Multi-parameter coupling detection method and related equipment of integrated optical sensor

Abstract

The present invention discloses a multi-parameter coupling detection method and related equipment of an integrated optical sensor, the detection method comprising: injecting different combinations of trace drugs into the sample to be tested in a plurality of microfluidic channels according to a preset time sequence; performing phase shift coding acquisition on the differentiated optical response signal, splitting the broadband light source into a plurality of wavelength sub-intervals, and generating a coded light signal with a preset phase shift amount; calculating the cross-correlation function between channels, and performing signal amplification or suppression processing on the channel group with a high coupling degree; performing multi-dimensional feature extraction, and calculating the differential interference amount, amplitude ratio and phase synchronization between channels; monitoring the baseline drift and noise distribution within a preset time window, and performing parameter adjustment; performing self-learning evaluation on the detection data, updating the recognition model parameters, and generating a multi-parameter fingerprint spectrum. The technical solution of the present invention effectively solves the problems of signal aliasing and insufficient sensitivity in multi-parameter detection, and can perform high-precision synchronous detection of multiple targets.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-parameter optical sensors, and in particular to a multi-parameter coupling detection method of an integrated optical sensor and related equipment. Background Art

[0002] Multi-parameter optical sensors are usually composed of micro-nano chips, optical waveguides, microfluidic channels, and surface modification layers. By integrating different types of detection units on the same substrate, multiple target indicators (such as ion concentrations, protein markers, or volatile substances) can be monitored synchronously or sequentially. In actual use, the operator will introduce the sample to be tested (liquid or gas) into the microfluidic system to interact with the surface functionalized sensing area; at this time, different detection channels produce unique optical responses to specific external substances (such as changes in refractive index, fluorescence, or scattering intensity), and then transmit these response signals to the readout end via optical waveguides. Through data acquisition and analysis, quantitative or qualitative information of multiple parameters can be obtained in a relatively short time, providing efficient and real-time detection effects for environmental monitoring, biochemical diagnosis, and even industrial process control.

[0003] However, when the sensor faces a more complex multi-component system or a dynamically changing detection environment, conventional optical sensing modes often find it difficult to accurately identify multiple target parameters, especially when the responses of different channels overlap or interfere with each other, and data is prone to aliasing or decreased sensitivity. Since the coupling mechanism between multiple channels is very complex, and the concentration and physical and chemical properties of the target to be monitored in each channel are different, simple channel separation or single-point correction is not enough to eliminate cross-influences; coupled with the possibility of tiny temperature drifts and refractive index disturbances during the detection process, the signal coupling between multiple channels is further amplified. Therefore, how to flexibly regulate and finely manage each detection channel is becoming a core problem facing the development of multi-parameter optical sensors. Summary of the invention

[0004] The main purpose of the present invention is to solve the technical problem of multi-channel cross interference existing in the existing multi-parameter optical sensor.

[0005] A first aspect of the present invention provides a multi-parameter coupling detection method for an integrated optical sensor, the multi-parameter coupling detection method for an integrated optical sensor comprising:

[0006] Injecting different micro-agent combinations into the sample to be tested in a plurality of microfluidic channels according to a preset time sequence, and obtaining differentiated optical response signals of each channel through the selective reaction of the micro-agent combination with the sample to be tested;

[0007] Performing phase shift coding acquisition on the differentiated optical response signal, splitting the broadband light source into a plurality of wavelength sub-intervals, generating a coded light signal with a preset phase shift amount for each wavelength sub-interval, inputting the coded light signal into a corresponding detection channel according to a timing allocation scheme, and obtaining multi-channel spectral evolution data;

[0008] Calculating the cross-correlation function between channels according to the multi-channel spectral evolution data, grouping the detection channels based on the coupling characteristics of the cross-correlation function, performing signal amplification processing on the channel group with high coupling degree and positive correlation, and performing signal suppression processing on the channel group with high coupling degree and negative correlation, to obtain optimized multi-channel response data;

[0009] Performing multi-dimensional feature extraction on the optimized multi-channel response data, calculating the differential interference, amplitude ratio and phase synchronization between channels, constructing a high-dimensional feature vector in combination with the characteristic reaction time of the trace drug combination, and obtaining multi-dimensional feature data of the target object by analyzing the interaction mode of the high-dimensional feature vector;

[0010] Within a preset time window of the detection process, based on the multi-dimensional feature data of the target object, the baseline drift and noise distribution of each channel are monitored using phase-shift coded pulses. When the baseline drift exceeds a preset threshold, the phase-shift modulation parameters and the drug injection ratio are adjusted to obtain corrected detection data;

[0011] The corrected detection data is subjected to self-learning evaluation, wherein the self-learning evaluation includes updating the recognition model parameters by using a progressive stochastic gradient algorithm and a multivariate regression analysis, dynamically correcting the channel coupling strength and the feature weight based on the updated recognition model parameters, and generating a multi-parameter fingerprint spectrum characterizing the characteristics of the sample to be tested.

[0012] Optionally, injecting different trace drug combinations into the sample to be tested in a plurality of microfluidic channels according to a preset time sequence, and obtaining differentiated optical response signals of each channel through the selective reaction of the trace drug combination with the sample to be tested, comprises:

[0013] Acquire background component data of the sample to be tested, classify and mark the sample to be tested according to the background component data, and determine the reference ratio parameters of the sample to be tested;

[0014] Calculate the amount of medicine injected into each microfluidic channel according to the reference ratio parameter, set a corresponding micro-medicine combination for each microfluidic channel, and determine the injection sequence of the micro-medicine combination;

[0015] Performing a flow splitting process on the sample to be tested, respectively injecting the corresponding micro-agent combination into the plurality of microfluidic channels according to the injection sequence, and recording the injection time of the micro-agent combination;

[0016] Calculating the selective reaction time of the combination of the sample to be tested and the trace drug, and setting the data acquisition time window of each channel according to the injection time and the selective reaction time;

[0017] Tracking the reaction process of the sample to be tested according to the preset time sequence within the data acquisition time window, and recording the selective reaction data of the combination of the sample to be tested and the trace drug at different times;

[0018] The light intensity variation trend of each channel is calculated according to the selective response data, and the light intensity variation trend is normalized in combination with the reference ratio parameter of the sample to be tested to obtain the differentiated optical response signal of each channel.

[0019] Optionally, the phase-shift encoding acquisition is performed on the differentiated optical response signal, a broadband light source is split into a plurality of wavelength sub-intervals, a coded light signal with a preset phase shift is generated for each wavelength sub-interval, and the coded light signal is input into a corresponding detection channel according to a timing allocation scheme to obtain multi-channel spectral evolution data, including:

[0020] Performing spectrum decomposition on the differentiated optical response signal, setting a sampling interval for the differentiated optical response signal according to the characteristic frequency of each channel, and generating a multi-channel sampling sequence;

[0021] According to the frequency distribution of the multi-channel sampling sequence, the broadband light source is divided into sub-intervals, the broadband light source is split into a plurality of wavelength sub-intervals, and a spectral component of each wavelength sub-interval is obtained;

[0022] Performing phase modulation calculation on the spectral components, setting a phase offset parameter for each wavelength sub-interval according to the timing characteristics of the multi-channel sampling sequence, and generating an initial coded optical signal;

[0023] Calculating the phase compensation amount of each wavelength sub-interval according to the amplitude distribution of the differentiated optical response signal, performing phase compensation processing on the initial coded optical signal, and obtaining a coded optical signal with a preset phase shift amount;

[0024] Based on the time distribution characteristics of the multi-channel sampling sequence, the coded optical signal with the preset phase shift amount is time-sequence divided to generate an interleaved time sequence allocation scheme;

[0025] According to the interleaved timing distribution scheme, the coded optical signal with the preset phase shift amount is sequentially input into each detection channel, the spectrum response process of each channel is recorded, and multi-channel spectrum evolution data is obtained.

[0026] Optionally, the cross-correlation function between channels is calculated according to the multi-channel spectral evolution data, the detection channels are grouped based on the coupling characteristics of the cross-correlation function, a signal amplification process is performed on the channel group with high coupling and positive correlation, and a signal suppression process is performed on the channel group with high coupling and negative correlation to obtain optimized multi-channel response data, including:

[0027] Performing time series analysis on the multi-channel spectral evolution data, extracting the spectral response value of each detection channel at a corresponding moment according to a preset sampling period, and obtaining the spectral response sequence of the detection channel;

[0028] The spectral response sequence is divided into time windows, the sequence length is determined according to the preset analysis time, and the cross-correlation function between any two detection channels is calculated to obtain the coupling characteristic data between the channels, wherein the formula for calculating the cross-correlation function is as follows:

[0029] ;

[0030] in, For detection channel and detection channels The cross-correlation value of is the sequence length, and Detection channels and detection channels At sampling time The spectral response sequence, is the time delay, and ;

[0031] Threshold analysis is performed on the coupling characteristic data between the channels to calculate the cross-correlation function. =0, and divide the detection channels into high coupling degree channel pairs and low coupling degree channel pairs according to the preset coupling degree threshold, to obtain the channel grouping result;

[0032] According to the channel grouping result, the high coupling channel pairs are polarity classified, the channel pairs whose cross-correlation function peaks are greater than zero are divided into positive correlation channel groups, and the channel pairs whose cross-correlation function peaks are less than zero are divided into negative correlation channel groups, so as to obtain channel group polarity data;

[0033] Performing synchronous amplification processing on the spectral response sequence of the positive correlation channel group, calculating the gain coefficient of each channel pair according to the positive peak value of the cross-correlation function, performing amplitude modulation on the spectral response value, and obtaining positive correlation signal modulation data;

[0034] Suppression processing is performed on the spectral response sequence of the negative correlation channel group, the suppression coefficient of each channel pair is calculated according to the negative peak value of the cross-correlation function, the crosstalk component is selectively attenuated, and negative correlation signal suppression data is obtained;

[0035] The positive correlation signal modulation data and the negative correlation signal suppression data are integrated to obtain optimized multi-channel response data.

[0036] Optionally, performing multidimensional feature extraction on the optimized multi-channel response data, calculating the differential interference, amplitude ratio and phase synchronization between channels, combining the characteristic reaction time of the trace drug combination, constructing a high-dimensional feature vector, and obtaining multidimensional feature data of the target object by analyzing the interaction mode of the high-dimensional feature vector, including:

[0037] The optimized multi-channel response data is divided into characteristic frequency bands, and the spectral data of each channel is segmented according to the characteristic absorption intervals of heavy metal ions, organic matter and protein to obtain multi-interval spectral distribution data;

[0038] Calculating the differential interference between any two detection channels according to the multi-interval spectral distribution data, and analyzing the amplitude difference and phase difference between the channels using a coherent spectrum analysis method to obtain interference characteristic data between the channels;

[0039] Normalizing the interference characteristic data between the channels, calculating the amplitude ratio of adjacent detection channels, extracting the phase difference as a phase synchronization parameter, and obtaining channel correlation characteristic data;

[0040] Dynamically analyzing the reaction process between the trace drug combination and the target substance, calculating the characteristic reaction time according to the inflection point moment of the spectral response of each channel, and obtaining drug reaction characteristic data;

[0041] Combining the interference feature data between the channels, the channel association feature data and the drug reaction feature data to construct a high-dimensional feature vector, and obtaining feature space distribution data;

[0042] Cluster analysis is performed on the interaction patterns in the feature space distribution data, and feature parameters of each dimension are extracted according to the feature distribution rules of different target objects to obtain multi-dimensional feature data of the target objects.

[0043] Optionally, within a preset time window of the detection process, according to the multi-dimensional feature data of the target object, the baseline drift and noise distribution of each channel are monitored using phase-shifted coded pulses, and when the baseline drift exceeds a preset threshold, the phase-shift modulation parameters and the drug injection ratio are adjusted to obtain the corrected detection data, including:

[0044] Classifying the detection channels according to the multidimensional feature data of the target object, setting different monitoring parameters according to the heavy metal detection channel, the organic matter detection channel and the protein detection channel, and obtaining a classified monitoring plan;

[0045] Based on the classification monitoring scheme, phase-shifted coded pulse sequences of different periods are generated, and various detection channels are monitored in an interlaced manner within the preset time window to obtain channel response timing data;

[0046] Performing baseline extraction on the channel response time series data, calculating the baseline drift and noise standard deviation at short time scales and long time scales respectively, and obtaining channel stability data;

[0047] The baseline drift of each detection channel is compared with a preset threshold value according to the channel stability data, and the channels with excessive baseline drift are divided into an immediate correction group and a periodic correction group to obtain a graded correction scheme;

[0048] Optimizing the phase shift modulation parameters of the detection channel of the instant correction group, determining the phase compensation amount and the duty cycle parameters according to the difference in the target category, and obtaining the modulation correction data;

[0049] The medicine injection ratio of the instant correction group is reconfigured according to the modulation correction data, the injection sequence is set in combination with the characteristic reaction time of different targets, the detection channel is calibrated according to the graded correction scheme, and the corrected detection data is obtained.

[0050] Optionally, the self-learning evaluation is performed on the corrected detection data, and the self-learning evaluation includes updating the recognition model parameters using a progressive stochastic gradient algorithm and a multivariate regression analysis, and dynamically correcting the channel coupling strength and the feature weight based on the updated recognition model parameters to generate a multi-parameter fingerprint spectrum characterizing the characteristics of the sample to be tested, including:

[0051] Establishing a target feature set for the corrected detection data, respectively setting a heavy metal ion response feature group, a volatile organic compound response feature group, and a protein marker response feature group to obtain group training data;

[0052] The grouped training data are subjected to feature difference analysis using a progressive stochastic gradient algorithm, and the identification parameter deviations of each target feature group in different concentration ranges are calculated to obtain feature deviation data;

[0053] Performing multivariate regression analysis on the characteristic deviation data, calculating the identification model coefficients and fitting errors of each response characteristic group, and obtaining updated identification model parameters;

[0054] Calculating the coupling strength correction coefficient of each detection channel according to the updated recognition model parameters, distinguishing and assigning weights of the high coupling channel group and the low coupling channel group, and obtaining coupling weight data;

[0055] Calculating the response intensity weights of different target feature groups based on the coupling weight data, dynamically weighting the crosstalk features and complementary features, and obtaining feature weight data;

[0056] The coupling weight data and the characteristic weight data are combined in multiple dimensions, a response fingerprint database is constructed according to the characteristic distribution of different targets, and a multi-parameter fingerprint spectrum characterizing the characteristics of the sample to be tested is generated.

[0057] A second aspect of the present invention provides a multi-parameter coupling detection device of an integrated optical sensor, the multi-parameter coupling detection device of an integrated optical sensor comprising:

[0058] A drug injection module is used to inject different trace drug combinations into the sample to be tested in a plurality of microfluidic channels according to a preset time sequence, and obtain differentiated optical response signals of each channel through the selective reaction of the trace drug combination with the sample to be tested;

[0059] A spectrum acquisition module, used for performing phase shift coding acquisition on the differentiated optical response signal, splitting the broadband light source into a plurality of wavelength sub-intervals, generating a coded light signal with a preset phase shift amount for each wavelength sub-interval, inputting the coded light signal into a corresponding detection channel according to a timing allocation scheme, and obtaining multi-channel spectrum evolution data;

[0060] A signal processing module, used to calculate the cross-correlation function between channels according to the multi-channel spectral evolution data, group the detection channels based on the coupling characteristics of the cross-correlation function, perform signal amplification processing on the channel group with high coupling degree and positive correlation, and perform signal suppression processing on the channel group with high coupling degree and negative correlation, so as to obtain optimized multi-channel response data;

[0061] A feature extraction module is used to extract multi-dimensional features from the optimized multi-channel response data, calculate the differential interference, amplitude ratio and phase synchronization between channels, and construct a high-dimensional feature vector in combination with the characteristic reaction time of the micro-drug combination, and obtain multi-dimensional feature data of the target object by analyzing the interaction mode of the high-dimensional feature vector;

[0062] A correction control module is used to monitor the baseline drift and noise distribution of each channel using phase-shifted coded pulses according to the multi-dimensional feature data of the target object within a preset time window of the detection process, and when the baseline drift exceeds a preset threshold, adjust the phase shift modulation parameters and the drug injection ratio to obtain corrected detection data;

[0063] A self-learning evaluation module is used to perform self-learning evaluation on the corrected detection data. The self-learning evaluation includes updating the recognition model parameters using a progressive stochastic gradient algorithm and multivariate regression analysis, dynamically correcting the channel coupling strength and feature weight based on the updated recognition model parameters, and generating a multi-parameter fingerprint spectrum that characterizes the characteristics of the sample to be tested.

[0064] A third aspect of the present invention provides a multi-parameter coupling detection device for an integrated optical sensor, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected via lines; the at least one processor calls the instructions in the memory so that the multi-parameter coupling detection device for the integrated optical sensor executes the steps of the multi-parameter coupling detection method for the integrated optical sensor.

[0065] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of the multi-parameter coupling detection method of the integrated optical sensor.

[0066] This scheme first injects different combinations of trace reagents into multiple microchannels at different times, so that the sample to be tested produces an optical response with differentiated characteristics in each channel. The reason why higher accuracy and resolution can be achieved is that each channel no longer passively accepts the same reaction conditions, but instead realizes directional labeling or complexation of the target according to the pre-set timing and reagent ratio. In this way, even if different channels encounter similar interference factors, different response patterns will be reflected in the spectral signal due to the difference in reagents, which will help the subsequent process to more effectively separate and identify these responses. If it only relies on uniform injection or indifferent treatment, it is often impossible to significantly distinguish the subtle differences between multiple channels, and it is difficult to deal with the interactive effects of multiple components in complex samples; this scheme, through the pretreatment of differentiated reagents, is equivalent to establishing a separate "reaction benchmark" for each channel, so that multi-parameter detection has inherent complementarity and scalability.

[0067] After obtaining the preliminary multi-channel differentiated response, the scheme continues to use phase shift encoding to inject multiple bands of broadband light sources into each detection channel in turn. At this time, the optical changes of each channel in different wavelength sub-intervals and different phase shift offsets are recorded as multi-dimensional spectral evolution data. Based on the principle of phase modulation and staggered timing of light, the unintentional crosstalk of each channel can be avoided or weakened in the same hardware structure, allowing the system to clearly capture the real coupling relationship between channels. The subsequent cross-correlation function operation can determine the positive correlation (cooperative detection of amplifiable signals) or negative correlation (interference to be suppressed) according to the similarity of each pair of channels in time or phase, and amplify or attenuate the grouped channel signals, thereby realizing interactive management between channels. Since the cross-correlation characteristics under different bands and different phase shift conditions can more accurately characterize the degree of coupling, this scheme can not only suppress common crosstalk interference, but also co-amplify positively correlated channels at the right time, further improving the detection sensitivity of trace components. Afterwards, the multi-dimensional feature extraction step constructs a high-dimensional feature vector for each channel with the help of parameters such as interference and amplitude ratio, and then combines factors such as drug reaction time to complete the detailed identification of the target object; if the system detects that the baseline drift exceeds the established threshold, the channel response is corrected by re-adjusting the phase shift modulation and drug ratio to ensure stability during long-term or multiple tests. Finally, the recognition model is progressively updated using a self-learning evaluation method to achieve high adaptability to complex environments and multiple types of samples. Through this set of step-by-step, interlocking processes, this solution, with methods as the main axis, significantly improves the flexibility, reconfigurability and accuracy of multi-parameter detection without changing the main structure of the optical device, and successfully resolves the difficulties of existing detection modes in channel coupling management and dynamic correction. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0069] Figure 1 A schematic diagram of an embodiment of a multi-parameter coupling detection method of an integrated optical sensor in an embodiment of the present invention;

[0070] Figure 2 A schematic diagram of an embodiment of a multi-parameter coupling detection device with an integrated optical sensor in an embodiment of the present invention;

[0071] Figure 3 Schematic diagram of an embodiment of a multi-parameter coupling detection device with an integrated optical sensor in an embodiment of the present invention.

[0072] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0073] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0074] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back...), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0075] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, "and / or" in the full text includes three solutions. Taking A and / or B as an example, it includes technical solution A, technical solution B, and technical solution that satisfies both A and B. In addition, the technical solutions between the various embodiments can be combined with each other, which must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0076] An embodiment of the present application provides a multi-parameter coupling detection method of an integrated optical sensor. Figure 1 A flow chart of a multi-parameter coupling detection method of an integrated optical sensor provided in an embodiment of the present application. In this embodiment, the method includes:

[0077] See also Figure 1 , injecting different trace drug combinations into the sample to be tested in a plurality of microfluidic channels according to a preset time sequence, and obtaining differentiated optical response signals of each channel through the selective reaction of the trace drug combination with the sample to be tested;

[0078] In one embodiment of the present invention, the method of injecting different trace drug combinations into the sample to be tested in a plurality of microfluidic channels according to a preset time sequence, and obtaining differentiated optical response signals of each channel through the selective reaction of the trace drug combination with the sample to be tested, comprises: obtaining background component data of the sample to be tested, classifying and marking the sample to be tested according to the background component data, and determining a reference ratio parameter of the sample to be tested; calculating the drug injection amount of each microfluidic channel according to the reference ratio parameter, setting a corresponding trace drug combination for each microfluidic channel, and determining an injection sequence of the trace drug combination; performing flow diversion processing on the sample to be tested, and injecting the trace drug combination into the plurality of microfluidic channels according to the injection sequence. The corresponding trace drug combination is injected into the channel respectively, and the injection time of the trace drug combination is recorded; the selective reaction time of the sample to be tested and the trace drug combination is calculated, and the data acquisition time window of each channel is set according to the injection time and the selective reaction time; within the data acquisition time window, the reaction process of the sample to be tested is tracked according to the preset timing, and the selective reaction data of the sample to be tested and the trace drug combination at different times are recorded; the light intensity change trend of each channel is calculated according to the selective reaction data, and the light intensity change trend is normalized in combination with the reference ratio parameter of the sample to be tested to obtain the differentiated optical response signal of each channel.

[0079] Specifically, when obtaining the background component data of the sample to be tested, it is necessary to first use common laboratory instruments such as spectrometers, ion chromatographs or inductively coupled plasma emission spectrometers to preliminarily identify the main chemical components and trace components in the sample, and form a background component list in conjunction with the data processing algorithm. After the analysis is completed, the detected components are compared with the existing database to determine the distribution and approximate concentration range of the required detection indicators. The background component list will be corrected according to the peak position, ion intensity or spectral peak area output by the analytical instrument to form one or more component groups. Subsequently, these groups can be classified and labeled with the help of automatic classification algorithms (such as K-means clustering or multidimensional classification methods based on hierarchical clustering), thereby generating a set of benchmark ratio parameters for each type of sample.

[0080] The benchmark ratio parameter refers to the concentration ratio and order of the reagents distributed in multiple microfluidic channels. A stronger reagent combination will be set for high-concentration heavy metals or volatile compounds, and an appropriate amount of probes or color developers will be set for substances with relatively small background interference. In the example of wastewater multi-metal monitoring, the background composition data will show that the concentrations of certain metal ions in the water sample are at a high level, while the content of organic impurities and inorganic salts is secondary. During the classification and labeling process, the water sample will be assigned a reagent ratio that is more suitable for heavy metal detection, so that each channel can perform its duties in subsequent analysis. In this way, different channels obtain more targeted chemical reaction conditions, creating favorable conditions for subsequent improvements in accuracy and resolution.

[0081] After calculating the amount of reagent injection for each microfluidic channel, it is necessary to differentiate and combine these reagent injection amounts according to the established reference ratio parameters to form an injection sequence. The injection sequence consists of the channel number, injection time and reagent component, and a precision pump or valve switching device is used to quantitatively inject the corresponding reagent into the sample to be tested at the appropriate time. The diversion processing link is responsible for dividing the original sample into multiple branches in the pipeline, so that it enters different channels respectively, contacts with their respective reagent combinations and reacts selectively. To ensure that subsequent data can be accurately attributed to the corresponding channel, the system will register the trigger time of each injection operation and record the target channel number. Subsequently, by studying the reaction rate of the sample and the reagent in different channels, the selective reaction time can be obtained, which represents the time from the start of mixing to the generation of a stable signal that can be detected by optical sensing. According to the injection time and the selective reaction time, the data acquisition window of each channel is set to a critical period that can capture an obvious reaction process. Assuming that a channel is dedicated to detecting a specific heavy metal ion, and it takes about fifteen seconds for the heavy metal to significantly change the fluorescence intensity after mixing with a reagent containing a fluorescent probe, the channel will be intensively sampled within fifteen to thirty seconds after the reaction starts. Outside this period, the sampling frequency can be reduced to reduce the accumulation of invalid data. Through this process, each channel retains the optimal acquisition interval that matches its reagent combination.

[0082] Subsequently, the reaction between the sample to be tested and the agent is tracked in real time according to the preset injection sequence within the data acquisition time window, and the optical detector records the spectral intensity, refractive index change or fluorescence peak characteristics at different reaction stages. Combined with the injection time, it can be known which agent is added at what time, and combined with the selective reaction time, the peak value of its active reaction period can be grasped. After the recording is completed, the obtained light intensity change trend is normalized, the relative differences caused by the dosage of the agent and the background concentration of different channels are unified into the same reference standard, and all response curves are amplitude corrected and baseline corrected based on the previous benchmark ratio parameters. If the amplitude of the reaction signal of channel one is too high, it can maintain a similar dimension with other channels after normalization, which is convenient for comprehensive judgment of multiple indicators. The differentiated optical response formed in this way can not only reflect the difference in sensitivity of each channel to different target components, but also significantly reduce noise interference and the probability of misjudgment in the subsequent analysis link. The entire step-by-step implementation only requires a common precision infusion pump or valve control system at the instrument level, as well as an optical detector that can meet multi-channel parallel sampling, so it can be used in various sample detection platforms without large-scale changes to the main structure of the sensor.

[0083] In the implementation of this method, each sub-step is of great significance: through the preliminary background component data collection and classification labeling, the most appropriate combination of reagents can be selected for the sample; by calculating and executing the differentiated injection sequence of each channel, the accuracy of identifying multi-component mixed environments can be improved; through the setting of selective reaction time and appropriate data acquisition window planning, the most effective optical signal within a precise time period can be obtained; finally, normalizing the differentiated optical response signal can reduce system deviation and significantly improve the reliability and pertinence of the results.

[0084] Please continue reading Figure 1 , performing phase shift encoding acquisition on the differentiated optical response signal, splitting the broadband light source into a plurality of wavelength sub-intervals, generating a coded light signal with a preset phase shift amount for each of the wavelength sub-intervals, inputting the coded light signal into a corresponding detection channel according to a timing allocation scheme, and obtaining multi-channel spectral evolution data;

[0085] In one embodiment of the present invention, the phase-shift coded acquisition of the differentiated optical response signal, splitting the broadband light source into a plurality of wavelength sub-intervals, generating a coded light signal with a preset phase shift amount for each wavelength sub-interval, and inputting the coded light signal into a corresponding detection channel according to a timing allocation scheme to obtain multi-channel spectral evolution data, includes: performing spectral decomposition on the differentiated optical response signal, setting a sampling interval for the differentiated optical response signal according to the characteristic frequency of each channel, and generating a multi-channel sampling sequence; sub-dividing the broadband light source according to the frequency distribution of the multi-channel sampling sequence, splitting the broadband light source into a plurality of wavelength sub-intervals, and obtaining the spectral components of each wavelength sub-interval; The phase modulation calculation is performed on the spectral component, and a phase offset parameter is set for each wavelength sub-interval according to the timing characteristics of the multi-channel sampling sequence to generate an initial coded optical signal; the phase compensation amount of each wavelength sub-interval is calculated according to the amplitude distribution of the differentiated optical response signal, and the initial coded optical signal is subjected to phase compensation processing to obtain a coded optical signal with a preset phase shift amount; based on the time distribution characteristics of the multi-channel sampling sequence, the coded optical signal with the preset phase shift amount is time-series divided to generate an interleaved timing allocation scheme; according to the interleaved timing allocation scheme, the coded optical signal with the preset phase shift amount is sequentially input into each detection channel, the spectral response process of each channel is recorded, and multi-channel spectral evolution data is obtained.

[0086] Specifically, when performing spectrum decomposition on the differentiated optical response signal, it is necessary to first convert the original optical signal obtained in each detection channel into a corresponding spectrum distribution diagram according to a fixed sampling method or discrete Fourier transform algorithm. This is done to extract the energy distribution at different frequencies from the time domain signal and lock the target frequency band according to the expected signal characteristics or interference characteristics. By analyzing this part of the results, the main resonance peak or characteristic frequency of each channel can be determined, so that the characteristic frequency and the frequency band near it can be sampled with higher precision in the subsequent steps. In order to facilitate the simultaneous management of the spectrum information of multiple channels, the main characteristic frequencies of each channel will be sorted into a list, and the sampling interval of the differentiated optical response signal will be set based on this list. If it is found in the detection application that channel one has a significant resonance peak in the frequency band of several hundred hertz, and the main power of channel two and channel three is concentrated in a higher or lower frequency band, a narrowband sampling interval will be set near the characteristic frequency of each channel, thereby generating a more refined multi-channel sampling sequence. The multi-channel sampling sequence can be understood as a sampling allocation plan for all channels in their respective key frequency bands to ensure that subsequent encoding and modulation can be carried out based on the information of these key frequency bands, avoiding excessive sampling resources in irrelevant or extremely noisy frequency bands.

[0087] According to the frequency distribution of the multi-channel sampling sequence, when the broadband light source is divided into sub-intervals, it is necessary to first split the originally continuous or quasi-continuous broadband light source into several wavelength sub-intervals, each of which corresponds to a relatively narrow spectral range, which corresponds to the characteristic frequency bands identified in the previous step. In this division mode, if it is identified that some channels prefer to receive wavelengths distributed in the visible light region, while other channels present a higher signal-to-noise ratio in the near-infrared region, the broadband light source will be cut into several sub-bands so that each sub-band can accurately illuminate the high-sensitivity band of the target channel. In this way, the output of the light source can be maximized and the energy waste in irrelevant bands can be reduced, thereby obtaining clearer spectral components. Each sub-interval has a corresponding central wavelength and bandwidth, which coincides with the frequency range of the multi-channel sampling sequence. Subsequently, by managing these spectral components, the system can cooperate with specific phase shift strategies in subsequent steps to ensure that the excitation and acquisition of each channel are more targeted.

[0088] When calculating the phase modulation of the spectral components, it is necessary to first establish phase offset parameters that match the timing characteristics of each channel, and then apply these parameters to the spectral components of the corresponding wavelength sub-intervals, so that each sub-interval has a phase offset corresponding to the characteristic frequency or characteristic time window of the target channel. This process is completed by a digital modulation controller or a programmable optical modulator. The initial coded optical signal formed in this way has different phase distributions, which can help the system distinguish the possible impact of each sub-interval on different channels when multiple channels work in parallel. Assuming that channel one is most sensitive to light with a phase offset of zero degrees, channel two has a significant response to the phase offset area of ​​90 degrees, and channel three obtains the best response near a higher phase shift angle, by setting the phase offset parameters for each wavelength sub-interval, each channel can obtain a relatively independent excitation signal in the same broadband light source.

[0089] Subsequently, according to the amplitude distribution of the differentiated optical response signal, the phase compensation amount of each wavelength sub-interval is calculated, and the initial coded optical signal is subjected to phase compensation processing to obtain a coded optical signal with a preset phase shift amount. The role of the compensation amount is to correct the amplitude differences caused by the channel itself or the external environment, so that each wavelength sub-interval can maintain the desired relatively balanced or controllable energy level before entering the channel. For example, when detecting a high-sensitivity channel, if the amplitude deviation is too large, it will cause resonance saturation or invalid response. Phase compensation can be used to equalize the unbalanced power distribution. In addition, when the original phase characteristics of the light source cannot fully meet the requirements of each channel for response delay or sensitivity, additional compensation can be used for fine-tuning so that the coded optical signal that finally enters the channel is closer to the target modulation curve.

[0090] Next, based on the time distribution characteristics of the multi-channel sampling sequence, the coded optical signal with a preset phase shift is time-sequenced to generate an interleaved timing allocation scheme. The so-called "interleaved" means setting non-overlapping or partially overlapping windowing moments for each wavelength sub-interval in the time dimension, so that each channel can obtain the main excitation light source in a specific time slice, while other channels enter a low-power or standby state to reduce crosstalk interference. Through this time division, different wavelength sub-intervals can be projected to different channels with different phase shift offsets at the same time, and the time division mechanism is used to ensure that the channels do not interfere with each other. If the system detects that channel one and channel two may cross in the spectrum, then the timing of the coded optical signals of the two is staggered to effectively distinguish the responses of the two. Assume that channel one receives a sub-interval light with a phase of zero degrees and a central wavelength of 650 nanometers in the first five seconds, and channel two receives a sub-interval light with a phase of sixty degrees and a central wavelength of 700 nanometers in the subsequent five seconds. This can fully utilize the same light source and avoid unnecessary signal overlap in the same time period.

[0091] Finally, according to the interleaved timing allocation scheme, the coded optical signal with a preset phase shift is input into each detection channel in order, and the spectral response process of each channel is recorded by the spectral detection equipment to form multi-channel spectral evolution data. These data show the optical change law of different channels in different sub-intervals, different phase shifts and different time windows, and include all-round measurement of amplitude, phase and interference peak position. In some environmental monitoring scenarios, it can be observed that the special absorption peak of channel one changes with phase shift, while channel two presents unique luminescence characteristics with phase shift modulation. The fusion of the spectral evolution data of the two can significantly enhance the recognition of complex pollutants. Due to the precise management of spectrum decomposition, phase shift and timing allocation throughout the process, the interference of optical signals obtained by each channel during recording is greatly reduced, and the sensitivity is significantly improved. Such an interleaved modulation method not only requires less hardware changes, but also has a clear implementation principle. By digitally modulating and subdividing the optical signal, it can flexibly adapt to the needs of multi-channel parallel work, so as to achieve good detection results under various complex sample conditions.

[0092] This embodiment can avoid different channels competing for resources in the same time slice and the same wavelength sub-interval, while ensuring that the phase shift parameters of each channel can be accurately applied to each sub-interval. Compared with traditional single-band continuous light source excitation, this method can cleverly distinguish the responses of multiple channels based on refined frequency division and phase shift modulation, reducing performance losses caused by insufficient contrast or spectral aliasing. The multi-channel spectral evolution data obtained in this way is also easier to dig out the true and differentiated characteristics of each channel in subsequent data analysis and decoupling links, thereby providing a clear and reliable foundation for multi-parameter detection.

[0093] Please continue reading Figure 1 , calculating the cross-correlation function between channels according to the multi-channel spectral evolution data, grouping the detection channels based on the coupling characteristics of the cross-correlation function, performing signal amplification processing on the channel group with high coupling degree and positive correlation, and performing signal suppression processing on the channel group with high coupling degree and negative correlation, to obtain optimized multi-channel response data;

[0094] In one embodiment of the present invention, the cross-correlation function between channels is calculated according to the multi-channel spectral evolution data, the detection channels are grouped based on the coupling characteristics of the cross-correlation function, the channel group with high coupling and positive correlation is subjected to signal amplification processing, and the channel group with high coupling and negative correlation is subjected to signal suppression processing to obtain optimized multi-channel response data, including: performing time series analysis on the multi-channel spectral evolution data, extracting the spectral response value of each detection channel at the corresponding moment according to a preset sampling period, and obtaining the spectral response sequence of the detection channel; dividing the spectral response sequence into time windows, determining the sequence length according to a preset analysis time length, calculating the cross-correlation function between any two detection channels, and obtaining the coupling characteristic data between the channels, wherein the formula for calculating the cross-correlation function is as follows:

[0095] ;

[0096] in, For detection channel and detection channels The cross-correlation value of is the sequence length, and Detection channels and detection channels At sampling time The spectral response sequence, is the time delay, and ,like If it is greater than 0, it means the channel The response lags behind or precedes the channel , it is necessary to cross the corresponding number of sampling points during calculation.

[0097] Threshold analysis is performed on the coupling characteristic data between the channels to calculate the cross-correlation function. =0, and divide the detection channel into a high-coupling channel pair and a low-coupling channel pair according to a preset coupling threshold to obtain a channel grouping result; classify the high-coupling channel pair by polarity according to the channel grouping result, divide the channel pair with a cross-correlation function peak value greater than zero into a positively correlated channel group, and divide the channel pair with a cross-correlation function peak value less than zero into a negatively correlated channel group to obtain channel group polarity data; synchronously amplify the spectral response sequence of the positively correlated channel group, calculate the gain coefficient of each channel pair according to the positive peak value of the cross-correlation function, and perform amplitude modulation on the spectral response value to obtain positively correlated signal modulation data; suppress the spectral response sequence of the negatively correlated channel group, calculate the suppression coefficient of each channel pair according to the negative peak value of the cross-correlation function, selectively attenuate the crosstalk component, and obtain negatively correlated signal suppression data; integrate the positively correlated signal modulation data with the negatively correlated signal suppression data to obtain optimized multi-channel response data.

[0098] Specifically, when performing time series analysis on the multi-channel spectral evolution data, the spectral response value of each detection channel at each time point is first extracted from the record according to a fixed sampling period, and these discrete spectral response values ​​are concatenated into a spectral response sequence for analysis. This process usually relies on a digital sampling device or software-based data decoding, which can separate the timing information matching each channel from the raw data of multi-channel synchronous recording. In this way, the optical behavior of each channel within a preset sampling time can be refined into several equidistant time slices, and a column of numerical sequences with a fixed order is obtained. Assuming that channel one, which is used to monitor specific pollutant molecules in a certain scene, collects spectral responses at a frequency of once per second during the entire detection process, the time series analysis will arrange the spectral values ​​of channel one in the first second, second second, third second... in sequence to form a response sequence; similar processing will be performed on channels two to n, thereby obtaining a multi-channel synchronous sampling sequence.

[0099] After obtaining these spectral response sequences, in order to analyze the mutual influence between different channels, it is necessary to divide the time and set the sequence length. The sequence length here can be determined by the pre-set analysis time or detection window. The goal of time division is to truncate the overly long data sequence into a suitable interval in order to strike a balance between computational complexity and practicality. If in the example scenario it is planned to only analyze the first two minutes when the sample is most active after the addition of the agent, the timing analysis will focus on the spectral response sequence within these two minutes, and use this time period as the input for the subsequent calculation of the cross-correlation function. The core of the calculation of the cross-correlation function is to measure the similarity of different channels under different time delay conditions, according to the formula:

[0100] ;

[0101] This formula can be used to determine whether the response curves of the two channels tend to be synchronous or show mirror-like changes under a given time offset. When the delay interval is large, the peak position and size of the cross-correlation can be found in the entire delay range to determine whether there is significant coupling between channels.

[0102] After obtaining the cross-correlation function curve, it is necessary to group the channels based on a predetermined coupling threshold. This threshold is usually pre-set by the experiment, which means =0 or near it, if the cross-correlation peak exceeds this value, it can be determined that the two channels have a strong coupling relationship; otherwise, it means that the correlation between the two is weak. When the peak is detected to be much larger than the positive threshold, it can be determined that the corresponding channel pair has a high degree of coupling. If the peak falls below the threshold, it is classified as a low degree of coupling. This will form a channel grouping result: high-coupling channel pairs are relatively separated from low-coupling channel pairs. In complex scenarios, this classification method can quickly locate which channels are strongly linked to each other at the signal level and which channels are relatively independent, which will help the subsequent steps to process the coupling relationship more finely.

[0103] Next, according to the channel grouping results obtained, it is necessary to perform polarity difference processing on the channel pairs with high coupling, that is, to determine whether the correlation is positive or negative based on the peak sign of the cross-correlation function. If the peak is in the positive direction, it means that the responses of the two channels change in the same direction in the same time period, and such channel pairs can be classified as positive correlation channel groups; on the contrary, if the peak is obviously negative, it means that the two channels mostly show an inverse relationship in the response peaks and valleys, and such channel pairs should be classified as negative correlation channel groups. Polarity difference processing allows the system to quickly distinguish whether it is a "combined amplification" relationship or a "mutual cancellation" relationship, thereby providing a decision-making basis for the next step of signal amplification or suppression. For example, if it is found that both channel one and channel two produce synchronous enhancement peaks for a certain pollutant during water quality testing, the positive peak of the cross-correlation function will be significantly higher than zero, and these channels can be marked as positively correlated; if channel three shows opposite periodic signals for the same substance, that is, when channel one responds to an increase, channel three is suppressed or reduced, then the cross-correlation peak will fall in the negative value range, and they can be classified as negative correlation channel groups.

[0104] For channel groups that have been identified as positively correlated, local amplification processing is required to make better use of their synergistic gain effect. This operation can determine the gain coefficient of each channel by calculating the size of the positive peak, and then apply the corresponding "range-extended modulation" to the respective spectral response values. In this process, if the positive peaks of channel one and channel two are particularly high, the gain coefficient will also be relatively large, and the system will multiply the original spectral values ​​of the corresponding channels to highlight the signal advantages brought by the superposition of the two. This is equivalent to focusing on strengthening the channel pairs that can produce complementarity or synergy during the analysis stage, so as to obtain clearer indicator signals in the subsequent identification of pollutants or target molecules.

[0105] For the detection channels that are classified into the negative correlation channel group, signal suppression processing is required to control the crosstalk components that are useless or offset each other. For example, if the peak value of the cross-correlation function of two channels falls in the negative interval and reaches a certain amplitude, it means that the two channels produce obvious reverse disturbances in certain detection dimensions, which is easy to cause confusion in the overall analysis. At this time, the suppression coefficient will be calculated according to the size of the negative peak, and the crosstalk part will be selectively attenuated. Attenuation can be achieved by directly reducing or deducting the interference terms at the data level, or by setting penalty weights in subsequent algorithms so that these obvious offsetting signals do not dominate the results. For example, if channel one has a peak for a certain complex organic matter, while channel two always maintains a low value in the same time period, and the negative correlation peaks of the two are very significant, there is reason to regard this pair of channels as negative interference in this frequency band, and reduce its impact on the overall results through suppression processing.

[0106] After completing the amplification of the positive correlation group and the suppression of the negative correlation group, it is necessary to integrate the positive mutual correlation signal modulation data and the mutual correlation signal suppression data to obtain the final optimized multi-channel response data. This process means that the waveform that has been "amplified" and the waveform that has been "suppressed" are re-aggregated into a unified coordinate system or the same data structure, so that the subsequent multi-channel fusion analysis can more easily identify the target or make accurate judgments on the premise of amplifying the signal components that are beneficial to the detection and weakening the adverse interference. Through this method, abnormal judgments caused by random noise, negative correlation crosstalk, etc. can be significantly reduced, and the effective signals of multiple collaborative channels can be integrated into a more robust and explanatory response curve. In a complex field environment, the weak signals output by some channels may be masked or difficult to distinguish if observed independently, but when they are amplified in the positive correlation group, they can work together with other channels to indicate the presence of a specific pollutant or a higher concentration distribution, thereby improving the sensitivity and accuracy of the entire multi-channel detection system.

[0107] Please continue reading Figure 1, extracting multi-dimensional features from the optimized multi-channel response data, calculating the differential interference, amplitude ratio and phase synchronization between channels, constructing a high-dimensional feature vector in combination with the characteristic reaction time of the trace drug combination, and obtaining multi-dimensional feature data of the target object by analyzing the interaction mode of the high-dimensional feature vector;

[0108] In one embodiment of the present invention, multi-dimensional feature extraction is performed on the optimized multi-channel response data, the differential interference, amplitude ratio and phase synchronization between channels are calculated, and a high-dimensional feature vector is constructed in combination with the characteristic reaction time of the trace drug combination. The multi-dimensional feature data of the target object is obtained by analyzing the interaction mode of the high-dimensional feature vector, including: characteristic frequency band division of the optimized multi-channel response data, segmenting the spectral data of each channel according to the characteristic absorption interval of heavy metal ions, organic matter and protein, and obtaining multi-interval spectral distribution data; the differential interference between any two detection channels is calculated according to the multi-interval spectral distribution data, and the amplitude difference and phase difference between the channels are analyzed by the coherent spectrum analysis method. , obtain interference feature data between channels; perform normalization processing on the interference feature data between channels, calculate the amplitude ratio of adjacent detection channels, extract the phase difference as a phase synchronization parameter, and obtain channel association feature data; perform dynamic analysis on the reaction process of the trace drug combination and the target, calculate the characteristic reaction time according to the inflection point moment of the spectral response of each channel, and obtain drug reaction feature data; combine the interference feature data between channels, the channel association feature data and the drug reaction feature data to construct a high-dimensional feature vector, and obtain feature space distribution data; perform cluster analysis on the interaction pattern in the feature space distribution data, extract feature parameters of each dimension according to the feature distribution law of different targets, and obtain multi-dimensional feature data of the target.

[0109] Specifically, when the optimized multi-channel response data is divided into characteristic frequency bands, the detected spectral range will be divided into multiple sections according to the pre-established spectral absorption characteristics of the target objects such as heavy metal ions, organic matter and proteins. The specific method is: after reading the optimized spectral response data of each detection channel, the significant absorption peaks of different substances in specific wavelengths or frequency bands are locked according to the spectral fingerprint information in the database, and then these sections are respectively classified as heavy metal absorption areas, organic absorption areas or protein absorption areas. When it is detected that some bands may have overlapping effects, the discrimination degree of the band for the target object will be evaluated. If the discrimination degree is high, it will be retained as an independent interval, otherwise it may be merged with the adjacent interval. Through this band division method, each detection channel forms a spectral distribution data consisting of multiple intervals, and thereby presents the differentiated response of the channel to different types of substances. For example, in a water quality monitoring instance, if the system recognizes that the wavelength range of 200-300 nanometers is closely related to the ultraviolet absorption of heavy metals, the range of 300-400 nanometers coincides with the characteristic peaks of organic matter, and certain peak positions between 280-320 nanometers may indicate the possible presence of proteins, then these ranges will be separately marked and included in the subsequent analysis process to ensure that the response segments to different targets can be fully processed.

[0110] After obtaining this multi-interval spectral distribution data, it is necessary to further calculate the differential interference between any two detection channels, and use the coherent spectrum analysis method to analyze the amplitude difference and phase difference between the channels to obtain the interference characteristic data between the channels. The differential interference refers to the interference result formed by subtracting or superimposing the response values ​​of two channels to the same target in the same spectral interval. Through this link, it can be observed whether channel one has an amplitude amplification effect with channel two in the heavy metal interval, or whether channel three produces a strong phase cancellation with channel four on the organic absorption peak. The coherent spectrum analysis method generally uses Fourier transform or phase correlation function to superimpose the signals of the two channels in the frequency domain or phase domain, and then judges the degree of matching between the two in spectral intensity and phase change. When the amplitude difference is significant and the phase is similar, it often indicates that the two channels have similar sensitivity to the same target; when the amplitude is small but the phase is opposite, it may indicate that one of the two channels is insensitive to the target or has a large interference. Through this complementary analysis, the performance of the test substance in different bands can be cross-validated with the help of data from multiple channels, providing a richer basis for subsequent accurate identification.

[0111] After obtaining the interference characteristic data between channels, it is necessary to normalize them. By calculating the amplitude ratio of adjacent detection channels and extracting the phase difference to evaluate the phase synchronization, the channel-related characteristic data can be obtained. The purpose of normalization is to map the interference quantity with different numerical scales to a more comparable interval for comparison and aggregation in the same dimension. For example, if the amplitude of channel one is generally higher than that of channel two, this difference may be due to the different detector gains. If it is not processed in advance, it will cause systematic deviations in phase or amplitude evaluation. By uniformly calibrating the amplitude of the interference quantity in each interval and defining the phase synchronization using the phase difference of adjacent channels, it can be determined whether the two have similar peak periods or absorption transitions in the same detection window. If the phase synchronization is high, these channels often produce a coordinated response to the same target. If the phase difference is large, it may indicate that they have different detection mechanisms for the target. This method is very effective in distinguishing mixed signals brought by multiple targets, because when multiple targets absorb in the same band, the detection principles or surface functional modifications of different channels often lead to obvious differences in phase and amplitude.

[0112] After completing the normalization of the interference characteristics between channels, it is necessary to dynamically analyze the reaction process of the trace reagent combination and the target, calculate the characteristic reaction time based on the inflection point moment of the spectral response of each channel, and form the reagent reaction characteristic data. The inflection point moment refers to the key node where the spectral response curve gradually transitions from the rapid rise or fall stage to the stable period (or suddenly jumps from the stable period), so as to judge the start and completion time of the actual chemical reaction. By automatically detecting these inflection points, it is possible to know when a stable signal is obtained and when the reaction ends. For example, if a channel shows an obvious jump in the heavy metal absorption section at the 10th second and completely flattens at the 15th second, it can be considered that the period between 10 seconds and 15 seconds is the stage where the reagent and heavy metal ions react critically. During this period, the sampling frequency should be increased to capture more complete curve changes, and finally the average reaction time of this stage is assigned to the corresponding reagent characteristic data. In this way, each channel can combine its own spectral response curve to obtain a set of reaction cycle data that is closer to reality, providing a timing reference for overall judgment.

[0113] Subsequently, the interference feature data between channels, channel association feature data, and drug reaction feature data are combined to construct a high-dimensional feature vector to form feature space distribution data. The core here is to pack key indicators such as amplitude ratio, phase synchronization, and characteristic reaction time into a multidimensional vector, so that the system can jointly model the responses of multiple channels in the subsequent analysis process. This high-dimensional vector may contain seven or eight or even more than a dozen dimensions, each of which represents an important parameter, such as the amplitude ratio of a specific band, or the phase synchronization of a specific time window. When multiple channels are involved in multi-parameter detection at the same time, this multi-dimensional characterization method can maximize the complementary advantages of data generated by different detection principles, different wavelength ranges, and different time periods. In some complex scenarios (such as industrial wastewater containing a variety of heavy metals and organic matter), it is difficult to distinguish the source of pollutants using only a single feature. With the help of high-dimensional feature vectors, the "fingerprint patterns" of different substances can be more accurately divided.

[0114] After completing the construction of the high-dimensional feature vector, it is necessary to perform cluster analysis on the interaction patterns in the feature space distribution data, and extract the feature parameters of each dimension based on the feature distribution law of different targets in the space, and finally obtain the multi-dimensional feature data of the target. Cluster analysis can select common algorithms, such as K-means or hierarchical clustering, so that the data points in the feature space are divided into different clusters according to the degree of similarity. Each cluster represents a type of substance that is relatively concentrated in high-dimensional features. When a large cluster is found to have similar characteristics in the heavy metal absorption area, phase synchronization, and reaction time, it can be determined that it corresponds to a specific type of heavy metal ion; if another type of cluster shows a strong response in the organic absorption area and a short reaction time, it can be further summarized as a certain volatile organic compound. In this process, the characteristic parameters of each dimension are equivalent to the coordinate axis of the fingerprint information. When the cluster analysis is completed, each type of substance will occupy a specific distribution interval on these coordinate axes, thereby providing strong support for subsequent automatic recognition and classification. Such multidimensional clustering has significant advantages in high-throughput or multi-parameter scenarios, because it is no longer limited to a single indicator or a single path, but incorporates all detection information of different properties (such as interference intensity, phase synchronization, reaction time, etc.) into the same analysis framework, greatly improving recognition accuracy and robustness. In summary, by dividing the optimized multi-channel response data into characteristic frequency bands, calculating differential interference, analyzing phase differences, capturing drug reaction timing, and clustering high-dimensional feature spaces, it can effectively break through the limitations of traditional single-band or dual-band analysis and achieve accurate perception and differentiation of multiple targets such as heavy metals, organic matter, and proteins.

[0115] Please continue reading Figure 1, within a preset time window of the detection process, based on the multi-dimensional feature data of the target object, the baseline drift and noise distribution of each channel are monitored using phase-shift coded pulses, and when the baseline drift exceeds a preset threshold, the phase-shift modulation parameters and the drug injection ratio are adjusted to obtain corrected detection data;

[0116] In one embodiment of the present invention, within a preset time window of the detection process, according to the multidimensional feature data of the target object, the baseline drift and noise distribution of each channel are monitored using phase-shift coded pulses. When the baseline drift exceeds a preset threshold, the phase-shift modulation parameters and the reagent injection ratio are adjusted to obtain corrected detection data, including: classifying the detection channels according to the multidimensional feature data of the target object, setting different monitoring parameters according to the heavy metal detection channel, the organic detection channel and the protein detection channel, and obtaining a classified monitoring scheme; generating phase-shift coded pulse sequences of different periods based on the classified monitoring scheme, and performing interleaved monitoring on various detection channels within the preset time window to obtain channel response timing data; and Baseline extraction is performed based on the data, and the baseline drift and noise standard deviation in short time scales and long time scales are calculated respectively to obtain channel stability data; the baseline drift of each detection channel is compared with a preset threshold value according to the channel stability data, and the channels with excessive baseline drift are divided into an immediate correction group and a periodic correction group to obtain a graded correction scheme; the phase shift modulation parameters of the detection channel of the immediate correction group are optimized, and the phase compensation amount and duty cycle parameters are determined according to the differences in target categories to obtain modulation correction data; the drug injection ratio of the immediate correction group is reconfigured according to the modulation correction data, and the injection sequence is set in combination with the characteristic reaction time of different targets, and the detection channel is calibrated according to the graded correction scheme to obtain corrected detection data.

[0117] Specifically, when classifying the detection channels according to the multidimensional characteristic data of the target, it is necessary to first extract the response intensity, phase synchronization and characteristic reaction time between each detection channel and the specific target from the analysis results obtained in the previous stage, and judge whether each channel has higher sensitivity or stronger selectivity for different categories of substances such as heavy metals, organic matter or proteins based on this information. In order to make the classification process more accurate, a matching database or classification rule set can be established to distinguish the three major categories of heavy metal detection channels, organic matter detection channels and protein detection channels. If a channel has the highest degree of consistency with heavy metal targets in multidimensional features (such as phase synchronization, amplitude gain), it will be marked as a heavy metal detection channel. If it is most sensitive to the absorption band and characteristic timing of organic matter, it will be classified into the organic matter detection channel. If it mainly matches the spectral peak and interference pattern of protein, it will be classified into the protein detection channel. In this way, each detection channel can better focus on the detection of the most suitable substance in subsequent use, reducing the scattered response to other types of substances. The classification process will eventually output a classification monitoring plan, which clarifies the correspondence between channels and targets, and also provides monitoring parameters suitable for heavy metal detection (such as wavelength sub-ranges that require higher sensitivity or longer sampling time sequences), monitoring parameters suitable for organic matter detection (such as the need to take into account multiple wavelength sub-ranges and medium sampling frequencies), and monitoring parameters suitable for protein detection (such as special requirements for the sensitivity of surface modification layers and characteristic absorption peaks).

[0118] When generating phase-shift coded pulse sequences of different periods based on the classification monitoring scheme, it is necessary to comprehensively consider the target characteristics and measurement needs of each type of detection channel. For example, if the heavy metal detection channel is most sensitive to a specific ultraviolet band, the corresponding sub-intervals and phase shifts will be planned in the pulse sequence to ensure that high-intensity or more frequent excitation light is provided in the critical time window. At the same time, for the protein detection channel, if it has a long reaction time and strict requirements on the signal-to-noise ratio, a larger duty cycle or a lower repetition frequency can be set in the pulse sequence, so that the detection channel has a more stable optical input to capture weak signals. Subsequently, these coded pulses with differentiated parameters perform interleaved monitoring of various detection channels within a preset time window. The system excites the heavy metal detection channel, the organic matter detection channel, and the protein detection channel in turn according to the arranged timing, thereby obtaining the channel response timing data. Such an interleaved method can not only take into account the parallel operation of multiple channels, but also greatly reduce the coupling interference between different channels. For example, if the sampling period of the heavy metal detection channel only partially overlaps with the organic matter detection channel, the possible crosstalk can be limited to a specific window, and it is easier to separate and correct the respective signal differences at the algorithm level.

[0119] When performing baseline extraction on the channel response time series data, it is necessary to first separate the output level of the channel under target-free or low concentration conditions from the long-term or short-term sampling, and this part of the output is used as the baseline reference value. Then, the baseline drift and noise standard deviation under short-time scale and long-time scale can be calculated respectively by statistical methods (such as sliding average or least squares fitting). If the short-time scale baseline drift is small, it indicates that the channel is relatively stable in a certain round of continuous detection, while the long-time scale drift is used to reflect the chronic offset caused by environmental factors such as multiple rounds of detection or temperature fluctuations. Channel stability data formed by these indicators can provide a basis for subsequent graded correction. When a protein detection channel shows obvious drift after thirty minutes of continuous detection, it can be determined that it may be affected by rising ambient temperature, aging of the agent, or attenuation of light source energy, so that the channel is divided into a specific level in the correction scheme, and correction is carried out in an immediate or periodic manner as appropriate.

[0120] When the baseline drift of each detection channel is compared with the preset threshold according to the channel stability data, all channels will be divided into a normal group with drift within the threshold and an abnormal group with drift exceeding the limit, and the abnormal group will be further divided into an immediate correction group and a periodic correction group. If the baseline drift exceeds the threshold by a small amount and only occurs in a short period of time, it can be classified as an immediate correction group, and stability can be restored through relatively fast local correction means; if the drift exceeds the threshold by a large amplitude or lasts for a long time, it will be classified into a periodic correction group, indicating that the channel has obvious environmental disturbances or device aging, and more thorough compensation is required after regular maintenance or longer-term maintenance. Through this hierarchical correction scheme, resources and time can be more efficiently invested in channels that really need emergency treatment, while ensuring that all channels can be accurately corrected at an appropriate frequency, and there will be no waste of resources caused by large-scale, repetitive interventions.

[0121] When optimizing the phase shift modulation parameters of the detection channel of the instant correction group, it is first necessary to detect the specific drift source currently faced by the channel, and determine the phase compensation amount and duty cycle parameters in combination with the target category to which it belongs. If a channel is a heavy metal detection channel, its sensitive range is mostly in the ultraviolet band and the light source is very sensitive to phase changes, the phase compensation amount will be appropriately increased and the duty cycle will be reduced for this band to quickly concentrate energy and shorten the invalid detection time, so that the channel can obtain sufficient detection signals in the shortest time. If another channel mainly detects organic matter and relies more on the visible light region, the pulse width can be extended in the phase shift strategy, and the phase offset can be adjusted slightly in a targeted manner, so that the channel has a larger receiving window, thereby overcoming temporary drift problems and ensuring the signal-to-noise ratio. The modulation correction data obtained in this way include parameters such as phase shift modulation intensity, phase offset angle, and duty cycle timing, which are used to accurately guide the subsequent injection ratio of the agent and the modulation method of the light source, so that the instant correction group can correct its own measurement benchmark in a short time.

[0122] Finally, when the reagent injection ratio of the instant correction group is reconfigured according to the modulation correction data, a new injection sequence needs to be set in combination with the characteristic reaction time of different targets, so that the channel currently affected by drift is in a more favorable chemical reaction condition and optical coupling environment in the subsequent acquisition. For example, if it is detected that the reaction time of the channel is prolonged due to temperature increase or reagent concentration deviation, the moment of the next injection can be postponed in the injection sequence, or the proportion of more active reagents in the ratio can be slightly increased, so that the reaction cycle returns to the interval aligned with the light source pulse. Through this flexible adjustment, not only can the instant correction group be restored to a relatively stable baseline output, but also the original periodic correction plan of other channels can be not disrupted in the multi-channel parallel detection mode. When the correction operation is completed, all channels of the instant correction group will output the corrected detection data in the new round of sampling according to the updated modulation and ratio, and these corrected detection data will be merged with the previous records, providing a more reliable and continuous basis for the subsequent multi-parameter comprehensive analysis.

[0123] Please continue reading Figure 1 , performing self-learning evaluation on the corrected detection data, the self-learning evaluation includes updating the recognition model parameters using a progressive stochastic gradient algorithm and multivariate regression analysis, dynamically correcting the channel coupling strength and feature weight based on the updated recognition model parameters, and generating a multi-parameter fingerprint spectrum that characterizes the characteristics of the sample to be tested.

[0124] In one embodiment of the present invention, the self-learning evaluation of the corrected detection data includes using a progressive stochastic gradient algorithm and multivariate regression analysis to update the recognition model parameters, dynamically correcting the channel coupling strength and feature weights based on the updated recognition model parameters, and generating a multi-parameter fingerprint spectrum characterizing the characteristics of the sample to be tested, including: establishing a target feature set for the corrected detection data, setting a heavy metal ion response feature group, a volatile organic compound response feature group, and a protein marker response feature group respectively, to obtain grouped training data; using a progressive stochastic gradient algorithm to perform feature difference analysis on the grouped training data, calculating the recognition parameter deviation of each target feature group in different concentration ranges, and obtaining to feature deviation data; perform multivariate regression analysis on the feature deviation data, calculate the recognition model coefficient and fitting error of each response feature group, and obtain updated recognition model parameters; calculate the coupling strength correction coefficient of each detection channel according to the updated recognition model parameters, distinguish and assign weights of the high-coupling channel group and the low-coupling channel group, and obtain coupling weight data; calculate the response strength weights of different target feature groups based on the coupling weight data, dynamically weight the crosstalk features and complementary features, and obtain feature weight data; perform multi-dimensional combination of the coupling weight data and the feature weight data, construct a response fingerprint database according to the feature distribution of different targets, and generate a multi-parameter fingerprint map characterizing the characteristics of the sample to be tested.

[0125] Specifically, when establishing a target feature set for the corrected detection data, it is necessary to extract a variety of detection results from the output records of the channels that have completed the correction processing, and divide these results into "heavy metal ion response feature group", "volatile organic compound response feature group" and "protein marker response feature group" in combination with the target category information. The "feature set" here not only includes the spectral intensity or absorption peak position of each channel, but also includes the phase synchronization, interference amount and data correction value after baseline drift, thereby forming a high-dimensional data structure. If it is found in actual wastewater analysis that the absorption peak of a certain detection channel for cadmium ions is particularly obvious in the range of 200-250 nanometers, and there is almost no response to organic matter and protein signals, then in this link, the relevant curves and parameters of the channel will be classified into the heavy metal ion response feature group; on the contrary, if another channel mainly produces significant spectral changes for aromatic organic matter, it will be regarded as a key reference in the volatile organic compound response feature group. Grouping can better manage these multi-channel and multi-band differentiated data, laying the foundation for subsequent algorithm training.

[0126] After completing the grouping of feature sets, it is necessary to use the progressive stochastic gradient algorithm to analyze the feature differences of each feature group, and then calculate the recognition parameter deviations under different concentration ranges. The focus of the progressive stochastic gradient algorithm is to allow the model to update the model parameters immediately after processing a small number of training samples (or small batch samples) each time, so that it can reflect the latest observed data characteristics in real time. If it is found in the heavy metal ion response feature group that some ions will have saturated absorption or obvious drift at high concentrations, the algorithm will gradually correct the fitting function of the high concentration area so that the final recognition model maintains a reasonable deviation in both high and low concentration ranges. Feature deviation data refers to the error obtained by comparing the measured value with the model prediction value. This error may be at the amplitude level, the phase offset level, or the synchronization difference between different channels. Through multiple iterations of each concentration range, the system can count the deviation rules of all targets at different concentrations and form a comprehensive feature deviation data report.

[0127] Subsequently, it is necessary to conduct multivariate regression analysis on these characteristic deviation data to further calculate the identification model coefficients and fitting errors of each response feature group, and finally generate updated identification model parameters. The "multivariate regression analysis" here will take the amplitude, phase difference, reaction time and other factors contained in the data as independent variables, and incorporate them into the regression equation, and fit the model as a whole by means of least squares method, ridge regression or partial least squares regression. If some feature groups still maintain good linear response under high temperature conditions or high salinity environments, the model will give higher weights to these feature dimensions during the fitting process; if other features fluctuate greatly within a specific range, deviation correction or adaptive weighting strategies will be adopted. The final identification model parameters include the comprehensive recognition ability of each channel to the target in various concentration ranges and various external environments, providing a reasonable technical support for the subsequent coupling strength correction and weight allocation.

[0128] After obtaining the updated identification model parameters, the coupling strength correction coefficient can be used to distinguish high-coupling channel groups and low-coupling channel groups, and weights can be assigned to form "coupling weight data". In specific practice, the results of the previous cross-correlation or interference analysis will be used to determine which channel combinations have significant synergistic amplification or strong coupling relationships, and then combined with the feedback in the identification model parameters to confirm whether the coupling has a positive contribution to the final detection accuracy. If a pair of high-coupling channels jointly provide an excellent signal-to-noise ratio improvement in heavy metal detection, the system will label them as "high weight" to make them contribute more to the final judgment; if other channel pairs often have reverse interference in a certain band, they will be classified as low coupling or negative coupling, and their conjugate effects will be weakened during integration. In this way, the system can more accurately select valuable signal pathways when fusing multi-channel data, avoiding the erroneous amplification of crosstalk caused by reverse coupling.

[0129] Next, it is necessary to calculate the response intensity weights of different target feature groups based on the coupling weight data, so as to dynamically weight the crosstalk features and complementary features that exist in the detection process, and then generate "feature weight data". For example, if some naphthalene-based organics in the water sample show partially complementary absorption peaks in channel A and channel B, and the model concludes that the two channels have a medium to high degree of positive coupling, then a higher synchronous detection weight can be given to the feature group of naphthalene-based organics; and for other channel combinations that may offset or negatively correlate each other, the combined weight in the feature group is reduced. Dynamic weighting means that the actual detected deviations will be continuously detected during the operation of the model. Once it is found that some channels begin to experience baseline shifts or a sudden increase in noise, their contribution to the overall inference can be immediately weakened to prevent local anomalies from spreading to global judgments.

[0130] Finally, the coupling weight data and the feature weight data are combined in multiple dimensions. By incorporating the performance of each target under different coupling conditions, different spectral channels, and different phase shifts into the same mapping space, a response fingerprint database is constructed, and a multi-parameter fingerprint spectrum reflecting the characteristics of the sample to be tested is generated. The so-called "fingerprint database" is to store these high-dimensional feature coordinates in a customized manner as a set of reference samples. When the system monitors a new sample, its multi-channel response will be mapped to these coordinates and a similarity search will be performed. If a certain heavy metal ion presents a characteristic peak in the UV band and phase synchronization at the same time, and it is highly overlapped with the existing records in the database, the specific concentration range of the ion can be quickly inferred. Such a multi-parameter fingerprint spectrum is very valuable for future batch testing or on-site monitoring, because once an observation value with a high similarity to the recorded cluster appears, the target type can be locked in a very short time and an alarm or concentration range estimate can be given.

[0131] The above describes the multi-parameter coupling detection method of the integrated optical sensor in the embodiment of the present invention. The following describes the multi-parameter coupling detection device of the integrated optical sensor in the embodiment of the present invention. Figure 2 An embodiment of a multi-parameter coupling detection device with an integrated optical sensor in an embodiment of the present invention includes:

[0132] The reagent injection module 101 is used to inject different trace reagent combinations into the sample to be tested in a plurality of microfluidic channels according to a preset time sequence, and obtain differentiated optical response signals of each channel through the selective reaction of the trace reagent combination with the sample to be tested;

[0133] The spectrum acquisition module 102 is used to perform phase shift coding acquisition on the differentiated optical response signal, split the broadband light source into multiple wavelength sub-intervals, generate a coded light signal with a preset phase shift amount for each wavelength sub-interval, input the coded light signal into a corresponding detection channel according to a timing allocation scheme, and obtain multi-channel spectrum evolution data;

[0134] A signal processing module 103 is used to calculate the cross-correlation function between channels according to the multi-channel spectral evolution data, group the detection channels based on the coupling characteristics of the cross-correlation function, perform signal amplification processing on the channel group with high coupling and positive correlation, and perform signal suppression processing on the channel group with high coupling and negative correlation, so as to obtain optimized multi-channel response data;

[0135] The feature extraction module 104 is used to extract multi-dimensional features from the optimized multi-channel response data, calculate the differential interference, amplitude ratio and phase synchronization between channels, and construct a high-dimensional feature vector in combination with the characteristic reaction time of the micro-drug combination, and obtain multi-dimensional feature data of the target object by analyzing the interaction mode of the high-dimensional feature vector;

[0136] The correction control module 105 is used to monitor the baseline drift and noise distribution of each channel using phase-shifted coded pulses according to the multi-dimensional characteristic data of the target object within a preset time window of the detection process, and when the baseline drift exceeds a preset threshold, adjust the phase shift modulation parameters and the drug injection ratio to obtain the corrected detection data;

[0137] The self-learning evaluation module 106 is used to perform self-learning evaluation on the corrected detection data. The self-learning evaluation includes using a progressive stochastic gradient algorithm and multivariate regression analysis to update the recognition model parameters, dynamically correcting the channel coupling strength and feature weight based on the updated recognition model parameters, and generating a multi-parameter fingerprint spectrum that characterizes the characteristics of the sample to be tested.

[0138] above Figure 2The multi-parameter coupling detection device of the integrated optical sensor in the embodiment of the present invention is described in detail from the perspective of modular functional entity. The multi-parameter coupling detection device of the integrated optical sensor in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0139] Figure 3 2 is a schematic diagram of the structure of a multi-parameter coupling detection device of an integrated optical sensor provided by an embodiment of the present invention. The multi-parameter coupling detection device 200 of the integrated optical sensor may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 210 (for example, one or more processors) and a memory 220, and one or more storage media 230 (for example, one or more mass storage device terminals) storing application programs 233 or data 232. Among them, the memory 220 and the storage medium 230 may be temporary storage or permanent storage. The program stored in the storage medium 230 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the multi-parameter coupling detection device 200 of the integrated optical sensor. Furthermore, the processor 210 may be configured to communicate with the storage medium 230, and execute a series of instruction operations in the storage medium 230 on the multi-parameter coupling detection device 200 of the integrated optical sensor to implement the steps of the multi-parameter coupling detection method of the integrated optical sensor.

[0140] The multi-parameter coupling detection device 200 with integrated optical sensor may further include one or more power supplies 240, one or more wired or wireless network interfaces 250, one or more input and output interfaces 260, and / or one or more operating systems 231, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 3 The structure of the multi-parameter coupling detection device of the integrated optical sensor shown does not constitute a limitation on the multi-parameter coupling detection device of the integrated optical sensor provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0141] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the multi-parameter coupling detection method of the integrated optical sensor.

[0142] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0143] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0144] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A multi-parameter coupling detection method of an integrated optical sensor, characterized in that: include: Injecting different micro-agent combinations into the sample to be tested in a plurality of microfluidic channels according to a preset time sequence, and obtaining differentiated optical response signals of each channel through the selective reaction of the micro-agent combination with the sample to be tested; Performing phase shift coding acquisition on the differentiated optical response signal, splitting the broadband light source into a plurality of wavelength sub-intervals, generating a coded light signal with a preset phase shift amount for each wavelength sub-interval, inputting the coded light signal into a corresponding detection channel according to a timing allocation scheme, and obtaining multi-channel spectral evolution data; Calculating the cross-correlation function between channels according to the multi-channel spectral evolution data, grouping the detection channels based on the coupling characteristics of the cross-correlation function, performing signal amplification processing on the channel group with high coupling degree and positive correlation, and performing signal suppression processing on the channel group with high coupling degree and negative correlation, to obtain optimized multi-channel response data; Performing multi-dimensional feature extraction on the optimized multi-channel response data, calculating the differential interference, amplitude ratio and phase synchronization between channels, constructing a high-dimensional feature vector in combination with the characteristic reaction time of the trace drug combination, and obtaining multi-dimensional feature data of the target object by analyzing the interaction mode of the high-dimensional feature vector; Within a preset time window of the detection process, based on the multi-dimensional feature data of the target object, the baseline drift and noise distribution of each channel are monitored using phase-shift coded pulses. When the baseline drift exceeds a preset threshold, the phase-shift modulation parameters and the drug injection ratio are adjusted to obtain corrected detection data; The corrected detection data is subjected to self-learning evaluation, wherein the self-learning evaluation includes updating the recognition model parameters by using a progressive stochastic gradient algorithm and a multivariate regression analysis, dynamically correcting the channel coupling strength and the feature weight based on the updated recognition model parameters, and generating a multi-parameter fingerprint spectrum characterizing the characteristics of the sample to be tested.

2. The multi-parameter coupling detection method of integrated optical sensor according to claim 1, characterized in that: The method of injecting different trace drug combinations into the sample to be tested in a plurality of microfluidic channels according to a preset time sequence, and obtaining differentiated optical response signals of each channel through the selective reaction of the trace drug combination with the sample to be tested, comprises: Acquire background component data of the sample to be tested, classify and mark the sample to be tested according to the background component data, and determine the reference ratio parameters of the sample to be tested; Calculate the amount of medicine injected into each microfluidic channel according to the reference ratio parameter, set a corresponding micro-medicine combination for each microfluidic channel, and determine the injection sequence of the micro-medicine combination; Performing a flow splitting process on the sample to be tested, respectively injecting the corresponding micro-agent combination into the plurality of microfluidic channels according to the injection sequence, and recording the injection time of the micro-agent combination; Calculating the selective reaction time of the combination of the sample to be tested and the trace drug, and setting the data acquisition time window of each channel according to the injection time and the selective reaction time; Tracking the reaction process of the sample to be tested according to the preset time sequence within the data acquisition time window, and recording the selective reaction data of the combination of the sample to be tested and the trace drug at different times; The light intensity variation trend of each channel is calculated according to the selective response data, and the light intensity variation trend is normalized in combination with the reference ratio parameter of the sample to be tested to obtain the differentiated optical response signal of each channel.

3. The multi-parameter coupling detection method of integrated optical sensor according to claim 1, characterized in that: The phase-shift encoding acquisition is performed on the differentiated optical response signal, a broadband light source is split into a plurality of wavelength sub-intervals, a coded light signal with a preset phase shift is generated for each wavelength sub-interval, and the coded light signal is input into a corresponding detection channel according to a timing allocation scheme to obtain multi-channel spectral evolution data, including: Performing spectrum decomposition on the differentiated optical response signal, setting a sampling interval for the differentiated optical response signal according to the characteristic frequency of each channel, and generating a multi-channel sampling sequence; According to the frequency distribution of the multi-channel sampling sequence, the broadband light source is divided into sub-intervals, the broadband light source is split into a plurality of wavelength sub-intervals, and a spectral component of each wavelength sub-interval is obtained; Performing phase modulation calculation on the spectral components, setting a phase offset parameter for each wavelength sub-interval according to the timing characteristics of the multi-channel sampling sequence, and generating an initial coded optical signal; Calculating the phase compensation amount of each wavelength sub-interval according to the amplitude distribution of the differentiated optical response signal, performing phase compensation processing on the initial coded optical signal, and obtaining a coded optical signal with a preset phase shift amount; Based on the time distribution characteristics of the multi-channel sampling sequence, the coded optical signal with the preset phase shift amount is time-sequence divided to generate an interleaved time sequence allocation scheme; According to the interleaved timing distribution scheme, the coded optical signal with the preset phase shift amount is sequentially input into each detection channel, the spectrum response process of each channel is recorded, and multi-channel spectrum evolution data is obtained.

4. The multi-parameter coupling detection method of an integrated optical sensor according to claim 1, characterized in that: The method comprises: calculating the cross-correlation function between channels according to the multi-channel spectral evolution data, grouping the detection channels based on the coupling characteristics of the cross-correlation function, performing signal amplification processing on the channel group with high coupling degree and positive correlation, and performing signal suppression processing on the channel group with high coupling degree and negative correlation, to obtain optimized multi-channel response data, including: Performing time series analysis on the multi-channel spectral evolution data, extracting the spectral response value of each detection channel at a corresponding moment according to a preset sampling period, and obtaining the spectral response sequence of the detection channel; The spectral response sequence is divided into time windows, the sequence length is determined according to the preset analysis time, and the cross-correlation function between any two detection channels is calculated to obtain the coupling characteristic data between the channels, wherein the formula for calculating the cross-correlation function is as follows: ； in, For detection channel and detection channels The cross-correlation value of is the sequence length, and Detection channels and detection channels At sampling time The spectral response sequence, is the time delay, and ; Threshold analysis is performed on the coupling characteristic data between the channels to calculate the cross-correlation function. =0, and divide the detection channels into high coupling degree channel pairs and low coupling degree channel pairs according to the preset coupling degree threshold, to obtain the channel grouping result; According to the channel grouping result, the high coupling channel pairs are polarity classified, the channel pairs whose cross-correlation function peaks are greater than zero are divided into positive correlation channel groups, and the channel pairs whose cross-correlation function peaks are less than zero are divided into negative correlation channel groups, so as to obtain channel group polarity data; Performing synchronous amplification processing on the spectral response sequence of the positive correlation channel group, calculating the gain coefficient of each channel pair according to the positive peak value of the cross-correlation function, performing amplitude modulation on the spectral response value, and obtaining positive correlation signal modulation data; Suppression processing is performed on the spectral response sequence of the negative correlation channel group, the suppression coefficient of each channel pair is calculated according to the negative peak value of the cross-correlation function, the crosstalk component is selectively attenuated, and negative correlation signal suppression data is obtained; The positive correlation signal modulation data and the negative correlation signal suppression data are integrated to obtain optimized multi-channel response data.

5. The multi-parameter coupling detection method of integrated optical sensor according to claim 1, characterized in that: The multi-dimensional feature extraction is performed on the optimized multi-channel response data, the differential interference, amplitude ratio and phase synchronization between channels are calculated, and a high-dimensional feature vector is constructed in combination with the characteristic reaction time of the trace drug combination. The multi-dimensional feature data of the target object is obtained by analyzing the interaction mode of the high-dimensional feature vector, including: The optimized multi-channel response data is divided into characteristic frequency bands, and the spectral data of each channel is segmented according to the characteristic absorption intervals of heavy metal ions, organic matter and protein to obtain multi-interval spectral distribution data; Calculating the differential interference between any two detection channels according to the multi-interval spectral distribution data, and analyzing the amplitude difference and phase difference between the channels using a coherent spectrum analysis method to obtain interference characteristic data between the channels; Normalizing the interference characteristic data between the channels, calculating the amplitude ratio of adjacent detection channels, extracting the phase difference as a phase synchronization parameter, and obtaining channel correlation characteristic data; Dynamically analyzing the reaction process between the trace drug combination and the target substance, calculating the characteristic reaction time according to the inflection point moment of the spectral response of each channel, and obtaining drug reaction characteristic data; Combining the interference feature data between the channels, the channel association feature data and the drug reaction feature data to construct a high-dimensional feature vector, and obtaining feature space distribution data; Cluster analysis is performed on the interaction patterns in the feature space distribution data, and feature parameters of each dimension are extracted according to the feature distribution rules of different target objects to obtain multi-dimensional feature data of the target objects.

6. The multi-parameter coupling detection method of integrated optical sensor according to claim 1, characterized in that: The method comprises: monitoring the baseline drift and noise distribution of each channel by using phase-shifted coded pulses according to the multi-dimensional characteristic data of the target object within a preset time window of the detection process; and adjusting the phase-shifted modulation parameters and the drug injection ratio when the baseline drift exceeds a preset threshold value to obtain the corrected detection data, including: Classifying the detection channels according to the multidimensional feature data of the target object, setting different monitoring parameters according to the heavy metal detection channel, the organic matter detection channel and the protein detection channel, and obtaining a classified monitoring plan; Based on the classification monitoring scheme, phase-shifted coded pulse sequences of different periods are generated, and various detection channels are monitored in an interlaced manner within the preset time window to obtain channel response timing data; Performing baseline extraction on the channel response time series data, calculating the baseline drift and noise standard deviation at short time scales and long time scales respectively, and obtaining channel stability data; The baseline drift of each detection channel is compared with a preset threshold value according to the channel stability data, and the channels with excessive baseline drift are divided into an immediate correction group and a periodic correction group to obtain a graded correction scheme; Optimizing the phase shift modulation parameters of the detection channel of the instant correction group, determining the phase compensation amount and the duty cycle parameters according to the difference in the target category, and obtaining the modulation correction data; The medicine injection ratio of the instant correction group is reconfigured according to the modulation correction data, the injection sequence is set in combination with the characteristic reaction time of different targets, the detection channel is calibrated according to the graded correction scheme, and the corrected detection data is obtained.

7. The multi-parameter coupling detection method of integrated optical sensor according to claim 1, characterized in that: The self-learning evaluation of the corrected detection data includes updating the recognition model parameters by using a progressive stochastic gradient algorithm and a multivariate regression analysis, dynamically correcting the channel coupling strength and the feature weight based on the updated recognition model parameters, and generating a multi-parameter fingerprint spectrum characterizing the characteristics of the sample to be tested, including: Establishing a target feature set for the corrected detection data, respectively setting a heavy metal ion response feature group, a volatile organic compound response feature group, and a protein marker response feature group to obtain group training data; The grouped training data are subjected to feature difference analysis using a progressive stochastic gradient algorithm, and the identification parameter deviations of each target feature group in different concentration ranges are calculated to obtain feature deviation data; Performing multivariate regression analysis on the characteristic deviation data, calculating the identification model coefficients and fitting errors of each response characteristic group, and obtaining updated identification model parameters; Calculating the coupling strength correction coefficient of each detection channel according to the updated recognition model parameters, distinguishing and assigning weights of the high coupling channel group and the low coupling channel group, and obtaining coupling weight data; Calculating the response intensity weights of different target feature groups based on the coupling weight data, dynamically weighting the crosstalk features and complementary features, and obtaining feature weight data; The coupling weight data and the characteristic weight data are combined in multiple dimensions, a response fingerprint database is constructed according to the characteristic distribution of different targets, and a multi-parameter fingerprint spectrum characterizing the characteristics of the sample to be tested is generated.

8. A multi-parameter coupling detection device with integrated optical sensor, characterized in that: The multi-parameter coupling detection device of the integrated optical sensor adopts the multi-parameter coupling detection method of the integrated optical sensor according to any one of claims 1 to 7, and the multi-parameter coupling detection device of the integrated optical sensor comprises: A drug injection module is used to inject different trace drug combinations into the sample to be tested in a plurality of microfluidic channels according to a preset time sequence, and obtain differentiated optical response signals of each channel through the selective reaction of the trace drug combination with the sample to be tested; A spectrum acquisition module, used for performing phase shift coding acquisition on the differentiated optical response signal, splitting the broadband light source into a plurality of wavelength sub-intervals, generating a coded light signal with a preset phase shift amount for each wavelength sub-interval, inputting the coded light signal into a corresponding detection channel according to a timing allocation scheme, and obtaining multi-channel spectrum evolution data; A signal processing module, used to calculate the cross-correlation function between channels according to the multi-channel spectral evolution data, group the detection channels based on the coupling characteristics of the cross-correlation function, perform signal amplification processing on the channel group with high coupling degree and positive correlation, and perform signal suppression processing on the channel group with high coupling degree and negative correlation, so as to obtain optimized multi-channel response data; A feature extraction module is used to extract multi-dimensional features from the optimized multi-channel response data, calculate the differential interference, amplitude ratio and phase synchronization between channels, and construct a high-dimensional feature vector in combination with the characteristic reaction time of the micro-drug combination, and obtain multi-dimensional feature data of the target object by analyzing the interaction mode of the high-dimensional feature vector; A correction control module is used to monitor the baseline drift and noise distribution of each channel using phase-shifted coded pulses according to the multi-dimensional feature data of the target object within a preset time window of the detection process, and when the baseline drift exceeds a preset threshold, adjust the phase shift modulation parameters and the drug injection ratio to obtain corrected detection data; A self-learning evaluation module is used to perform self-learning evaluation on the corrected detection data. The self-learning evaluation includes updating the recognition model parameters using a progressive stochastic gradient algorithm and multivariate regression analysis, dynamically correcting the channel coupling strength and feature weight based on the updated recognition model parameters, and generating a multi-parameter fingerprint spectrum that characterizes the characteristics of the sample to be tested.

9. A multi-parameter coupling detection device with integrated optical sensor, characterized in that: The multi-parameter coupling detection device with integrated optical sensor comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the multi-parameter coupling detection device of the integrated optical sensor to perform the steps of the multi-parameter coupling detection method of the integrated optical sensor as described in any one of claims 1-7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the steps of the multi-parameter coupling detection method of the integrated optical sensor as claimed in any one of claims 1 to 7 are implemented.

Images