the anchored trajectories and segmented trajectories, and determine the existing trajectory in the updated trajectory set as the formant trajectory.

[0136] By traversing all anchored trajectories and segmented trajectories, the existing trajectories in the updated trajectory set are identified as resonance peak trajectories, which can be used for inversion calculations to determine key physical parameters such as aerosol-related particle size and complex refractive index.

[0137] Using the method of this embodiment, based on the physical constraint that the relative wavelength distance between resonance peaks usually remains stable over time, the relative wavelength difference corresponding to each candidate resonance peak is calculated. From the candidate peak position dataset containing a large amount of noise, all hidden real resonance peak trajectories are accurately identified and separated, and these trajectories are repaired and improved.

[0138] The resonance peak trajectories obtained by the method of this invention not only effectively remove background noise and stray point interference, repair internal time discontinuities, and successfully separate multiple concurrent physical trajectories, but also ensure, through interpolation alignment operations, that all recorded wavelength values ​​accurately correspond to the physical measurement grid points of the spectrometer. These high-quality trajectory data (typically organized as a data structure containing the time series and corresponding wavelength series of each trajectory) are output and saved in a standardized data format (such as a text file).

[0139] This invention provides an aerosol optical tweezers resonance peak trajectory recognition system based on iterative clustering, comprising: The dataset generation module acquires multiple frames of aerosol spectral data and constructs a time-peak dataset; wherein, the time-peak dataset includes time frames and multiple candidate resonance peaks corresponding to each time frame; The anchor trajectory determination module, based on the time-peak dataset and a segmented anchoring rule, uses a dynamic trajectory tracking algorithm to determine at least one anchor trajectory and its corresponding time-peak data subset. The wavelength difference calculation module is used to calculate the relative wavelength difference of each candidate resonance peak in the corresponding time-peak data subset based on the anchoring trajectory. The segmented trajectory determination module is used to perform cluster analysis on all relative wavelength differences corresponding to the time-peak data subset to determine the corresponding segmented trajectory. The trajectory determination module determines the resonance peak trajectory based on the anchored trajectory and the segmented trajectory corresponding to all the time-peak data subsets.

[0140] The above-described method and system embodiments are based on the same principles, and their related aspects can be referenced from each other to achieve the same technical effects. For specific implementation processes, please refer to the foregoing embodiments, which will not be repeated here.

[0141] In summary, the aerosol spectral resonance peak trajectory identification method and system based on cluster analysis according to embodiments of the present invention has at least one of the following beneficial effects: 1. By using segmented anchoring rules, segmented processing of spectral data with particularly long time series or extremely high data point density is achieved, i.e., time-peak data subsets. For each subset, the relative wavelength difference is calculated based on the anchoring trajectory. Further cluster analysis is performed to determine the segmented trajectory, and then the resonance peak trajectory is determined. Using the physical constraint of stable relative distance between resonance peaks, clustering denoising and trajectory reconstruction are performed to achieve automated and high-precision identification of multiple resonance peak trajectories in complex noisy spectral data. The original accuracy of peak position information is preserved, improving data processing efficiency and the accuracy, objectivity and repeatability of subsequent aerosol physical parameter inversion.

[0142] 2. By using segmented anchoring rules and dynamic trajectory tracking algorithms, the anchoring trajectory and the corresponding time-peak data subset are determined, ensuring that the anchoring trajectory within each subset is effective, stable, and highly continuous. This effectively addresses the complexities in actual spectral data and serves as a benchmark for subsequent relative wavelength differences, thereby achieving accurate segmented clustering analysis and segmented trajectory identification and separation.

[0143] 3. By clustering relative wavelength differences to obtain clusters, similar and dense points can be identified and aggregated, thereby determining multiple segmented trajectories. This can effectively address complex situations commonly found in actual spectral data, such as transient signal loss, a large number of stray points, and multiple physical trajectories occurring concurrently, close to, or even intersecting. It has achieved accurate and automated identification of multiple resonance peak trajectories from noisy and complex spectral data.

[0144] 4. For segmented trajectories, the integrity of the segmented trajectories is ensured by judging time continuity and interpolation completion. Trajectory updates are performed by judging trajectory continuity to track the resonance peak trajectory. While effectively controlling computing resources, the accuracy and efficiency of data processing are guaranteed, the quality of the final results is ensured, and the needs of large-scale data analysis are met.

[0145] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0146] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for identifying the resonance peak trajectory of aerosol optical tweezers based on iterative clustering, characterized in that, include: Acquire multiple frames of aerosol spectral data to construct a time-peak dataset; wherein, the time-peak dataset includes time frames and multiple candidate resonance peaks corresponding to each time frame; Based on the time-peak dataset, and using a dynamic trajectory tracking algorithm based on segmented anchoring rules, at least one anchoring trajectory and its corresponding time-peak data subset are determined. Based on the anchoring trajectory, the relative wavelength difference of each candidate resonance peak in the corresponding time-peak data subset is calculated; Cluster analysis is performed on all relative wavelength differences corresponding to the time-peak data subset to determine the corresponding segmented trajectories; The resonance peak trajectory is determined based on the anchored trajectory and the segmented trajectory corresponding to all the time-peak data subsets.

2. The method according to claim 1, characterized in that, Based on the aforementioned time-peak dataset, and using a dynamic trajectory tracking algorithm based on segmented anchoring rules, at least one anchoring trajectory and its corresponding time-peak data subset are determined, including: Based on the time-peak dataset, segmented anchoring rules are executed to determine the current time frame and the initial formant. Based on the initial resonance peak, starting from the next time frame of the current time frame, the dynamic trajectory tracking algorithm is used to search for and locate the anchor resonance peak frame by frame in the time-peak dataset in chronological order. Based on the initial resonance peak and all the anchoring resonance peaks, the anchoring trajectory is generated, and the current time period covered by the anchoring trajectory is determined. Based on the current time period, determine the corresponding time-peak data subset.

3. The method according to claim 2, characterized in that, The segmented anchoring rules include a first segmentation rule and a first anchoring rule; Based on the time-peak dataset, segmented anchoring rules are executed to determine the current time frame and the initial resonance peak, including: Determine if this is the first time the segmented anchoring rule has been executed; If this is the first time the segmented anchoring rule is executed, the current time frame is determined from the time-peak dataset based on the first segmentation rule. From all the candidate resonance peaks corresponding to the current time frame, select the initial resonance peak according to the first anchoring rule.

4. The method according to claim 3, characterized in that, The segmented anchoring rule also includes a second segmentation rule and a second anchoring rule; Based on the time-peak dataset, segmented anchoring rules are executed to determine the current time frame and the initial formant, including: If this is not the first time the segmented anchoring rule is executed, then the current time period determined after the previous execution is obtained, and the current time frame is determined from the current time period based on the current time period and the second segmentation rule; From all the candidate resonance peaks corresponding to the current time frame, an initial resonance peak is selected according to the second anchoring rule.

5. The method according to claim 4, characterized in that, From all the candidate resonants corresponding to the current time frame, an initial resonant is selected according to the second anchoring rule, including: Obtain the anchored resonance peak corresponding to the current time frame, and compare it with the difference of each candidate resonance peak corresponding to the current time frame; Based on the aforementioned differences, an initial resonance peak is selected according to the second anchoring rule.

6. The method according to claim 2, characterized in that, Based on the anchoring trajectory, the relative wavelength difference of each candidate resonance peak in the corresponding time-peak data subset is calculated, including: The wavelength values ​​of the initial resonance peak and each of the anchoring resonance peaks in the anchoring trajectory are extracted as the anchoring wavelength values; Select one frame from the corresponding time-peak data subset as the target time frame, extract the wavelength value of each candidate resonance peak corresponding to the target time frame, and calculate the difference between the wavelength value and the anchor wavelength value corresponding to the target time frame to obtain the relative wavelength difference of each candidate resonance peak corresponding to the target time frame. By traversing each frame in the time-peak data subset, the relative wavelength difference of each candidate resonance peak in the time-peak data subset is obtained.

7. The method according to claim 1, characterized in that, Cluster analysis is performed on all relative wavelength differences corresponding to the time-peak data subset to determine the corresponding segmented trajectories, including: The relative wavelength differences are clustered to obtain at least one segmented cluster; wherein, the segmented cluster includes multiple relative wavelength differences; Obtain the candidate resonance peak corresponding to each relative wavelength difference in the segmented cluster, and use it as a segmented resonance peak cluster; The segmented trajectory is determined based on the segmented resonance peak cluster.

8. The method according to claim 7, characterized in that, Based on the segmented resonance peak cluster, the segmented trajectory is determined, including: Detect the temporal continuity between the candidate resonance peaks in the segmented resonance peak cluster; For two candidate resonance peaks that do not conform to time continuity, the first interpolation method is used to determine the complete resonance peak; The completed resonance peak is added to the segmented resonance peak cluster to form a calibration segmented resonance peak cluster; The segmented trajectory is determined based on the calibrated segmented resonance peak cluster.

9. The method according to claim 1, characterized in that, The resonance peak trajectory is determined based on the segmented trajectories corresponding to all the time-peak data subsets, including: For each anchored trajectory and each segmented trajectory, determine its continuity with each existing trajectory in the trajectory set; If the anchored trajectory or the segmented trajectory meets the trajectory continuity requirement with the existing trajectory, then the existing trajectory is updated based on the anchored trajectory or the segmented trajectory; If the anchored trajectory or the segmented trajectory does not conform to trajectory continuity with all the existing trajectories, then the anchored trajectory or the segmented trajectory is regarded as a new existing trajectory, and the trajectory set is updated. Traverse all the anchored trajectories and segmented trajectories, and determine the existing trajectories in the updated trajectory set as the formant trajectories.

10. A system for identifying the resonance peak trajectory of aerosol optical tweezers based on iterative clustering, characterized in that, include: The dataset generation module acquires multiple frames of aerosol spectral data and constructs a time-peak dataset; wherein, the time-peak dataset includes time frames and multiple candidate resonance peaks corresponding to each time frame; The anchor trajectory determination module, based on the time-peak dataset and a segmented anchoring rule, uses a dynamic trajectory tracking algorithm to determine at least one anchor trajectory and its corresponding time-peak data subset. The wavelength difference calculation module is used to calculate the relative wavelength difference of each candidate resonance peak in the corresponding time-peak data subset based on the anchoring trajectory. The segmented trajectory determination module is used to perform cluster analysis on all relative wavelength differences corresponding to the time-peak data subset to determine the corresponding segmented trajectory. The trajectory determination module determines the resonance peak trajectory based on the anchored trajectory and the segmented trajectory corresponding to all the time-peak data subsets.

Images