Intelligent early warning system for plant diseases and insect pests based on quantum dot light emitting characteristics

Abstract

The application relates to the technical field of agricultural monitoring, and discloses a plant disease and pest intelligent early warning system based on quantum dot light emitting characteristics, which comprises a multivariate composite quantum dot sensing array unit, a coherent time sequence coding excitation unit, a single photon time correlation counting detection unit and a central processing and alarm decision unit. In work, the sensing array unit is acted on by plant stress metabolites and generates a differentiated fluorescent lifetime response; the detection unit acquires time-resolved data flow of the response; the central processing and alarm decision unit processes data and constructs a multimodal composite fingerprint fused by a real-time lifetime vector, a real-time cross-correlation matrix and a synergistic response dynamic curve; the composite fingerprint is input to a preset classification model for matching operation to output accurate prediction of the plant health state. Through construction of a multidimensional composite fingerprint, early and accurate intelligent early warning of plant stress events is realized, and the accuracy and specificity of diagnosis are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural monitoring, in particular to a plant disease and pest intelligent early warning system based on quantum dot luminescence characteristics. BACKGROUND

[0002] Early and accurate warning of plant diseases and pests is crucial for food security and sustainable agricultural development. Traditional field management mainly relies on manual inspection and visual diagnosis by experts. This method not only consumes a lot of manpower, but also has strong subjectivity. When the symptoms of the disease become obvious on the appearance of the plant, the best prevention and control opportunity has usually been missed, which may result in significant yield loss.

[0003] To achieve early diagnosis, various technical means have been developed in the industry. Laboratory molecular diagnostic methods, such as polymerase chain reaction (PCR) or enzyme-linked immunosorbent assay (ELISA), can provide high-precision pathogen identification, but the process requires destructive sampling and is complex, time-consuming and costly, making it difficult to apply to in-situ and continuous monitoring in large-scale fields.

[0004] In recent years, non-destructive monitoring technologies such as spectral imaging and thermal imaging have been applied to some extent. These technologies analyze changes in plant reflectance spectra or surface temperature to assess their health status. However, these technologies mainly capture changes in macro physiological parameters such as chlorophyll content, water stress or crown temperature anomalies exhibited by plants under stress, and their response often lags behind the underlying biochemical changes, so there is still a lack of timeliness in early warning. Another type of technology is to monitor volatile organic compounds (VOCs) released by plants under stress using sensor arrays such as electronic noses. Although this method has the potential for non-invasive and early monitoring, the response signal of existing sensors is usually a comprehensive result of the cross-reaction of multiple gases, which leads to insufficient selectivity and sensitivity, making it difficult to accurately distinguish different stress sources or disease types, and its performance is easily disturbed by environmental temperature and humidity and complex background gases, affecting the reliability of the early warning results.

[0005] The existing technology generally has the technical bottleneck of single dimension of early warning information, which is difficult to reveal the intrinsic biochemical correlation of stress events, in realizing timely, accurate and deep diagnostic capability of early warning of plant diseases and pests.

[0006] Therefore, the present application proposes a plant disease and pest intelligent early warning system based on quantum dot luminescence characteristics to solve the deficiencies of the prior art. SUMMARY

[0007] In view of the deficiencies of the prior art, the present application provides a plant disease and pest intelligent early warning system based on quantum dot luminescence characteristics, solves the problems of single diagnostic dimension, lagging early warning, insufficient specificity and difficulty in revealing the internal dynamic correlation of stress events in the plant disease and pest monitoring technical means, thereby limiting the realization of early and accurate early warning.

[0008] To achieve the above object, the present application is implemented by the following technical solutions:

[0009] The first aspect of the present application provides a plant disease and pest intelligent early warning system based on quantum dot luminescence characteristics. The system comprises a multi-element composite quantum dot sensing array unit, a coherent time sequence coding excitation unit, a single photon time correlation counting detection unit and a central processing and alarm decision unit.

[0010] The multi-element composite quantum dot sensing array unit has a plurality of differently surface-modified quantum dots integrated on the substrate, which are used to interact with plant stress metabolites to produce differential fluorescence lifetime responses and emit fluorescence after being excited.

[0011] The coherent time sequence coding excitation unit is used to emit precisely timed pulsed light to the multi-element composite quantum dot sensing array unit.

[0012] The single photon time correlation counting detection unit is used to detect the fluorescence emitted by the multi-element composite quantum dot sensing array unit and generate a time-resolved data stream recording the arrival time of fluorescence photons relative to the excitation pulse.

[0013] The central processing and alarm decision unit is connected with the coherent time sequence coding excitation unit and the single photon time correlation counting detection unit, respectively. The central processing and alarm decision unit is used to control the coherent time sequence coding excitation unit to emit precisely timed pulsed light, receive and process the time-resolved data stream from the single photon time correlation counting detection unit to calculate the fluorescence lifetime representing the differential fluorescence lifetime response, construct a composite fingerprint based on the fluorescence lifetime, then input the composite fingerprint into a pre-set classification model for matching operation to output a prediction label and a confidence, and finally make an alarm decision based on the prediction label and the confidence to output an alarm signal.

[0014] As a preferred technical solution, the differential surface modification strategy of the multi-element composite quantum dot sensing array unit comprises at least one of the following: constructing a pH response channel, modifying a ligand with a protonatable or deprotonatable group on the surface of the quantum dot, so that the fluorescence lifetime of the quantum dot responds to the change of pH value; constructing a redox response channel, fixing a molecule sensitive to active oxygen on the surface of the quantum dot, so that the fluorescence lifetime of the quantum dot responds to the redox reaction; constructing a volatile organic compound affinity response channel, modifying a functional group or polymer film with physical adsorption or weak chemical action on the surface of the quantum dot, so that the fluorescence lifetime of the quantum dot responds to the adsorption of volatile organic compounds; constructing a specific signal molecule response channel, preparing a molecularly imprinted polymer film on the surface of the quantum dot, so that the fluorescence lifetime of the quantum dot responds to the capture of the target signal molecule.

[0015] As a preferred technical solution, the coherent time sequence coding excitation unit is controlled by the central processing and alarm decision unit, and is used for generating at least two excitation modes: one is a single-channel time sequence coding pulse sequence for reconnaissance scanning and adaptive interrogation; and the other is a time-accurately phase-locked double-channel or multi-channel pulse sequence for performing cooperative response dynamics measurement.

[0016] As a preferred technical solution, the single-photon time correlation counting detection unit comprises a time correlation counting electronic module, which receives a synchronization output signal from the coherent time sequence coding excitation unit and a single-photon response pulse from a single-photon detector, and constructs a curve of fluorescence intensity decay with time by counting the time interval of the single-photon response pulse relative to the synchronization output signal in multiple excitation events, and the central processing and alarm decision unit calculates the fluorescence lifetime by fitting the curve.

[0017] As a preferred technical solution, the composite fingerprint constructed by the application innovatively integrates multiple dimensions of data, specifically including: a real-time lifetime vector, a real-time cross-correlation matrix, and one or more cooperative response dynamics curves.

[0018] As a preferred technical solution, the specific way in which the central processing and alarm decision unit constructs the composite fingerprint is: constructing the real-time lifetime vector based on the real-time fluorescence lifetime of each sensing channel; calculating the normalized covariance between different channels based on the time sequence of the fluorescence lifetime of each sensing channel, and constructing the real-time cross-correlation matrix, the matrix element of which is calculated by the following formula:

[0019] ;

[0020] Wherein, represents the real-time cross-correlation matrix In particular, the matrix element located at the i-th row and the j-th column; In particular, the matrix element located at the i-th row and the j-th column; In particular, the matrix element located at the i-th row and the j-th column; In particular, the matrix element located at the i-th row and the j-th column; In particular, the matrix element located at the i-th row and the j-th column; In particular, the matrix element located at the i-th row and the j-th column; In particular, the matrix element located at the i-th row and the j-th column; In particular, the matrix element located at the i-th row and the j-th column; In particular, the matrix element located at the i-th row and the j-th column; In particular, the matrix element located at the i-th row and the j-th column; In particular, the matrix element located at the i-th row and the j-th column; In particular, the matrix element located at the i-th row and the j-th column; In particular, the matrix element located at the i-th row and the j-th column; represents the mathematical expectation. And by performing a pump-probe scan, the fluorescence lifetime of the probe channel is measured as a function of the pump-probe delay time, and the cooperative response dynamics curve is constructed, which is defined as:

[0021] ;

[0022] wherein, is the cooperative response dynamics curve function; is the fluorescence lifetime of the probe channel measured when the pump-probe delay is is the fluorescence lifetime of the probe channel measured when the pump-probe delay is is the reference fluorescence lifetime of the probe channel in the healthy state.

[0023] As a preferred technical solution, the classification model stored in the central processing and alarm decision unit is a multi-modal fusion Transformer network, or a tensor decomposition algorithm combined with a support vector machine classifier.

[0024] As a preferred technical solution, the central processing and alarm decision unit executes a two-stage intelligent monitoring strategy, specifically: in the normal monitoring mode, periodic reconnaissance scanning and anomaly detection are performed, and when a normalized deviation score based on real-time fluorescence lifetime calculation exceeds a preset threshold, the system automatically switches from the normal monitoring mode to the deep analysis mode to perform the construction of the composite fingerprint.

[0025] As a preferred technical solution, when the central processing and alarm decision unit makes an alarm decision, an alarm signal is output only when the predicted label is not the healthy state and the confidence exceeds a preset alarm threshold.

[0026] The second aspect of the present application provides a plant disease and pest intelligent early warning method based on the luminescence characteristics of quantum dots, which comprises the following steps:

[0027] The coherent time sequence coding excitation unit controlled by the central processing and alarm decision unit transmits precise time sequence pulse light to excite the multi-element composite quantum dot sensing array unit to generate fluorescence;

[0028] The fluorescence is detected by the single photon time correlation counting detection unit, and a time resolution data stream is generated;

[0029] Then, the central processing and alarm decision unit processes the time resolution data stream to calculate the fluorescence lifetime, and constructs a composite fingerprint fused by a real-time lifetime vector, a real-time cross-correlation matrix and a synergistic response dynamic curve;

[0030] The composite fingerprint is input into a preset classification model for matching operation, and a prediction label and a confidence are output;

[0031] When the prediction label is not a healthy state and the confidence exceeds a preset alarm threshold, an alarm signal is output by the central processing and alarm decision unit.

[0032] The present application provides a plant disease and pest intelligent early warning system based on the luminescence characteristics of quantum dots.

[0033] 1、The present application can comprehensively characterize plant stress events from three dimensions of state, correlation and process by constructing a multi-modal composite fingerprint fused by a real-time lifetime vector, a real-time cross-correlation matrix and a synergistic response dynamic curve; compared with traditional monitoring technologies relying on a single parameter, this multi-dimensional data fusion method provides more rich and in-depth diagnostic information, so that the system can effectively distinguish different types or different stages of diseases and pests, thereby improving the accuracy and specificity of early warning.

[0034] 2、The present application adopts a dual-mode monitoring strategy combining periodic reconnaissance scanning and deep analysis; the system is in a low-power reconnaissance mode during regular monitoring, uses a high-sensitivity multi-element composite quantum dot sensing array unit to continuously capture weak biochemical signal changes in the environment, and immediately switches to a deep analysis mode for accurate construction of a composite fingerprint once an anomaly is detected; this strategy ensures that the system can timely discover early signs of stress events, realizes early warning, and gains valuable time for intervention measures in agricultural production.

[0035] 3、The present application innovatively introduces a synergistic response dynamic detection function; by controlling the coherent time sequence coding excitation unit to perform pump detection scanning, the system can quantify the conduction dynamic process of response signals between different sensing channels; the constructed synergistic response dynamic curve reveals the internal time sequence correlation and interaction strength of different plant stress metabolite production or conduction processes, so that the system can not only make state judgments, but also provide insights into stress response mechanisms, providing a new data dimension for pathological analysis. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the intelligent early warning system for plant diseases and pests of the present invention;

[0037] Figure 2 This is a schematic cross-sectional view of the multi-component composite quantum dot sensing array unit of the present invention;

[0038] Figure 3 This is a flowchart illustrating the plant disease and pest composite fingerprint early warning method of the present invention.

[0039] Among them, 10 is a multi-component composite quantum dot sensing array unit; 20 is a coherent timing coding excitation unit; 30 is a single-photon time-correlation counting detection unit; and 40 is a central processing and alarm decision unit. Detailed Implementation

[0040] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] See attached document Figure 1 , Figure 1 This is a schematic diagram of the structure of an intelligent early warning system for plant diseases and pests according to an embodiment of the present invention. The present invention provides an intelligent early warning system for plant diseases and pests, which may include: a multi-component composite quantum dot sensor array unit 10, a coherent timing coding excitation unit 20, a single photon time correlation counting detection unit 30, and a central processing and alarm decision unit 40.

[0042] The multi-component quantum dot sensing array unit 10 serves as the sensing interface of the system and is configured to interact with the plant under test or its surrounding environment. Multiple quantum dots are integrated on the substrate of this unit. Different types of quantum dots are modified with specific surface chemicals to produce differentiated fluorescence lifetime responses to different types of plant stress metabolites. When interacting with target metabolite molecules, the local chemical microenvironment of the quantum dots changes, thereby causing specific changes in their fluorescence lifetime.

[0043] A coherent time-encoding excitation unit 20, which is configured to provide a light source for the multi-element composite quantum dot sensing array unit 10; the coherent time-encoding excitation unit 20 is electrically connected to and controlled by the central processing and alarm decision unit 40, and is configured to emit pulsed light with precise timing to the multi-element composite quantum dot sensing array unit 10; according to the instruction of the central processing and alarm decision unit 40, the coherent time-encoding excitation unit 20 can generate at least two excitation modes: one is a single-channel time-encoding pulse sequence, which is used for reconnaissance scanning and adaptive interrogation; the other is a time-precise phase-locked dual-channel or multi-channel pulse sequence, which is used for performing pump-probe type cooperative response dynamics measurement;

[0044] A single-photon time-correlated counting detection unit 30, which is configured to detect the fluorescence emitted by the multi-element composite quantum dot sensing array unit 10; the detection end of the single-photon time-correlated counting detection unit 30 faces the multi-element composite quantum dot sensing array unit 10, and is configured to receive the fluorescence photons; the data output end of the single-photon time-correlated counting detection unit 30 is electrically connected to the central processing and alarm decision unit 40; the single-photon time-correlated counting detection unit 30 can accurately record the arrival time of each detected fluorescence photon relative to the excitation pulse emitted by the coherent time-encoding excitation unit 20, and generate a time-resolved raw data stream;

[0045] A central processing and alarm decision unit 40, which is the control and calculation core of the system, and is connected to the coherent time-encoding excitation unit 20 and the single-photon time-correlated counting detection unit 30 for control and data communication; the central processing and alarm decision unit 40 is configured to generate and send control instructions to the coherent time-encoding excitation unit 20 to control the timing and mode of the excitation light; at the same time, the central processing and alarm decision unit 40 is configured to receive and process the time-resolved data from the single-photon time-correlated counting detection unit 30; based on a preset algorithm, the central processing and alarm decision unit 40 is configured to perform fluorescence lifetime calculation, composite fingerprint construction and analysis; and according to the final analysis result, the central processing and alarm decision unit 40 is configured to make an alarm decision and output an alarm signal;

[0046] In this embodiment, the working principle of the system is as follows: the central processing and alarm decision unit 40 first instructs the coherent time-encoding excitation unit 20 to emit a series of excitation pulses to the multi-element composite quantum dot sensing array unit 10; the multi-element composite quantum dot sensing array unit 10 generates fluorescence after being excited, and the single-photon time-correlated counting detection unit 30 captures the fluorescence photons and records their time information; the time information is transmitted to the central processing and alarm decision unit 40 for subsequent data processing and analysis;

[0047] The central processing and alarm decision unit 40 constructs a curve of fluorescence intensity decay with time by analyzing the data collected by the single-photon time-correlated counting detection unit 30.

[0048] Referring to the accompanying Figure 1 and the accompanying Figure 2, attached Figure 2 is a schematic diagram of the structure profile of a multi-element composite quantum dot sensing array unit according to an embodiment of the present application.

[0049] The multi-element composite quantum dot sensing array unit 10 functions as the core sensing interface of the system, which is used to convert the trace amount of chemical information released by plants under different physiological states into fluorescent lifetime signals that can be recognized by the optical system; the specific physical form of the multi-element composite quantum dot sensing array unit 10 can be configured according to the application scenario; in one embodiment, the multi-element composite quantum dot sensing array unit 10 is prepared on the inner channel surface of a microfluidic chip composed of polydimethylsiloxane (PDMS) or glass material, and the channel structure is used to guide the measured gas sample to flow accurately through the sensing area; in another embodiment, the multi-element composite quantum dot sensing array unit 10 is constructed on the distal end surface of an optical fiber to form an optical fiber probe suitable for in-situ and minimally invasive monitoring of the environment inside the plant canopy or on the leaf surface; in yet another embodiment, the multi-element composite quantum dot sensing array unit 10 can be a planar glass or silicon substrate, on which different types of quantum dots are fixed in an array form by microarray spotting or inkjet printing technology, for deployment in a closed plant incubator or environmental monitoring cabin;

[0050] The quantum dots constituting the multi-element composite quantum dot sensing array unit 10 are selected in accordance with the principle of multi-signal channel construction; in one embodiment, quantum dots with different sizes of core / shell structures are selected, such as cadmium selenide / zinc sulfide (CdSe / ZnS) or indium phosphide / zinc sulfide (InP / ZnS) quantum dots; by precisely controlling the size of the quantum dot core, the peak wavelength of its emission spectrum can be precisely adjusted, so that different quantum dot types in the array have mutually separated and spectrally distinguishable emission channels, for example, set at 520 nanometers, 580 nanometers and 650 nanometers respectively; such spectral separation enables the single-photon time-correlated counting detection unit 30 to set corresponding bandpass filters in the detection light path, and acquire and analyze the fluorescent lifetime signals of each sensing channel in parallel without cross talk;

[0051] The core technical feature of the multi-element composite quantum dot sensing array unit 10 lies in the differential surface modification strategy implemented on the quantum dots; the purpose of this strategy is not to achieve absolute specific recognition of a single target molecule, but to construct a sensing system with differential response patterns to different categories of stress metabolites; by analyzing the response fingerprints of the entire array, complex biochemical events can be analyzed; the specific surface modification strategy can include the following implementation methods:

[0052] S101: Constructing pH-responsive channel, ligands with protonatable or deprotonatable groups, such as mercaptopropionic acid (MPA) or other short-chain mercapto carboxylic acids, are modified on the surface of a type of quantum dots by chemical bonding; when plants are infected by pathogens, the pH value of the local microenvironment will change; this change in pH will cause the degree of protonation of the carboxyl groups on the surface ligands of the quantum dots to change, thereby changing the net charge and dipole moment of the quantum dot surface, which will affect the non-radiative recombination rate of quantum dot excitons, ultimately resulting in a measurable change in its fluorescence lifetime ;

[0053] S102: Constructing redox-responsive channel, molecules sensitive to reactive oxygen species (ROS), such as glutathione (GSH) or viologen-based derivatives, are immobilized on the surface of another type of quantum dots; when plants encounter biological or abiotic stress, they will produce explosive reactive oxygen species (such as hydrogen peroxide, superoxide anion, etc.); these reactive oxygen species will undergo redox reactions with the sensitive molecules on the surface of the quantum dots, efficiently quenching or enhancing the fluorescence of the quantum dots through charge transfer or energy transfer processes, resulting in a significant change in their fluorescence lifetime ;

[0054] S103: Constructing volatile organic compound (VOCs) affinity-responsive channel, functional groups or polymer films with physical adsorption or weak chemical action on specific types of VOCs are modified on the surface of other quantum dot channels; in one specific implementation, ligands with primary amine groups (-NH2), such as cysteamine, can be modified on the surface of quantum dots to preferentially adsorb aldehyde or ketone VOCs released by plants under stress, which changes the dielectric constant around the quantum dots, thereby modulating their fluorescence lifetime ; in another specific implementation, ligands with hydrophobic alkyl long chains can be modified on the surface of quantum dots to enrich non-polar alkane or alkene VOCs, affecting the fluorescence lifetime by changing the solvation environment ;

[0055] S104: Constructing specific signal molecule-responsive channel, molecular probes capable of specific recognition with plant endogenous signal molecules (such as salicylic acid, jasmonic acid, etc.), such as molecularly imprinted polymer (MIP) thin layers, are prepared on the surface of another type of quantum dots; when the target signal molecule is captured by the MIP layer, it will cause a change in the conformation of the polymer or the local refractive index, which is transmitted to the surface of the quantum dots, also causing a predictable change in their fluorescence lifetime ;

[0056] By the combination of one or more of the above modification strategies, the multi-element composite quantum dot sensing array unit 10 can deconstruct a single, ambiguous disease event into a set of response vectors with clear physical and chemical meanings composed of the fluorescence lifetime changes of multiple channels, providing rich and high-dimensional raw sensing data for subsequent composite fingerprint analysis.

[0057] Referring to the drawings Figure 1 The coherent time-encoded excitation unit 20 is used to provide pulsed excitation light with programmable time structure and wavelength selectivity for the multi-element composite quantum dot sensing array unit 10 under the precise control of the central processing and alarm decision unit 40;

[0058] In a specific embodiment, the coherent time-encoded excitation unit 20 includes one or more picosecond pulsed laser diodes and corresponding beam shaping and combining optical elements; for example, the coherent time-encoded excitation unit 20 can integrate a laser diode array, where each laser diode emits pulsed light of a different central wavelength, each wavelength corresponding to the absorption peak of a different kind of quantum dot in the multi-element composite quantum dot sensing array unit 10; the drive circuit of each laser diode is connected to an independent time sequence control channel of the central processing and alarm decision unit 40;

[0059] In another embodiment, the coherent time-encoded excitation unit 20 can use a broadband ultrafast laser source (such as a supercontinuum laser source) in combination with a fast wavelength selection module; the wavelength selection module can be an acousto-optic tunable filter (AOTF), which can quickly switch the wavelength of the transmitted light on the time scale of nanoseconds to microseconds by applying radio frequency signals of different frequencies;

[0060] In yet another embodiment, the coherent time-encoded excitation unit 20 can use a single-wavelength pulsed laser source, whose beam is expanded to uniformly illuminate the entire multi-element composite quantum dot sensing array unit 10; a spatial light modulator, such as a digital micromirror device (DMD), is inserted into the optical path; the central processing and alarm decision unit 40 dynamically controls the deflection state of the micromirrors on the DMD by sending a series of binary images to the DMD, thereby achieving selective modulation of the excitation light in space, so that only the quantum dots at specific positions are illuminated by the excitation light;

[0061] The coherent time-encoded excitation unit 20, under the control of the central processing and alarm decision unit 40, is used to generate two core excitation modes, the generation methods of which are described in detail as follows:

[0062] S201: generating single-channel time-encoding code (TEC); this excitation mode is mainly used for periodic reconnaissance scanning and adaptive interrogation steps; the FPGA or microcontroller inside the central processing and alarm decision unit 40 generates a digitized time-encoding sequence according to a preset algorithm; the sequence is sent to the driving circuit of the coherent time-encoding excitation unit 20; if a laser diode array is used, the sequence directly controls the laser diodes of different wavelengths in the array to turn on and off in sequence or in combination at preset time points; if an AOTF is used, the sequence is converted into control instructions for the AOTF driving frequency, so that it quickly switches the transmission wavelength according to the encoding rule; if a DMD is used, the sequence is converted into a series of DMD image frames to control the illumination of a specific spatial region on the time axis;

[0063] S202: generating time-accurate phase-locked dual-channel "pump-probe" pulse sequence; this excitation mode is the key to realizing the cooperative response dynamics detection step; its generation process is strictly controlled by a high-precision clock and a programmable delay generator inside the central processing and alarm decision unit 40; when it is necessary to measure the channel (pump) and the channel (probe), the timing control is as follows:

[0064] First, at time point , the central processing and alarm decision unit 40 sends out a trigger signal to instruct the coherent time-encoding excitation unit 20 to generate a beam of excitation pulses for exciting quantum dots on the channel ;

[0065] Second, the programmable delay generator inside the central processing and alarm decision unit 40 starts timing and goes through an accurately set delay time ; the value of the delay time is programmable and can be step-scanned in the range of nanoseconds to milliseconds;

[0066] Finally, at time point , the central processing and alarm decision unit 40 sends out a second trigger signal to instruct the coherent time-encoding excitation unit 20 to generate another beam of excitation pulses for exciting quantum dots on the channel ;

[0067] The pulse sequence of the entire excitation event can be represented as:

[0068] ;

[0069] Wherein, represents the total excitation light intensity as a function of time; ​a waveform function representing the actual pump pulse; a time variable; a normalized waveform function representing a single excitation pulse; and representing the amplitudes of the pump pulse and the probe pulse, respectively; a waveform function representing the actual probe pulse, is a reference waveform function shifted in time by an amount whose center is aligned with the time point , which explicitly indicates that the second excitation pulse (probe pulse) occurs after the first excitation pulse (pump pulse) with a delay of , which fully describes the temporal characteristics of the second excitation pulse; is the starting time of the pump pulse; is the precisely controlled time delay between the pump pulse and the probe pulse; at the same time, the synchronization signal (SYNC) sent to the single-photon time-correlated counting detection unit 30 is strictly synchronized with the starting time point of the pump pulse to ensure the accuracy of the subsequent fluorescence decay time measurement.

[0070] Referring to the accompanying Figure 1 , the single-photon time-correlated counting detection unit 30 is responsible for collecting the fluorescence photons emitted by the multi-element composite quantum dot sensing array unit 10 and accurately recording each photon event with extremely high time resolution, providing raw data for the central processing and alarm decision unit 40 to perform fluorescence lifetime calculation;

[0071] In a specific embodiment, the single-photon time-correlated counting detection unit 30 includes a fluorescence collection optical system, a spectral separation system, and a single-photon detector array; the fluorescence collection optical system, such as a high numerical aperture microscope objective or lens group, is used to efficiently collect the scattered fluorescence emitted from the multi-element composite quantum dot sensing array unit 10; the collected fluorescence then passes through a dichroic mirror, which reflects or filters out the shorter wavelength excitation light and only allows the longer wavelength fluorescence to pass through; the spectral separation system, such as a motorized filter wheel equipped with multiple different center wavelength bandpass filters, or a set of fixedly installed spectral elements combined with bandpass filters, is responsible for spatially or temporally separating the fluorescence from different types of quantum dots and directing them to different detection units of the single-photon detector array;

[0072] The single-photon detector array can be composed of multiple single-photon avalanche diodes (SPADs) or silicon photomultiplier (SiPM) units; each detection unit corresponds to a sensing channel and can independently detect single-photon events and output a fast electrical pulse signal;

[0073] The core working principle of the single-photon time-correlated counting detection unit 30 is realized through the following steps:

[0074] S301: The time-correlated counting electronics module inside the single-photon time-correlated counting detection unit 30 receives two input signals: the first signal is a synchronization output signal (SYNC) from the coherent time sequence encoding excitation unit 20, and the pulse front of the signal is strictly synchronized with the pulse front of each excitation light; the second signal is a single-photon response pulse output from any detection unit in the single-photon detector array;

[0075] S302: Whenever the coherent time sequence encoding excitation unit 20 sends out an excitation pulse, a SYNC signal is sent to the time-correlated counting electronics module at the same time, and a high-precision time-to-digital converter (TDC) or time-to-amplitude converter (TAC) inside the time-correlated counting electronics module is started, beginning a timing period;

[0076] S303: After the excitation pulse, if the fluorescence photon emitted by the multi-element composite quantum dot sensing array unit 10 is captured by a certain detection unit in the single-photon detector array, the detection unit will generate an electrical pulse; this electrical pulse is sent to the time-correlated counting electronics module as a stop signal, thereby terminating the current timing period;

[0077] S304: The time-to-digital converter (TDC) directly outputs a digital value proportional to the time interval between the arrival of the SYNC signal and the arrival of the photon response pulse ; or, in the implementation of the time-to-amplitude converter (TAC), it outputs a voltage pulse whose amplitude is proportional to the time interval ; the voltage pulse is then converted to a digital address by an analog-to-digital converter (ADC);

[0078] S305: A multi-channel analyzer (MCA) or a memory address space inside the central processing and alarm decision unit 40, the address of which corresponds to the digital address generated in step S304; each time a time interval is measured, the count value in the corresponding memory address is increased by one;

[0079] S306: The steps S301 to S305 are repeated millions of times at an extremely high frequency (e.g., megahertz level); since fluorescence emission is a random process, after a large amount of statistics, the count values of different addresses in the memory form a distribution histogram; the abscissa of the histogram represents the arrival time of the photon , the ordinate represents the number of photons detected in the time window; this finally formed histogram, namely the curve of fluorescence intensity decay with time, is the direct experimental basis for calculating the fluorescence lifetime ; by fitting analysis of the histogram data, the fluorescence lifetime of each sensing channel can be accurately extracted.

[0080] Referring to the accompanying drawings Figure 1 , the central processing and alarm decision unit 40, as the control center and data processing core of the entire early warning system, is responsible for coordinating the operation of all other units in the system and executing the complete algorithm process from raw data to final alarm decision;

[0081] The hardware basis of the central processing and alarm decision unit 40 can be configured according to the performance requirements and deployment environment of the system; in a specific embodiment, the central processing and alarm decision unit 40 is an embedded system based on a field programmable gate array (FPGA); the parallel processing capability and high-precision timing control capability of the FPGA make it suitable for real-time generation of complex excitation coding sequences and high-speed processing of massive timestamp data from the single-photon time correlation counting detection unit 30; in another embodiment, the central processing and alarm decision unit 40 is a system on chip (SoC) that integrates a microprocessor (CPU), a digital signal processor (DSP), and a programmable logic unit, capable of implementing control, communication, and complex algorithm operation on a single chip; in yet another embodiment, the central processing and alarm decision unit 40 can be a microcontroller (MCU) responsible for executing on-site control and preliminary data processing, and uploading key data to a cloud server through a wireless communication module (such as LoRa or NB-IoT), with the cloud server performing the final composite fingerprint analysis and decision model operation with more powerful computing resources;

[0082] The core functions of the central processing and alarm decision unit 40 are realized by performing the following series of steps:

[0083] S401: Perform control signal generation function, the central processing and alarm decision unit 40 generates a sequence of digital control instructions according to the internal set method steps, and sends them to the coherent timing coding excitation unit 20 through its I / O interface; these instructions accurately define the wavelength, start time, duration, and repetition frequency of the excitation pulse, used to realize single-channel timing excitation coding (TEC) and accurate time delay control in pump-probe mode;

[0084] S402: Perform high-speed data acquisition and pre-processing function, the central processing and alarm decision unit 40 receives time-resolved histogram data from the single photon time correlation counting detection unit 30; for the histogram data of each sensing channel, the central processing and alarm decision unit 40 performs a curve fitting algorithm, such as a nonlinear least squares method or a maximum likelihood estimation algorithm, to calculate the fluorescence lifetime of the channel ;

[0085] S403: Perform system initialization and reference fingerprint calibration function, after the system is deployed in a healthy plant environment, the central processing and alarm decision unit 40 repeatedly performs step S402 to acquire and calculate fluorescence lifetime data over a period of time, which is used to establish and store the reference lifetime vector in the healthy state and the reference cross-correlation matrix ;

[0086] S404: Perform periodic reconnaissance and anomaly triggering function, in the regular monitoring mode, the central processing and alarm decision unit 40 repeatedly performs steps S401 and S402 at a preset time interval to obtain real-time fluorescence lifetime vector and calculate its normalized deviation from the reference lifetime vector ; when the deviation exceeds the preset threshold, the system automatically switches from the reconnaissance mode to the in-depth analysis mode;

[0087] S405: Perform adaptive interrogation and static fingerprint construction function, in the in-depth analysis mode, the central processing and alarm decision unit 40 first generates optimized excitation codes based on the abnormal results of step S404 and controls the coherent timing coded excitation unit 20 to execute; then, the central processing and alarm decision unit 40 repeatedly performs step S402 at a high frequency in a short time to obtain a high-resolution fluorescence lifetime time series of the response channel ; based on the time series, calculate the real-time cross-correlation matrix ; the matrix elements are calculated by the following formula:

[0088] ;

[0089] where, represents the matrix element in the real-time cross-correlation matrix located in the row and the column; and are the fluorescence lifetime time series of the th and the th channel, respectively; and are the fluorescence lifetime time series of the Mean and standard deviation of the channel fluorescence lifetime time series; and are the mean and standard deviation of the channel fluorescence lifetime time series, respectively, at the current state Mean and standard deviation of the channel fluorescence lifetime time series; represent the mathematical expectation;

[0090] S406: Perform the cooperative response kinetics probing function, the central processing and alarm decision unit 40 analyzes the real-time cross-correlation matrix obtained in step S405 , selects the channel pair with the strongest correlation, and then controls the coherent time sequence coding excitation unit 20 to perform a pump-probe scan, at each set delay time , repeats step S402 to calculate the fluorescence lifetime of the probe channel, and finally constructs the cooperative response kinetics curve ;

[0091] S407: Perform the composite fingerprint fusion and model matching function, the central processing and alarm decision unit 40 fuses the real-time lifetime vector, the real-time cross-correlation matrix, and the cooperative response kinetics curve into a final composite fingerprint, the non-volatile memory inside the central processing and alarm decision unit 40 stores a standard fingerprint library and a pre-trained classification model; the specific implementation of the classification model can be a multi-modal fusion Transformer network, a tensor decomposition algorithm combined with a support vector machine classifier, or other machine learning models suitable for processing heterogeneous data, the central processing and alarm decision unit 40 inputs the real-time constructed composite fingerprint into the classification model for matching operation;

[0092] S408: Perform the alarm decision and output function, after the classification model operation, output the predicted label of the current plant state and the corresponding confidence; the central processing and alarm decision unit 40 judges the output according to the preset logic rule; if the predicted label is not healthy state and the confidence exceeds a preset alarm threshold, the central processing and alarm decision unit 40 triggers an alarm through its output interface, and the output alarm information can include binary alarm state, predicted disease type, stress level, and diagnosis confidence value.

[0093] Referring to the accompanying Figure 3 , Figure 3 is a flowchart of a plant disease and pest composite fingerprint early warning method according to an embodiment of the present application; the early warning method of the present application is coordinated and executed by the central processing and alarm decision unit 40 in the aforementioned early warning system, and specifically includes the following steps:

[0094] S501: Perform system initialization and baseline state calibration, deploying the early warning system of the present invention in a plant or its growing environment that has been confirmed to be in a healthy state; the central processing and alarm decision unit 40 controls the system to enter calibration mode. In this mode, the system uses a standardized broadband excitation sequence at fixed time intervals to calibrate all elements of the multi-element composite quantum dot sensing array unit 10. Each sensing channel performs continuous measurements;

[0095] S502: During the calibration process, the central processing and alarm decision unit 40 checks each sensor channel. A baseline lifetime vector is calculated and stored by statistically averaging a series of fluorescence lifetime values ​​measured over a sufficiently long period of time (e.g., several hours). The data structure of this vector is as follows:

[0096] ;

[0097] in, Representing the The average fluorescence lifetime of each sensing channel under healthy conditions;

[0098] S503: Simultaneously, the central processing and alarm decision unit 40 records the minute time fluctuation sequence of fluorescence lifetime for each channel and calculates the different channels. and channels The normalized covariance of the fluctuation signals is used to construct and store a benchmark cross-correlation matrix. ;

[0099] S504: Perform periodic reconnaissance scans and anomaly detection. After completing the baseline calibration, the system enters a low-power routine monitoring mode. The central processing and alarm decision unit 40 performs a rapid scan at a relatively long time interval (e.g., minutes). In this mode, the coherent timing coding excitation unit 20 uses a universal excitation code that covers the absorption spectrum of all channels.

[0100] S505: After each scan, the central processing and alarm decision unit 40 calculates a normalized deviation score. This score is used to quantify the degree of deviation of the current system state from the baseline state; the formula for calculating this score is:

[0101] ;

[0102] in, It is the first The real-time fluorescence lifetime of the channel measured in the current scan cycle; when this score Exceeding a preset exception trigger threshold When the central processing and alarm decision unit 40 determines that a potential threat event is detected, it immediately terminates the surveillance scanning mode and automatically enters the next step of the in-depth analysis process;

[0103] S506: Perform adaptive interrogation. The central processing and alarm decision unit 40 first identifies all the sensing channels whose response values exceed the normal range in step S505, forming an “activated” channel set ; then, the central processing and alarm decision unit 40 dynamically generates a new, targeted time sequence excitation code ; this code is optimized to allocate excitation energy and detection time resources to the channels in the set ; the corresponding control instructions are sent to the coherent time sequence excitation unit 20 to execute this customized excitation sequence;

[0104] S507: In the adaptive interrogation mode, the central processing and alarm decision unit 40 collects and records the fluorescence lifetime time series of the activated channels at a high time resolution, and calculates the real-time cross-correlation matrix based on this data ; the calculation method of this matrix is the same as step S503, but using the data under the current threat state. This matrix reveals the newly established or changed internal physicochemical correlation patterns between different sensing microenvironments under external disturbance.

[0105] S508: Integrate the real-time fluorescence lifetime vector and the real-time cross-correlation matrix obtained in the adaptive interrogation stage to form a static composite fingerprint ; its data structure is: ;

[0106] S509: Perform cooperative response dynamics detection. The central processing and alarm decision unit 40 analyzes the real-time cross-correlation matrix obtained in step S507 , and automatically selects one or more channel pairs with the largest absolute value of the correlation coefficient as the target of dynamics detection.

[0107] S510: For the selected target channel pair, the central processing and alarm decision unit 40 controls the coherent time sequence excitation unit 20 to switch to the pump-probe scanning mode; the central processing and alarm decision unit 40 instructs the coherent time sequence excitation unit 20 to first send an excitation pulse to channel at time , after a precisely controlled delay time , send an excitation pulse to channel at time ; the single-photon time correlation counting detection unit 30 synchronously collects and calculates the channel Fluorescence lifetime under this condition ; this process is repeated by stepping through different delay times values.

[0108] S511: The central processing and alarm decision unit 40 calculates and constructs the collaborative response kinetic curve according to the scanning results , which is defined as:

[0109] ;

[0110] wherein, is the probe channel fluorescence lifetime measured when the pump-probe delay is ; is the baseline fluorescence lifetime of the probe channel under healthy state; the shape of this curve directly reflects the dynamics of the chemical disturbance spreading from the channel microenvironment to the channel microenvironment; the collection of kinetic curves of all measured channel pairs constitutes the dynamic fingerprint ;

[0111] S512: Perform composite fingerprint fusion; the central processing and alarm decision unit 40 fuses the static fingerprint obtained in step S508 with the dynamic fingerprint obtained in step S511 in data structure, to construct a high-dimensional, multi-modal ultimate composite fingerprint ; the complete data structure is:

[0112] ;

[0113] This fingerprint is a heterogeneous data set containing system state vectors, static correlation matrices, and dynamic process curves.

[0114] S513: Perform intelligent decision-making; the central processing and alarm decision unit 40 passes the constructed ultimate composite fingerprint as input to a pre-trained and stored classification model ; the structure of this model is designed to be able to process the above heterogeneous data, such as a multi-modal deep learning model; after the model operates, it outputs a predicted label of the current plant state (e.g., healthy, disease A, disease B, etc.) and the confidence of the prediction .

[0115] S514: According to the output results of the classification model, the central processing and alarm decision unit 40 finally triggers the alarm according to the preset logic condition; the logic condition is: when the predicted label is not healthy, and its corresponding confidence Exceeding a preset alarm threshold When the system triggers an alarm, the output alarm signal can include specific disease type inferences, stress development stage assessments, and diagnostic confidence values, providing a basis for subsequent precision agriculture management decisions.

[0116] To further clarify the feasibility and wide applicability of the technical solution of the present invention, more specific implementation details are provided below for some of the core technical modules involved in the aforementioned method flow.

[0117] In one specific embodiment, the fluorescence decay histogram collected by the single-photon time-correlation counting detection unit 30 is fitted to extract the fluorescence lifetime. The algorithm can employ an iteratively reweighted nonlinear least squares method; the objective of this algorithm is to minimize the chi-square statistic as defined below. :

[0118] ;

[0119] in, These are the histogram channel indices, totaling [number missing]. One channel; It is the first The time corresponding to each channel; It is in the Photon count values ​​recorded in the experiment in each channel; It is in time At this point, the fluorescence decay model is based on the undetermined parameter set. Calculated theoretical photon count; set of parameters to be determined It contains the fluorescence lifetime that is to be determined. ;

[0120] In another embodiment, maximum likelihood estimation (MLE) can be used; this method is based on the statistical property that photon counts follow a Poisson distribution, and its goal is to find the parameter set that maximizes the probability of the observed experimental data. For a multi-exponential decay process, i.e., fluorescence intensity decay follows... The model described above, with the fitting algorithm configured to simultaneously solve for multiple lifetime components, is given. and its corresponding amplitude coefficient ;

[0121] For the cooperative response dynamics curve characteristic extraction, the central processing and alarm decision unit 40 can perform the following steps: first, a function fitting algorithm is applied to the discrete curve data points, and the function model can describe a typical excitation relaxation process, for example, a function composed of an exponential rise and an exponential decay term; then, a set of characteristic parameters with clear physical meaning is extracted from the fitted analytical function, for example: the peak amplitude of the curve, the time required to reach the peak (i.e. the peak position ), the full width at half maximum (FWHM) of the curve, and the time constants of the rising and falling edges; these extracted scalar parameters collectively form a low-dimensional vector as a compact representation of the dynamic fingerprint for subsequent model classification;

[0122] Alarm decision model offline training and online deployment process, including the following steps:

[0123] S601: Establish a standard fingerprint library, apply known, different types and different degrees of biological stress (e.g. inoculation of specific pathogenic bacteria) or non-biological stress (e.g. drought, salinity) to a large number of plant samples in a controlled laboratory;

[0124] S602: Use the early warning system of the present application to continuously monitor the samples from the healthy state throughout the entire cycle, and simultaneously execute the complete method process of the foregoing S501 to S512, to generate a complete ultimate composite fingerprint for the plant state at each time point;

[0125] S603: For each collected fingerprint , label its corresponding ground truth label (e.g. healthy, early downy mildew, medium powdery mildew, water stress, etc.), thereby forming a large-scale, labeled training dataset;

[0126] S604: Based on the dataset, the decision model is trained offline, and in an optional implementation, the decision model is a multi-modal fusion Transformer network; the network has different input branches for processing different modal data in the fingerprint: one branch processes the real-time lifetime vector through a fully connected layer; another branch extracts spatial correlation features in the real-time cross-correlation matrix through a two-dimensional convolution layer or a graph neural network layer; and another branch extracts the cooperative response dynamics curve the timing features in the time series; the features extracted from each branch are then fused and input into a Transformer encoder for deep feature interaction, and finally the prediction result is output by the classification head; in another optional implementation, a tensor decomposition algorithm combined with a support vector machine (SVM) strategy can be used, and first the ultimate composite fingerprint is represented as a high-order tensor, and then the core feature vector is extracted through tensor decomposition (such as PARAFAC decomposition), and then the feature vector is input into a standard support vector machine classifier;

[0127] S605: After the training is completed, the model parameters obtained by optimization are solidified into the program memory of the central processing and alarm decision unit 40; after being deployed on site, the system can perform online and real-time reasoning and decision-making.

[0128] Although embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A plant disease and pest intelligent early warning system based on quantum dot light emitting characteristics, characterized in that, The application relates to a multi-element composite quantum dot sensing array unit, wherein a plurality of quantum dots with different surface modifications are integrated on the substrate of the multi-element composite quantum dot sensing array unit, and are used for reacting with plant stress metabolites to generate different fluorescent lifetime responses and emitting fluorescence after being excited; the different surface modifications are at least one of the following: Constructing a pH response channel: modifying a ligand with a protonatable or deprotonatable group on the surface of a quantum dot, so that the fluorescent lifetime of the quantum dot responds to pH value changes; Constructing a redox response channel: fixing a molecule sensitive to active oxygen on the surface of a quantum dot, so that the fluorescent lifetime of the quantum dot responds to redox reactions; Constructing a volatile organic matter affinity response channel: modifying a functional group or a polymer film with physical adsorption or weak chemical action on volatile organic matters on the surface of a quantum dot, so that the fluorescent lifetime of the quantum dot responds to the adsorption of volatile organic matters; Constructing a specific signal molecule response channel: preparing a molecularly imprinted polymer film on the surface of a quantum dot, so that the fluorescent lifetime of the quantum dot responds to the capture of target signal molecules; A coherent time sequence coding excitation unit is used for emitting precisely timed pulsed light to the multi-element composite quantum dot sensing array unit; A single-photon time correlation counting detection unit is used for detecting fluorescence emitted by the multi-element composite quantum dot sensing array unit and generating a time-resolved data stream recording the arrival time of fluorescent photons relative to excitation pulses; A central processing and alarm decision unit is connected with the coherent time sequence coding excitation unit and the single-photon time correlation counting detection unit respectively, and is used for: Controlling the coherent time sequence coding excitation unit to emit precisely timed pulsed light; Receiving the time-resolved data stream from the single-photon time correlation counting detection unit; Processing the time-resolved data stream to calculate fluorescent lifetimes representing different fluorescent lifetime responses; Constructing a composite fingerprint based on the fluorescent lifetimes; the composite fingerprint comprises a real-time lifetime vector, a real-time cross-correlation matrix and a cooperative response kinetics curve; the central processing and alarm decision unit is specifically used for: constructing the real-time lifetime vector based on real-time fluorescent lifetimes of each sensing channel; calculating normalized covariances between different channels based on time sequences of fluorescent lifetimes of each sensing channel to construct the real-time cross-correlation matrix; constructing the cooperative response kinetics curve by performing pump-probe scanning to measure the change of the fluorescent lifetime of a probe channel with pump-probe delay time; performing periodic reconnaissance scanning and anomaly detection in a regular monitoring mode, and automatically switching from the regular monitoring mode to a deep analysis mode when a normalized deviation score calculated based on real-time fluorescent lifetimes exceeds a preset threshold, so as to perform construction of the composite fingerprint; Inputting the composite fingerprint into a preset classification model to perform matching operation to output a prediction label and a confidence degree; And making an alarm decision based on the prediction label and the confidence degree to output an alarm signal. The coherent time sequence coding excitation unit is controlled by the central processing and alarm decision unit and is used for generating at least two excitation modes:

2. The plant disease and pest intelligent early warning system based on quantum dot luminescence characteristics according to claim 1, characterized in that, ​ Single-channel time-sequenced encoding pulse sequence for reconnaissance scanning and adaptive interrogation; Time-accurately phase-locked dual-channel or multi-channel pulse sequence for performing cooperative response kinetics measurement.

3. The plant disease and pest intelligent early warning system based on quantum dot luminescence characteristics according to claim 1, characterized in that, The single-photon time-correlated counting detection unit comprises a time-correlated counting electronics module that receives a synchronization output signal from the coherent time-sequenced encoding excitation unit and a single-photon response pulse from a single-photon detector, and constructs a curve of fluorescence intensity decay over time by statistically time intervals of the single-photon response pulse relative to the synchronization output signal in multiple excitation events, and the central processing and alarm decision unit calculates the fluorescence lifetime by fitting the curve.

4. The plant disease and pest intelligent early warning system based on quantum dot luminescence characteristics according to claim 1, characterized in that, The classification model stored in the central processing and alarm decision unit is a multi-modal fusion Transformer network or a tensor decomposition algorithm combined with a support vector machine classifier.

5. The plant disease and pest intelligent early warning system based on quantum dot luminescence characteristics according to claim 1, characterized in that, When making an alarm decision, the central processing and alarm decision unit only outputs an alarm signal when the predicted label is not a healthy state and the confidence exceeds a preset alarm threshold.

6. The plant disease and pest intelligent early warning method based on the quantum dot light-emitting characteristics, applied to the system of any one of claims 1-5, characterized in that, The method comprises the following steps: S1. The central processing and alarm decision unit controls the coherent time-sequenced encoding excitation unit to emit precisely timed pulsed light to excite the multi-element composite quantum dot sensor array unit to generate fluorescence; S2. The single-photon time-correlated counting detection unit detects the fluorescence and generates a time-resolved data stream; S3. The central processing and alarm decision unit processes the time-resolved data stream to calculate the fluorescence lifetime and constructs a composite fingerprint fused from a real-time lifetime vector, a real-time cross-correlation matrix, and a cooperative response kinetics curve; S4. The composite fingerprint is input to a preset classification model for matching operation, and a predicted label and a confidence are output; S5. When the predicted label is not a healthy state and the confidence exceeds a preset alarm threshold, the central processing and alarm decision unit outputs an alarm signal.

Images