A high-precision liquid level positioning method based on peak illuminance method

Abstract

This invention relates to the field of intelligent sensor analysis technology, specifically a high-precision liquid surface positioning method based on the peak illuminance method. The method includes: driving a probe to descend, utilizing the refraction and spectral evolution effect of the light beam caused by different curvatures of the meniscus to generate a travel data stream and a photoelectric signal data stream; constructing a dynamic noise model for filtering optimization, and combining the travel data to construct a digital illuminance-travel curve of interface characteristics; extracting the illuminance peak fingerprint of the probe, identifying the characteristic peaks of the primary spectral path of the liquid surface refraction, and establishing interface anchor points; analyzing the fingerprint morphological parameters, and inverting the physical parameters of interfacial tension and wetting contact angle; generating an interface dynamic correction vector to compensate for physical deviations in liquid surface height, and outputting the liquid surface coordinates. This invention achieves coordinated characterization of the intrinsic physicochemical parameters and spatial distribution of the fluid by analyzing the photomechanical response of the phase interface, and eliminates interfacial deformation interference by utilizing property inversion, ensuring the integrity and accuracy of the original physicochemical characteristics of the sample.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensor analysis technology, specifically a high-precision liquid level positioning method based on the peak illuminance method. Background Technology

[0002] In the field of intelligent sensor analysis technology, which involves the characterization of the physicochemical properties of materials and the analysis of complex fluid compositions, accurate identification of the spatial evolution boundary of the interface between the probe unit and the liquid material to be tested is a prerequisite for high-throughput sample analysis. Existing technologies primarily employ optical refractive index transition detection and total internal reflection interface analysis to identify the physical topology of material interfaces. These methods define the spatial physical boundary of the interface by observing the refractive index gradient change of the probe beam as it passes through the interface, or by utilizing the evolution of the critical state of total internal reflection caused by the difference in the intrinsic optical properties of the media on both sides of the interface.

[0003] However, the response characteristics of the interfacial optical sensing signal are deeply parametrically coupled with the microscopic physical properties and optical properties of the material under test. Specifically, the meniscus curvature generated by the interfacial tension alters the refraction evolution path and wavefront distribution of the probe beam, making the illuminance response characteristics captured by the photosensitive element a complex function of spatial coordinates and the intrinsic refractive index and surface dynamic parameters of the material. This nonlinear coupling of photomechanical parameters leads to a shift in the characterization signal of the interfacial boundary as the material's physical properties fluctuate when analyzing materials with different physicochemical compositions. Furthermore, existing optical detection methods easily disrupt the thermodynamic equilibrium state of the meniscus in the measurement area when interfering with the interface, thereby altering the original physicochemical characteristics of the material interface in the measurement area.

[0004] Therefore, this invention proposes a high-precision liquid level positioning method based on the peak illumination method. Summary of the Invention

[0005] The purpose of this invention is to provide a high-precision liquid surface positioning method based on the peak illuminance method. By using the peak illuminance method, the peak illuminance fingerprint generated by the optical focusing effect of the liquid meniscus is used as the positioning reference. Furthermore, the morphological parameters of the peak illuminance fingerprint are analyzed to dynamically compensate and calibrate the positioning results. This solves the fundamental contradiction between the measurement principle of existing liquid surface positioning technology and the application purpose of analyzing unknown or variable materials, as mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A high-precision liquid level positioning method based on peak illuminance method includes: The probe is driven to descend at a set speed while emitting horizontal light to illuminate the meniscus of the liquid under test. The vertical travel coordinates of the probe are recorded by utilizing the refraction and spectral evolution effect of the light beam at different curvature positions of the meniscus, generating a travel data stream. The transmitted light intensity response characteristics after passing through the container medium are continuously captured by a smart sensor, generating a photoelectric signal data stream. Based on the baseline signal in the non-interface region of the photoelectric signal data stream, a dynamic noise model is constructed; filtering optimization is performed according to the digital filter inside the dynamic noise model to generate a filtered data stream; combined with the travel data stream, a digital illuminance-travel curve is constructed. Based on the digital illuminance-stroke curve, the illuminance peak fingerprint generated by the probe in the meniscus section is extracted, and the characteristic peak corresponding to the first spectral path of refraction on the liquid surface is identified. Interface anchor points are established based on the travel coordinates of characteristic peaks, and the morphological parameters of the illuminance peak fingerprint are analyzed. The interfacial tension and wetting contact angle physical parameters of the liquid under test are calculated by combining the capillary action model. A dynamic interface correction vector is generated based on the physical parameters of interfacial tension and wetting contact angle to compensate for physical deviations in liquid level height and output liquid level coordinates.

[0007] Preferably, the generation of the travel data stream specifically includes: a process feedback unit coupled to a stepper drive unit that drives the probe to move vertically; the process feedback unit continuously outputs digital pulse signals corresponding to the probe's interactive trajectory when the stepper drive unit is running, and performs real-time counting and accumulation of the digital pulse signals through a signal processing unit, converts the accumulated count value into sample axial scanning coordinates, and associates a timestamp with each coordinate data to generate a travel data stream.

[0008] Preferably, the generation of the photoelectric signal data stream specifically includes: capturing the transmitted photon flux modulated by the refractive energy level of the liquid interface under test using a smart sensor, and converting it into an analog signal reflecting the photophysical properties of the sample interface; performing low-noise gain analysis on the signal through a photosensitive detection unit, and discretizing and quantizing it using an analog-to-digital converter at a preset sampling period to generate a photoelectric signal data stream carrying liquid surface features; the smart sensor is configured as a detector sensitive to a preset wavelength according to the spectral response characteristics of the liquid under test, thereby enhancing the sensitivity of the identification of the interface spectral signal.

[0009] Preferably, the construction of the dynamic noise model specifically includes: extracting stable baseline signal segments corresponding to the photoelectric signal data stream in the probe's gaseous medium region and when it is far from the liquid surface, as background reference signals; performing statistical analysis on the stable baseline signal segments, calculating the mean, standard deviation, and statistical variance, and quantifying the amplitude characteristics of random noise; applying fast Fourier transform to the stable baseline signal segments to analyze spectral characteristics and identify periodic interference introduced by ambient light and signal transmission links to the material property analysis; and constructing a dynamic noise model based on the set of noise amplitude characteristics and periodic noise frequency characteristics.

[0010] Preferably, the construction of the digital illuminance-stroke curve specifically includes: constructing a two-dimensional data structure of double-precision floating-point type; spatiotemporally fusion of filtered data streams reflecting the optical properties of materials and stroke data streams reflecting spatial distribution based on timestamps; for data points with perfectly matched timestamps, storing the illuminance value and stroke value as a set of feature mapping pairs in the two-dimensional data structure; for data points with mismatched timestamps, using a resampling algorithm based on cubic spline interpolation to calculate the corresponding illuminance value and stroke value at the timestamp, and generating the digital illuminance-stroke curve.

[0011] Preferably, the extraction of the illuminance peak fingerprint generated by the probe in the meniscus segment specifically includes: performing first and second order numerical differentiation on the digital illuminance-stroke curve; searching for zero-crossing points where the first derivative value changes from positive to negative, and determining whether the second derivative value corresponding to the zero-crossing point is less than a negative threshold determined based on the standard deviation of the baseline signal, thereby determining the illuminance peak point of the illuminance peak fingerprint; and, based on the response gradient threshold of the baseline signal, searching forward and backward from the illuminance peak point to positions where the absolute value of the curve slope is less than the response gradient threshold, thereby determining the start and end points of the illuminance peak fingerprint and extracting the illuminance peak fingerprint.

[0012] Preferably, the step of calculating the interfacial tension and wetting contact angle physical parameters of the liquid under test by combining the capillary action model specifically includes: extracting the response bandwidth and integral flux of the illuminance peak fingerprint, and analyzing the radius of curvature distribution of the meniscus at the container wall by combining the optical refractive index constant; substituting the radius of curvature distribution into the Young-Laplace model to calculate the interfacial pressure difference, and inverting and deriving the interfacial tension parameters of the liquid under test by combining the hydrostatic equilibrium conditions; and obtaining the physical parameters of the wetting contact angle of the liquid to the container wall and the interfacial tension of the liquid under test by analyzing the response asymmetry coefficient of the illuminance peak fingerprint.

[0013] Preferably, the generation of the interface dynamic correction vector specifically includes: taking the interface tension parameter and the wetting contact angle physical parameter as input variables, and the container's geometric radius as a structural parameter, and substituting them into the capillary action model; the capillary action model calculates the capillary rise height and fall height formed by the liquid surface at the container wall by establishing an interface morphology evolution equation under the equilibrium relationship of liquid surface tension, wettability, and gravity field; the output layer outputs the capillary rise height and fall height, and establishes them as the axial position correction amount for the interface anchor point, generating the interface dynamic correction vector; the capillary action model calculates and outputs the measurement deviation correction operator caused by the intrinsic properties of the material by establishing a mechanical evolution equation of liquid surface tension, wettability, and gravity field.

[0014] Preferably, the output liquid surface coordinates specifically include: using the interface anchor point as the reference coordinate, performing vector arithmetic operations on the interface dynamic correction vector as the compensation operator to obtain the corrected center space coordinates; and fusing the corrected center travel coordinates with the index position of the probe component in the planar scanning array to output the absolute boundary coordinates of the liquid under test in three-dimensional space.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention achieves high versatility and fundamental reliability by combining three processing methods: identifying illuminance peak fingerprints independent of the physicochemical properties of the analyte as the core localization basis, employing a non-contact detection path based on optical abrupt changes, and using an intelligent sensor to perform adaptive signal processing to ensure stable fingerprint extraction. This completely resolves the fundamental contradiction between measurement principles and the application purpose of analyzing unknown materials. Unlike traditional capacitive or pressure-based detection methods that require probe immersion or significant contact with liquid, this invention utilizes the meniscus optical lens effect generated instantaneously at the gas-liquid interface as a trigger signal. This non-contact localization method significantly reduces the amount of probe residue carried by physical contact, effectively avoiding the risk of cross-contamination. Furthermore, the intelligent sensor's dynamic modeling and adaptive filtering of environmental noise before each measurement ensures that this characteristic signal can be stably captured with a high signal-to-noise ratio even under complex or changing conditions. Therefore, this invention provides a truly universal and fundamentally reliable sample localization solution for various intelligent sensor analysis platforms.

[0016] 2. This invention achieves unprecedented positioning accuracy through a progressive processing method: first, decoding the location of the illuminance peak fingerprint to establish a highly repeatable interface anchor point; then, deeply analyzing the morphological parameters of the illuminance peak fingerprint; and finally, calculating the interface dynamic correction vector to compensate for optical refraction deviations. This method meets the precision handling requirements of micro- and nano-scale samples in automated analysis. The innovation of this invention lies in its coarse positioning + fine calibration second-order solution strategy. The illuminance peak signal generated by the meniscus lens effect has extremely sharp and distinct characteristics, providing a highly high-resolution and repeatable reference anchor point for initial positioning. Building upon this, this invention goes beyond simple positioning and further mines the illuminance peak fingerprint as an information-rich data volume. By analyzing its response amplitude, response bandwidth, and other morphological parameters, the differences in meniscus curvature caused by different liquid surface tensions can be deduced, and the interface dynamic correction vector can be calculated accordingly to provide real-time compensation at the micrometer level for the initial positioning results. This intelligent upgrade from identification and positioning to measurement and calibration elevates the accuracy of this method to a new order of magnitude, ensuring extreme precision in pipetting operations when processing trace samples. 3. This invention achieves high intelligence and noise resistance through a three-pronged approach: a smart sensor establishes a dynamic noise model in real time based on the baseline signal; the internal filter parameters are adaptively optimized based on the dynamic noise model to achieve the best signal-to-noise ratio; and a non-contact measurement principle is applied. This ensures that the positioning process is unaffected by environmental factors while guaranteeing zero sample contamination required for high-sensitivity material analysis. The intelligence of this invention is reflected in its proactive adaptability to the measurement environment. At the initial stage of each positioning task, the smart sensor learns the current background noise characteristics, such as fluctuations in ambient light and random noise from the circuit itself, and dynamically adjusts its digital filtering algorithm accordingly. This ensures that the core illuminance peak signal can be clearly extracted with the highest signal-to-noise ratio under any operating conditions, greatly enhancing the stability and robustness of the method. More importantly, this critical contact measurement method based on optical fingerprinting, compared to the total internal reflection method which requires an optical prism to press the liquid surface or the capacitance method which requires a probe to insert into the liquid surface, minimizes physical noise on the sample itself. Since the positioning process only involves changes in the optical properties of the microscopic interface and does not involve macroscopic replacement or compression of the sample by the probe body, it significantly reduces the risk of cross-contamination inherent in existing deep-contact technologies from an engineering application perspective. This is crucial for applications with stringent requirements for sample purity, such as PCR amplification and high-sensitivity immunoassay, ensuring the authenticity and validity of the final analytical results. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the overall method of a high-precision liquid level positioning method based on the peak illuminance method proposed in this invention application. Figure 2 This is a flowchart of an adaptive signal processing and curve construction method for a high-precision liquid level positioning method based on the peak illuminance method proposed in this invention application. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention provides a high-precision liquid level positioning method based on the peak illuminance method, the technical solution of which is as follows: A high-precision liquid level positioning method based on peak illuminance method, reference Figure 1 The specific implementation steps of the method proposed in this invention include: S1. Drive the probe to descend at a set speed while emitting horizontal light to illuminate the meniscus of the liquid to be tested. Utilize the refraction and spectral evolution effect of the light beam at different curvature positions of the meniscus to record the vertical travel coordinates of the probe and generate a travel data stream. The intelligent sensor continuously captures the transmitted light intensity response characteristics after passing through the container medium and generates a photoelectric signal data stream.

[0020] S2. Based on the baseline signal in the non-interface region of the photoelectric signal data stream, construct a dynamic noise model; perform filtering optimization according to the digital filter inside the dynamic noise model to generate a filtered data stream; combine the travel data stream to construct a digital illuminance-travel curve.

[0021] S3. Based on the digital illuminance-stroke curve, extract the illuminance peak fingerprint generated by the probe in the meniscus section and identify the characteristic peak corresponding to the first spectral path of the liquid surface refraction.

[0022] S4. Based on the travel coordinates of the characteristic peak, establish the interface anchor point, analyze the morphological parameters of the illuminance peak fingerprint, and combine the capillary action model to invert and calculate the physical parameters of the interfacial tension and wetting contact angle of the liquid under test.

[0023] S5. Generate a dynamic interface correction vector based on the physical parameters of interfacial tension and wetting contact angle to compensate for physical deviations in liquid level height and output liquid level coordinates.

[0024] Example 1: This embodiment provides a specific application of a high-precision liquid level positioning method based on the peak illuminance method, which is applicable to transparent or semi-transparent liquids.

[0025] Furthermore, the probe is driven to descend at a set speed while simultaneously emitting horizontal light to illuminate the meniscus of the liquid under test. Utilizing the refraction and spectral evolution effect of the light beam at different curvature positions of the meniscus, the vertical travel coordinates of the probe are recorded, generating a travel data stream. A smart sensor continuously captures the transmitted light intensity response characteristics after passing through the container medium, generating a photoelectric signal data stream. Corresponding to step S1 above, the specific process is as follows: A process feedback unit is adopted, which is coupled to the stepper drive unit that drives the probe to move vertically. When the stepper drive unit is running, the process feedback unit continuously outputs digital pulse signals corresponding to the probe's interactive trajectory. The signal processing unit counts and accumulates the digital pulse signals in real time, converts the accumulated count value into the sample axial scanning coordinate, and associates a timestamp with each coordinate data to generate a travel data stream.

[0026] The driving probe descends vertically at a set speed while simultaneously emitting a horizontally collimated and shaped beam with a diameter of 0.5-1 mm, illuminating the meniscus of the liquid under test. A smart sensor located along the beam's exit path continuously captures the deflected light rays refracted by the meniscus in a dark field configuration. A differential orthogonal process feedback unit with a resolution of 2000 lines per revolution is coupled to a high-precision stepper drive unit that moves vertically along the Z-axis of the driving probe. During the operation of the stepper drive unit, the differential orthogonal process feedback unit continuously outputs A / B phase and out-of-phase differential orthogonal pulse signals corresponding to minute angular travels. These differential orthogonal pulse signals are input to a programmable gate array (FPGA) or a dedicated motion control chip, where they are decoded and accumulated in real time by a built-in quadruple-frequency pulse counting circuit. This allows each physical scribe line to be resolved into four logical states, thus quadrupling the resolution. The accumulated count value is then converted in real time into absolute vertical position coordinate data representing the probe's sub-micron resolution, based on the lead parameters of the precision ball screw. To ensure synchronization with the optical signal, a crystal oscillator with a frequency of 50MHz or higher inside the FPGA is used as the master clock to associate a high-precision, nanosecond-resolution timestamp with each newly generated coordinate data, thereby generating the travel data stream.

[0027] By employing a high-resolution process feedback unit and real-time pulse counting, the generated travel data stream is ensured to have extremely high spatial resolution and time synchronization accuracy, providing a reliable position reference for the subsequent construction of accurate digital illuminance-travel curves.

[0028] The transmitted photon flux modulated by the refractive energy level of the liquid interface is captured by an intelligent sensor and converted into an analog signal reflecting the photophysical properties of the sample interface. The signal is analyzed by a photosensitive detection unit with low noise gain and discretized and quantized by an analog-to-digital converter at a preset sampling period to generate a photoelectric signal data stream carrying liquid surface features. The intelligent sensor is configured as a detector sensitive to a preset wavelength according to the spectral response characteristics of the liquid to be tested, thereby enhancing the sensitivity of the identification of the interface spectral signal.

[0029] The captured photon flux is converted into an analog photocurrent ranging from picoamperes to nanoamperes using a PIN photodiode optimized for the 850nm near-infrared band, exhibiting high quantum efficiency and nanosecond-level fast response. This analog photocurrent is then input to the first stage of a photosensitive detection unit composed of an ultra-low bias current junction field-effect transistor operational amplifier for low-noise amplification, and converted into a millivolt-level analog voltage signal. The analog voltage signal then undergoes a second-stage amplification and filtering circuit before being fed into a high-speed successive approximation analog-to-digital converter (ADC) with 16 bits or higher quantization bits and a sampling frequency stable at 50kHz. The ADC periodically discretizes the analog voltage signal, generating a digital photoelectric signal data stream with nanosecond-level resolution timestamps synchronized by the same master clock, capable of reflecting subtle changes in light intensity.

[0030] By employing a combination of high-sensitivity photodiodes, low-noise amplifier circuits, and high-speed analog-to-digital converters, the conversion process from weak optical signals to digital data streams is ensured to have high fidelity, high signal-to-noise ratio, and fast response characteristics, enabling the precise capture of instantaneously changing optical events.

[0031] Furthermore, based on the baseline signal in the non-interface region of the photoelectric signal data stream, a dynamic noise model is constructed; filtering optimization is performed according to the digital filter within the dynamic noise model to generate a filtered data stream; combined with the travel data stream, a digital illuminance-travel curve is constructed, corresponding to step S2 above, with reference to... Figure 2 The specific process is as follows: The stable baseline signal segments corresponding to the photoelectric signal data stream in the probe's gaseous medium region and far from the liquid surface are extracted as background reference signals. Statistical analysis is performed on the stable baseline signal segments to calculate the mean, standard deviation, and statistical variance, quantifying the amplitude characteristics of random noise. Fast Fourier Transform is applied to the stable baseline signal segments to analyze spectral characteristics and identify periodic interferences introduced by ambient light and signal transmission links to the material property analysis. Based on the combination of noise amplitude characteristics and periodic noise frequency characteristics, a dynamic noise model is constructed.

[0032] In the initial stage of the probe's interface scanning, within a preset reference distance where there is no interaction between the probe's optical path and the interface of the liquid under test, the embedded microprocessor inside the smart sensor extracts a stable photoelectric signal data stream of 1024 data points as a baseline signal sample. The processor performs statistical analysis on the baseline signal sample sequence, calculating the statistical variance of the baseline signal sample sequence to quantitatively characterize the average power level of random Gaussian white noise. Simultaneously, a fast Fourier transform algorithm with a Hanning window is applied to the baseline signal sample sequence to analyze its power spectral density. By performing a peak search algorithm on the power spectrum, periodic noise components with significant power peaks introduced by external ambient light (50Hz / 60Hz power frequency noise and harmonics) or internal circuitry (switching power supply noise) are identified. The peak search algorithm determines the center frequency, bandwidth, and amplitude of the periodic noise by locating the energy concentration points in the power spectrum. The dynamic noise model is a parameterized set of the calculated random noise power and the center frequency, bandwidth, and amplitude of all periodic noises. After establishing the dynamic noise model and calculating the optimized digital filter parameters, the processor inside the smart sensor must lock the cutoff frequency and order of the digital filter before the probe enters the preset safe distance for interaction with the liquid surface (more than 10 mm from the expected liquid surface). Thereafter, the filter coefficients remain constant throughout the entire vertical scan process to ensure uniform and constant phase and amplitude response characteristics for filtering the digital illuminance-stroke curve, thereby guaranteeing the accuracy and consistency of the extraction of peak illuminance fingerprint morphological parameters.

[0033] By performing both statistical and spectral analysis on the baseline signal, the noise characteristics under the current measurement environment can be comprehensively and quantitatively characterized, providing a precise and dynamic mathematical basis for subsequent adaptive filtering.

[0034] A two-dimensional data structure of double-precision floating-point type is constructed. Based on timestamps, the filtered data stream reflecting the optical properties of the material and the travel data stream reflecting the spatial distribution are spatiotemporally fused. For data points with perfectly matched timestamps, the illuminance value and travel value are stored as a set of feature mapping pairs in the two-dimensional data structure. For data points with mismatched timestamps, a resampling algorithm based on cubic spline interpolation is used to calculate the corresponding illuminance value and travel value at the timestamp, generating the digital illuminance-travel curve.

[0035] A two-dimensional floating-point data array structure is established, and spatiotemporal synchronization is performed between the run-length data stream and the filtered data stream based on the timestamps associated with their respective data points. Since the two data streams may have microsecond-level sampling time discrepancies, the processor employs a resampling algorithm based on cubic spline interpolation. To ensure the smoothness of the reconstructed curve at the endpoints of the run-length axis and minimize oscillations at the boundaries, the interpolation operation must use natural boundary conditions, i.e., setting the spline function S at the starting point of the data sequence. and termination point The second derivative is strictly zero. Using the timestamp sequence of the travel data stream as the reference time axis, for each reference time point, the two adjacent timestamps of each reference time point are found in the photoelectric signal data stream, and the interpolated illuminance value corresponding to the reference time point is calculated using a cubic spline function. Through this process, a resampled photoelectric signal data stream that is completely synchronized with the travel data stream is generated. Finally, these precisely matched illuminance values ​​and travel values ​​are stored as a set of (X, Y) feature mapping pairs in a two-dimensional data array to generate a high-resolution, time-distortion-free digital illuminance-travel curve. The digital illuminance-travel curve is implemented in the software as a two-dimensional data array structure of double-precision floating-point type. The two-dimensional data array structure is an ordered sequence aligned with timestamps as the reference, where each element is a feature mapping pair (Z, L), and the Z-axis corresponds to the absolute vertical coordinate travel value of the probe (in micrometers). The L-axis corresponds to the optimized filtered illuminance value (in lux or normalized relative illuminance). The two-dimensional data array structure is designed to support efficient random access and numerical differential computation.

[0036] By using a timestamp-based spatiotemporal matching and interpolation algorithm, the minor asynchronous issues that may exist between two independent data streams are resolved, ensuring that the final constructed digital illuminance-travel curve can truly and accurately reflect the functional relationship between light intensity and physical location.

[0037] The core inspiration of this invention lies in an adaptive signal processing and high-fidelity reconstruction method based on real-time environment awareness. This method aims to proactively identify and quantify dynamic noise sources in the measurement process, and adaptively optimize the signal processing chain accordingly.

[0038] At the initial stage of each localization task, the processor executes an online noise characterization program. This program captures a baseline signal and performs joint time-domain and frequency-domain analysis. The time-domain analysis quantifies the power of broadband random noise by calculating the sample variance; the frequency-domain analysis identifies the center frequency, bandwidth, and power of narrowband periodic noise by performing a Fast Fourier Transform and applying a peak detection algorithm. The analysis results are structured into a dynamic noise model containing multiple noise parameters. Subsequently, a parameter decision algorithm is activated, adaptively configuring the parameters of the internal multi-stage digital filters based on the specific parameters of the dynamic noise model. If the model identifies significant 50Hz power frequency noise, it automatically sets the center frequency of the IIR notch filter to 50Hz; simultaneously, it adjusts the cutoff frequency and order of the Butterworth low-pass filter according to the magnitude of the random noise power. The parameter decision algorithm uses a signal-to-noise ratio-based approach. The piecewise mapping rule is used to determine the order N of the Butterworth low-pass filter. This piecewise mapping rule aims to balance noise suppression and signal fidelity. when We set N=4th order to preserve signal details and reduce phase distortion.

[0039] when Setting N=6 provides adequate noise suppression.

[0040] when Setting N=8 provides the maximum noise suppression capability.

[0041] The selection of the order N, ranging from 4 to 8, is based on the preset requirement for the desired signal attenuation slope and is related to the average power of random noise. They are positively correlated.

[0042] Cutoff frequency The determination is based on the signal-to-noise ratio. Real-time estimation, The estimation formula is as follows ,in This represents the power of the peak illuminance signal. To determine the power, the following lookup table or piecewise linear function is used, provided that the dominant frequency component (below 1kHz) of the peak illuminance signal remains unaffected. : when When, set 5kHz (high suppression); when When, set 10kHz; when When, set 15kHz (low suppression to preserve signal details).

[0043] For periodic noise, if the model identifies 50Hz or 60Hz noise with an amplitude exceeding 5mV, it automatically activates a second-order infinite impulse response digital notch filter and sets the center frequency of the second-order infinite impulse response digital notch filter. The identified power frequency is precisely matched, and the quality factor Q is fixed at 20 to ensure high attenuation in narrowband.

[0044] This invention, through a closed-loop adaptive mechanism, ensures that the final digital illuminance-travel curve maintains a high signal-to-noise ratio and high fidelity regardless of whether the laboratory is bright or dim, or how the circuit status fluctuates, thus laying a solid and reliable foundation for subsequent accurate fingerprint recognition.

[0045] Furthermore, based on the digital illuminance-stroke curve, the illuminance peak fingerprint generated by the probe in the meniscus segment is extracted, and the characteristic peak corresponding to the first spectral path of refraction at the liquid surface is identified. Corresponding to step S3 above, the specific process is as follows: Perform first and second-order numerical differentiation on the digital illuminance-stroke curve; search for zero-crossing points where the first derivative value changes from positive to negative, and determine whether the second derivative value corresponding to the zero-crossing point is less than the negative threshold determined based on the standard deviation of the baseline signal, thereby determining the illuminance peak point of the illuminance peak fingerprint; based on the response gradient threshold of the baseline signal, search forward and backward from the illuminance peak point to the position where the absolute value of the curve slope is less than the response gradient threshold, thereby determining the start and end points of the illuminance peak fingerprint, and extracting the illuminance peak fingerprint.

[0046] The Savitzky-Gore smoothing differential filter is applied to the digital illuminance-stroke curve generated in the previous step. This filter can accurately calculate the first and second derivatives while smoothing the data. By searching for zero-crossing points in the first derivative sequence where the value changes from positive to negative, and verifying that the corresponding second derivative value is less than a preset negative threshold, the core of the illuminance peak fingerprint—the illuminance peak point—is located with sub-pixel precision. Subsequently, based on a preset response gradient threshold dynamically determined by the baseline noise level, a search is performed along the positive and negative directions of the stroke axis from the illuminance peak point until the absolute value of the curve slope is consistently less than the response gradient threshold. This determines the start and end points of the illuminance peak fingerprint. Thus, the complete illuminance peak fingerprint, containing all key morphological features, is completely separated and extracted from the entire data curve and stored as an independent data segment.

[0047] By combining first-order and second-order derivative analysis, a mathematically robust peak point localization method is provided, effectively avoiding misjudgments caused by signal spikes or gentle peaks, and ensuring the accuracy and uniqueness of core feature localization.

[0048] The fundamental inspiration of this invention lies in an optical fingerprinting method that is completely decoupled from the physicochemical properties of the material under test, bypassing all uncertain measurement variables related to the intrinsic properties of the material under test, and instead using more universal and stable pure physical optical phenomena as the reference marker for positioning.

[0049] When a collimated beam of light passes through the container wall and the internal medium at a fixed angle of incidence, its propagation follows Snell's law of refraction. At the gas-liquid interface, due to the significant difference in refractive indices between air and liquid, this invention captures the abrupt illuminance peak caused by the formation of the meniscus, thereby achieving sub-micron level precision in capturing the liquid level height. When the central axis of the probe beam sweeps across this structure, light focusing or divergence inevitably occurs, resulting in an instantaneous, rapid, and distinctive change in illuminance at the light receiver, typically manifested as an illuminance peak. The generation of this illuminance peak fingerprint depends solely on the presence of the gas-liquid interface, the difference in refractive indices between the two media, and the fundamental physical laws of light propagation, and is insensitive to macroscopic physicochemical properties such as the liquid's chemical composition, conductivity, color, and viscosity. In contrast, the capacitance method, based on Gauss's law, directly correlates the signal with the dielectric constant; the pressure method, based on fluid dynamics, couples the signal with parameters such as viscosity and density.

[0050] By stably identifying and precisely locating the characteristic optical fingerprint of the liquid surface, the determination of this positioning benchmark is unaffected by macroscopic physicochemical properties such as sample conductivity and color. This enables the present invention to become a truly reliable and universal standard in the field of intelligent sensor analysis that does not require repeated calibration for different samples.

[0051] Furthermore, interface anchor points are established based on the travel coordinates of characteristic peaks, and the morphological parameters of the illuminance peak fingerprint are analyzed. Combined with the capillary action model, the interfacial tension and wetting contact angle physical parameters of the liquid under test are calculated. Corresponding to step S4 above, the specific process is as follows: The response bandwidth and integral flux of the illuminance peak fingerprint are extracted, and the radius of curvature distribution of the meniscus at the container wall is analyzed in conjunction with the optical refractive index constant. The radius of curvature distribution is substituted into the Young-Laplace model to calculate the interfacial pressure difference, and the interfacial tension parameters of the liquid under test are derived by combining the hydrostatic equilibrium conditions. By analyzing the response asymmetry coefficient of the illuminance peak fingerprint, the physical parameters of the wetting contact angle of the liquid to the container wall and the interfacial tension of the liquid under test are obtained.

[0052] In the two-dimensional data array of the digital illuminance-travel curve, the illuminance peak point identified in the previous step is directly retrieved and located by index. The travel coordinate value corresponding to the illuminance peak point without any compensation is read, and the illuminance peak point is designated as the interface anchor point as the initial reference for subsequent fine positioning.

[0053] By directly mapping the most significant and repeatable peak points in the signal to physical locations, a stable and reliable initial positioning reference is provided for subsequent precision calibration.

[0054] Numerical calculations are performed on the extracted complete peak illuminance fingerprint data segment to obtain its morphological parameters, specifically including: response amplitude, which is defined as the difference between the illuminance value at the peak illuminance point and the average illuminance value of the baseline signal; response bandwidth, which is obtained by linear interpolation at half the response amplitude to find two precise abscissas and calculating the difference between the abscissas; integral flux, which is numerically integrated by applying the trapezoidal rule to the integral of the region enclosed between the signal envelope and the baseline; and response asymmetry coefficient, which is defined as the value of the right half-bandwidth of the peak point divided by the left half-bandwidth.

[0055] By performing multi-dimensional quantitative analysis on characteristic signals, the qualitative signal "shape" is transformed into a set of quantitative mathematical parameters that can be used for precise calculation. This is a key step in realizing the transformation from "positioning" to "measurement".

[0056] The calculated response amplitude, response bandwidth, and other morphological parameters are used as feature vectors and input into a multidimensional lookup table pre-established by calibrating a series of standard liquids with known gradient surface tensions. This multidimensional lookup table stores the mapping relationship between different combinations of morphological parameters and corresponding stroke correction values. Through multilinear interpolation operations performed on the multidimensional grid of the lookup table, a stroke correction value is output that accurately compensates for the optical refraction deviation between the peak point and the actual liquid surface; this stroke correction value is the interface dynamic correction vector. The multidimensional lookup table is a four-dimensional data structure containing four input dimensions: normalized response amplitude... Normalized response bandwidth Normalized integral flux and response asymmetry coefficients The output is the travel correction value. (Unit: micrometers) The extracted morphological parameters need to be normalized to ensure the model's robustness to changes in optical gain (such as light source intensity drift). Normalized response amplitude. The calculation formula is: ,in, The original illuminance value at the peak illuminance point. The baseline signal average illuminance value. This represents the maximum illuminance response amplitude obtained during the calibration phase on the standard liquid (i.e., the global maximum value of the calibration liquid set). Normalized response bandwidth. and normalized integral flux The calculation formula is: , ;in, and These represent the global maximum values ​​of the response bandwidth and integral flux obtained during the calibration phase by measuring all standard liquids, respectively. This linear scaling ensures that all morphological parameters are mapped to the range [0, 1]. Response asymmetry coefficients As a ratio, its value ranges from 0 to 2. In this model, no additional normalization is required; it is determined by the right half-bandwidth. Divide by the left half bandwidth It can be obtained Left half bandwidth and right half bandwidth The calculations are all performed by linearly interpolating two adjacent data points at the half-response amplitude positions on both sides of the illuminance peak point to determine the precise travel coordinate Z. In calculating the left half-bandwidth... and right half bandwidth At that time, half-bandwidth illuminance value Uniform numerical values ​​must be used: ,in It is the difference between the illuminance value at the peak illuminance point and the average illuminance value of the baseline signal. This uniform interpolation is used for both sides. Value, to ensure Objectivity of measurement. Regarding illuminance values. The interval Its corresponding illuminance value is Precise half-response amplitude travel coordinates Calculated using linear interpolation formula: ,in, and These are the travel coordinates of adjacent data points. and This corresponds to the illuminance value. This sub-pixel-level calculation ensures that the accuracy of the half-bandwidth measurement reaches the sub-micron level. A response asymmetry coefficient is established. With wetting contact angle Empirical mapping function , where the coefficient , C is obtained through calibration experiments on standard liquids.

[0057] The capillary action model includes an input layer, a physical logic layer, and an output layer, and performs property inversion and stroke compensation through the hierarchical logic described below. In the input layer, the processor first performs normalization processing on the received response amplitude, response bandwidth, and integral flux. This process converts the original signal into a dimensionless feature vector distributed in the interval between 0 and 1 by dividing the real-time extracted values ​​by the global maximum value of the same parameters obtained in the calibration phase. This step eliminates the influence of light intensity drift on subsequent physical parameter calculations through linear scaling.

[0058] In the physical logic layer, the processor invokes a pre-established four-dimensional multidimensional lookup table, which serves as a discretization mapping of the Young-Laplace mechanical equilibrium equations. The computation process consists of two stages: The first stage involves the inverse calculation of interfacial tension. Since the interfacial tension of the liquid directly determines the curvature of the meniscus formed at the container wall, and the meniscus curvature determines the focusing shape of the light beam, the processor performs multilinear interpolation within the physical mapping space of the lookup table based on the normalized response bandwidth and response amplitude. By analyzing the physical property weights of adjacent grid points, the interfacial tension coefficients required to maintain the current meniscus morphology are inversely resolved.

[0059] The second stage involves the inversion calculation of the physical parameters of the wetting contact angle. The wettability of the liquid to the container wall determines the degree of asymmetry of the meniscus in the vertical direction. The processor extracts the response asymmetry coefficients and integral flux, identifies the difference in waveform broadening on both sides of the travel axis, and thus maps and calculates the physical parameters of the wetting contact angle between the liquid and the container wall solid-liquid interface.

[0060] In the output layer, the processor uses the calculated interfacial tension and wetting contact angle as core physical variables, combined with the known container geometric radius, and substitutes them into the capillary rise and fall equilibrium equation for final solution. This equation describes the evolution of the equilibrium between the upward pulling force of surface tension and the downward attraction of the liquid column's gravity. By multiplying the interfacial tension by the cosine of the contact angle and eliminating the influence of gravity and container radius, the specific values ​​of the capillary rise and fall heights caused by the liquid surface tension at the container wall are calculated.

[0061] Finally, the processor establishes the calculated capillary rise and fall heights as interface dynamic correction vectors, and performs axial coordinate compensation on the interface anchor points determined in the initial stage through vector arithmetic operations. This step fundamentally eliminates the meniscus height deviation caused by the differences in the physical properties of different liquids, achieving high-precision output of the true physical coordinates of the liquid interface.

[0062] The multidimensional lookup table is established by calibrating at least five standard liquids with different surface tension gradients (such as pure water, aqueous ethanol solution, aqueous glycerol solution, etc.). This invention establishes a mapping model based on the following physical optical convergence law: the surface tension coefficient of a liquid is positively correlated with the curvature of the meniscus formed at the probe tip. Specifically, the greater the surface tension, the smaller the contact angle and radius of curvature of the meniscus, and the shorter the focal length of the meniscus as an equivalent liquid lens, resulting in a more concentrated luminous flux density distribution of the focused light spot on the sensor target surface. This manifests morphologically as an increase in the illuminance peak point and a narrower response bandwidth. Conversely, low surface tension or high viscosity liquids result in a flatter meniscus, producing a lower and wider illuminance peak. Based on this law, the calibration process is as follows: First, the precise optical refraction deviation between the illuminance peak point position and the actual liquid surface position of each standard liquid is independently measured using a high-precision metrology device (such as a laser confocal microscope). Simultaneously, the corresponding morphological parameters are calculated using the method of this invention. (The following text appears to be incomplete and requires further context: "will ( , , , , The data pairs are used as grid point data in a lookup table. When facing an unknown liquid to be tested, its morphological parameters are first calculated, and these parameter points are usually located inside the hypercube formed by the grid points of the lookup table. To output an accurate interface dynamic correction vector, this invention does not use a simple nearest neighbor search, but instead performs multilinear interpolation. Specifically, the multilinear interpolation operation identifies the parameters surrounding the test parameter points. For each adjacent calibration grid point, a weighting coefficient is calculated based on the Euclidean distance between the parameter point to be measured and each grid point in four-dimensional space. The final travel correction value is then calculated by weighted averaging. Multilinear interpolation ensures that even if the physical properties of the sample being tested fall between those of standard liquids (such as serum or buffer), smooth and accurate correction results can be obtained based on the continuity of physical laws.

[0063] By establishing a mapping relationship between morphological parameters and optical refraction deviation, and cleverly utilizing the microscopic physical information of the sample contained in the signal morphology, intelligent and dynamic compensation for the initial positioning results can be achieved. This is the core of the invention's achievement of ultra-high precision.

[0064] A second-order solution and dynamic precision calibration method based on feature signal morphological analysis breaks through the first-order thinking of traditional positioning technology that only uses whether the signal amplitude reaches a certain threshold. Instead, it treats the feature signal itself as an information-rich data body and achieves second-order precision correction of the initial positioning result through in-depth analysis of its morphology.

[0065] The first-order solution is performed, which uses the positional information of the illuminance peak fingerprint on the travel axis to obtain the absolute position of the liquid surface. This step directly decodes the position coordinates of the peak point into interface anchor points. The second-order solution is performed, which uses the shape information of the illuminance peak fingerprint to correct the result of the first-order solution. This process is based on physical understanding: although the appearance of the illuminance peak fingerprint is not sensitive to changes in refractive index, the specific morphological parameters of the illuminance peak fingerprint have a subtle and quantifiable correlation with the microscopic physical properties of the liquid, mainly surface tension. For example, liquids with higher surface tension have a larger meniscus curvature, forming a shorter focal length optical lens, and the resulting illuminance peak signal may be higher and narrower. Through a pre-calibrated fluid optics compensation model (lookup table or neural network), the morphological parameters (response amplitude, full width at half maximum, etc.) calculated in real time can be converted into accurate travel correction values, i.e., interface dynamic correction vectors.

[0066] Through this dynamic, real-time signal morphology-based compensation mechanism, the present invention can intelligently adapt to minute physical differences between different samples, performing real-time corrections to the positioning results at the micrometer level. This transforms it from a simple liquid level detector into a self-calibrating precision measuring instrument, meeting the stringent requirements of modern automated analysis for extremely precise handling of micro- and nano-scale samples.

[0067] Furthermore, a dynamic interface correction vector is generated based on the physical parameters of interfacial tension and wetting contact angle to compensate for physical deviations in liquid level height, and the liquid level coordinates are output. Corresponding to step S5 above, the specific process is as follows: The interfacial tension parameter and the wetting contact angle physical parameter are used as input variables, and the geometric radius of the container is used as a structural parameter, which are substituted into the capillary action model. The capillary action model calculates the capillary rise and fall heights formed by the liquid surface at the container wall by establishing an interfacial morphology evolution equation under the equilibrium relationship of liquid surface tension, wettability and gravity field. The output layer outputs the capillary rise and fall heights and establishes them as axial position correction values ​​for the interfacial anchor points, generating the interfacial dynamic correction vector. The capillary action model calculates and outputs the measurement deviation correction operator caused by the intrinsic properties of the material by establishing a mechanical evolution equation of liquid surface tension, wettability and gravity field.

[0068] The capillary rise or fall height calculated by the capillary action model is used as an operator and vector arithmetic superposition is performed on the established sample interface anchor points along the vertical travel axis Z. Specifically, the directionality of the capillary effect is determined based on the inverted wetting contact angle, and the height deviation value is used as a directional offset to compensate for the coordinate values ​​of the interface anchor points. This numerically cancels the axial positioning deviation caused by the meniscus's physical morphology, reconstructing the final axial travel coordinate values ​​of the liquid surface representing the true physical boundary of the liquid volume. In the output layer, the processor uses the inverted interfacial tension... With wetting contact angle As a core variable, combined with the preset liquid density Given the gravitational acceleration g and the container's geometric radius r, the interface displacement compensation is calculated using the capillary lift equilibrium equation. One specific form of this equation is the evolution operator of Jurin's law: ; in, This refers to the height by which the liquid level rises or falls at the container wall due to capillary action; the processor will then... The interface dynamic correction vector is established, and the coordinates of the initially determined interface anchor points are corrected by algebraic superposition.

[0069] Using the interface anchor point as the reference coordinate, the interface dynamic correction vector is used as the compensation operator to perform vector arithmetic operations to obtain the corrected center space coordinates; the corrected center travel coordinates are fused with the index position of the probe component in the planar scanning array to output the absolute boundary coordinates of the liquid under test in three-dimensional space.

[0070] A coordinate transformation and mapping algorithm is employed to use the final axial coordinate value, after physical property correction, as the Z-axis component. This value is then fused with the two-dimensional planar index position (X-axis and Y-axis coordinates) of the sample container, which is input in real time by automated analysis. Specifically, a three-dimensional spatial vector is constructed to map the precise axial positioning result based on fluid property analysis to the absolute spatial coordinate system defined by the device, generating the final three-dimensional spatial parameters describing the physical position of the sample interface.

[0071] Based on the final three-dimensional spatial parameters, a closed-loop feedback control signal is generated through a communication interface driver. Specifically, the corrected three-dimensional coordinate data is encapsulated into a digital data frame of a preset format. This data frame employs a specific encoding protocol and its structure includes at least: an instruction type field (to distinguish between stop, aspiration, or immersion actions), a target axial coordinate field (containing the compensated final Z-axis value), an operation parameter field (containing the aspiration volume or immersion depth calculated based on the actual physical boundary), and a cyclic redundancy check (CRC) field. The encapsulated data frame is then transmitted in real-time to the probe's motion controller via a high-speed communication bus (such as a CAN bus or Ethernet protocol), driving the actuator to perform actions according to the corrected physical coordinates.

[0072] By organically combining high-precision data stream generation, intelligent adaptive signal processing, and a second-order dynamic calibration strategy based on optical fingerprinting, a complete high-performance liquid surface positioning solution is achieved. This solution not only fundamentally solves the reliability and cross-contamination problems faced in materials analysis applications through its non-contact and property decoupling principles, but also provides strong technical support for accurate sample processing in the field of intelligent sensor analysis through its unique second-order calculation and dynamic calibration mechanism.

[0073] Example 2: This embodiment further elaborates on how intelligent sensors perform adaptive optimization based on dynamic noise models. The intelligent sensor's internal processor incorporates a rule-based expert decision-making process. Once the established dynamic noise model is parameterized, the expert decision-making process is triggered, comprising a series of priority-ordered conditional judgments and parameter configuration actions. The high-priority rule set in the expert decision-making process handles periodic noise: First, the process checks the periodic noise parameter set in the model. If a frequency component exists, the power spectral density exceeds a preset noise threshold, and the frequency is within the power frequency noise band, the process automatically instantiates a second-order infinite impulse response digital notch filter. The process then calls an algorithm to calculate the numerator and denominator coefficients required for the filter's transfer function. This algorithm maps the continuous-time domain second-order notch filter design to the discrete-time domain using the bilinear transform method, with the input parameter being the identified precise noise frequency. (50Hz or 60Hz) and a fixed quality factor Q=20. Another set of rules addresses random noise: the expert decision process determines whether the random noise power parameter in the model exceeds a preset threshold. If it does, it indicates a low signal-to-noise ratio, and the process adjusts the configuration of the Savitzky-Gore smoothing differential filter applied in step S3 accordingly. Specifically, it increases the length of the smoothing window while appropriately reducing the order of the fitting polynomial to maintain signal characteristics without distortion. This complete decision process automates the filter optimization process, provides logical interpretability, and allows for targeted and refined parameter configuration based on different noise compositions. Furthermore, the expert decision process ensures that the parameter configuration of the Savitzky-Gore smoothing differential filter (window length W and fitting polynomial order M) remains synchronized with the current noise environment. Specifically, when the random noise power is high, the process increases the smoothing window length W of the SG filter to enhance the noise reduction effect while maintaining the order M of the fitting polynomial at second or third order to preserve fidelity. All of these SG filter parameter configurations must be completed and locked before the probe begins its descent to ensure that the parameters for the differential calculations are consistent throughout the step.

[0074] Example 3: This embodiment further elaborates on the interface dynamic correction vector calculation model, particularly the optimization scheme for handling complex nonlinear relationships. This embodiment employs a radial basis function (RBF) neural network model pre-trained using machine learning methods. The network structure of the RBF neural network model is precisely defined as follows: an input layer, where the number of neurons corresponds to the dimension of the calculated morphological parameters, containing four neurons that receive four normalized input values: response amplitude, response bandwidth, integral flux, and asymmetric factor; a hidden layer, containing several RBF neurons with isotropic Gaussian activation functions; and an output layer, consisting of linear neurons whose output is the scalar interface dynamic correction vector. The training process of the RBF neural network model is a two-stage hybrid learning process. The first stage is unsupervised learning, using the K-means clustering algorithm to cluster the input feature vectors of a large number of training samples to determine the optimal center point of each RBF neuron in the hidden layer, and calculating the width or radius of influence of each basis function based on the clustering results. The second stage is supervised learning, using the singular value decomposition algorithm to determine the connection weights from the hidden layer to the output layer. The direct solution is obtained from the hidden layer output and the target output (actual deviation). A linear least squares problem is constructed to obtain the optimal connection weights. The training dataset is obtained by using a series of standard liquids with known precise surface tensions. The morphological parameter vectors of each liquid are measured and calculated using the method of this invention. Simultaneously, the precise optical refraction deviation between the peak illuminance point and the actual liquid surface is independently determined using higher-precision metrology equipment, such as a laser confocal microscope. The training dataset consists of five standard liquids with different surface tension gradients: pure water (72 mN / m), 25% ethanol aqueous solution (36 mN / m), 50% ethanol aqueous solution (28 mN / m), 10% glycerol aqueous solution (73 mN / m, used to introduce viscosity variables), and isopropanol (21 mN / m). During the calibration process, for each liquid, the probe is scanned multiple times with a variable step size of 10 mm / s to 100 mm / s, collecting a total of 500 pairs of morphological parameter vector-actual optical refraction deviation samples. For the radial basis function neural network, the number of hidden layer neurons is set to 1 / 10 of the number of samples, i.e., 50, and the basis function width is set to 1.5 times the average distance between input vectors. A large number of morphological parameter vector-actual deviation data pairs are used as the training set. The trained neural network model is stored in the non-volatile memory of the smart sensor, which can fit the highly nonlinear relationship between morphological parameters and correction vectors more accurately than a lookup table, thereby achieving higher-precision dynamic calibration when faced with unknown samples.

[0075] Example 4: This embodiment further elaborates on illuminance peak fingerprint recognition and extraction, detailing how to use numerical differentiation technology to achieve sub-pixel accuracy illuminance peak point localization.

[0076] After obtaining high-resolution, high signal-to-noise ratio digital illuminance-travel curves, the processor applies a Savitzky-Gore smoothing differential filter to these curves. The core advantage of the Savitzky-Gore smoothing differential filter lies in its ability to combine polynomial least squares fitting with differential operations. This achieves smooth noise reduction while accurately calculating the first and second derivatives of the curves, avoiding the problem of inaccurate derivative calculations in noisy environments using the traditional finite difference method.

[0077] First, the parameter configuration of the SG filter: Quantitative selection of window length W: The window length W must satisfy W < The specific selection rule is based on the empirical scaling factor k and the current signal-to-noise ratio. The mapping relationship. The empirical scaling factor k is usually set to 0.5. <k<0.9。

[0078] Quantification of data points: Before calculating the SG filter window length W, the data points calculated in micrometers must be quantized. Converted to a quantity measured in data points First, the probe is scanned at a constant speed (unit: Divide the illuminance signal by the fixed sampling frequency (50kHz) to obtain the spatial sampling step size, then... Dividing by the spatial sampling step size yields the result. . This represents the average number of data points contained within the response bandwidth range of the peak illumination fingerprint.

[0079] The formula for calculating window length W is based on the ideal window length. By It can be obtained by multiplying it by the empirical scaling factor k. The complete segmentation mapping rules are as follows: At that time, set =0.6, to minimize the smoothing effect.

[0080] ,set up =0.7, providing moderate smoothness.

[0081] At that time, set =0.8, to suppress noise to the maximum extent.

[0082] Determining the final window length W: The calculated ideal value... Round to the nearest integer, denoted as . ,pass This is implemented to ensure that W is odd, which meets the implementation requirements of the SG filter.

[0083] The order M of the fitting polynomial is set to second or third order to accurately fit the quadratic or cubic curvature of the fingerprint. Simultaneously, the order M of the fitting polynomial (typically 2 or 3) must satisfy the fundamental constraint W > M. This rule ensures that W is always less than M. .

[0084] Furthermore, sub-pixel precision peak point localization: Coarse localization: By searching the first derivative sequence output by the SG filter, the zero-crossing point that changes from positive to negative is identified. The zero-crossing point is the approximate extreme point of the illuminance peak fingerprint on the discrete data.

[0085] Extreme point verification: Read the second derivative value on the Z-axis coordinate corresponding to the zero-crossing point. Since the illuminance peak point is a local maximum, the second derivative value must be negative. To eliminate signal glitches or smooth peaks, it is necessary to further verify that the negative value is less than a preset negative threshold. .

[0086] Fine positioning (subpixel interpolation): Once coarse positioning and verification are successful, the processor will use the SG fitting polynomial near the zero-crossing point (two data points on each side) to calculate the precise Z-axis coordinates by analytically finding the roots (i.e., setting the first derivative of the fitting polynomial to zero). The analytical coordinates are the positions of the illumination peak points with subpixel precision, thereby minimizing the impact of quantization errors on positioning accuracy.

[0087] Furthermore, the fingerprint boundaries were determined: Starting from the located peak illuminance point, search in both the positive and negative Z-axis directions until the absolute value of the slope (first derivative) of the curve output by the SG filter is consistently less than the response gradient threshold dynamically determined by the baseline noise level. These points are identified as the starting and ending points of the illuminance peak fingerprint. Through the above method, this invention achieves highly robust differential calculation of feature signals and sub-pixel accuracy peak localization in noisy environments, providing a highly repeatable data foundation for subsequent generation of interface anchor points (first-order solution).

[0088] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A high-precision liquid level positioning method based on an illumination peak value method, characterized in that, include: The probe is driven to descend at a set speed while emitting horizontal light to illuminate the meniscus of the liquid under test. The vertical travel coordinates of the probe are recorded by utilizing the refraction and spectral evolution effect of the light beam at different curvature positions of the meniscus, generating a travel data stream. The transmitted light intensity response characteristics after passing through the container medium are continuously captured by a smart sensor, generating a photoelectric signal data stream. A dynamic noise model is constructed based on the baseline signal in the non-interface region of the photoelectric signal data stream. The filtering optimization is performed based on the digital filter inside the dynamic noise model to generate a filtered data stream; combined with the travel data stream, a digital illuminance-travel curve is constructed. Based on the digital illuminance-stroke curve, the illuminance peak fingerprint generated by the probe in the meniscus section is extracted, and the characteristic peak corresponding to the first spectral path of refraction on the liquid surface is identified. Interface anchor points are established based on the travel coordinates of characteristic peaks, and the morphological parameters of the illuminance peak fingerprint are analyzed. The interfacial tension and wetting contact angle physical parameters of the liquid under test are calculated by combining the capillary action model. A dynamic interface correction vector is generated based on the physical parameters of interfacial tension and wetting contact angle to compensate for physical deviations in liquid level height and output liquid level coordinates.

2. The high-precision liquid level positioning method based on peak illuminance method according to claim 1, characterized in that, The generation of the itinerary data stream specifically includes: A process feedback unit is adopted, which is coupled to the stepper drive unit that drives the probe to move vertically. When the stepper drive unit is running, the process feedback unit continuously outputs digital pulse signals corresponding to the probe's interactive trajectory. The signal processing unit counts and accumulates the digital pulse signals in real time, converts the accumulated count value into the sample axial scanning coordinate, and associates a timestamp with each coordinate data to generate a travel data stream.

3. The high-precision liquid level positioning method based on the peak illuminance method according to claim 1, characterized in that, The generation of the photoelectric signal data stream specifically includes: The transmitted photon flux modulated by the refractive energy level of the liquid interface is captured by an intelligent sensor and converted into an analog signal reflecting the photophysical properties of the sample interface. The signal is analyzed by a photosensitive detection unit with low noise gain and discretized and quantized by an analog-to-digital converter at a preset sampling period to generate a photoelectric signal data stream carrying liquid surface features. The intelligent sensor is configured as a detector sensitive to a preset wavelength according to the spectral response characteristics of the liquid to be tested, thereby enhancing the sensitivity of the identification of the interface spectral signal.

4. The high-precision liquid level positioning method based on the peak illuminance method according to claim 1, characterized in that, The construction of the dynamic noise model specifically includes: The stable baseline signal segments corresponding to the photoelectric signal data stream in the probe's gaseous medium region and far from the liquid surface are extracted as background reference signals. Statistical analysis is performed on the stable baseline signal segments to calculate the mean, standard deviation, and statistical variance, quantifying the amplitude characteristics of random noise. Fast Fourier Transform is applied to the stable baseline signal segments to analyze spectral characteristics and identify periodic interferences introduced by ambient light and signal transmission links to the material property analysis. Based on the combination of noise amplitude characteristics and periodic noise frequency characteristics, a dynamic noise model is constructed.

5. The high-precision liquid level positioning method based on peak illuminance method according to claim 1, characterized in that, The construction of the digital illuminance-stroke curve specifically includes: A two-dimensional data structure of double-precision floating-point type is constructed. Based on timestamps, the filtered data stream reflecting the optical properties of the material and the travel data stream reflecting the spatial distribution are spatiotemporally fused. For data points with perfectly matched timestamps, the illuminance value and travel value are stored as a set of feature mapping pairs in the two-dimensional data structure. For data points with mismatched timestamps, a resampling algorithm based on cubic spline interpolation is used to calculate the corresponding illuminance value and travel value at the timestamp, generating the digital illuminance-travel curve.

6. The high-precision liquid level positioning method based on the peak illuminance method according to claim 1, characterized in that, The extraction of the fingerprint from the illuminance peak generated by the probe in the meniscus segment specifically includes: Perform first and second-order numerical differentiation on the digital illuminance-stroke curve; search for zero-crossing points where the first derivative value changes from positive to negative, and determine whether the second derivative value corresponding to the zero-crossing point is less than the negative threshold determined based on the standard deviation of the baseline signal, thereby determining the illuminance peak point of the illuminance peak fingerprint; based on the response gradient threshold of the baseline signal, search forward and backward from the illuminance peak point to the position where the absolute value of the curve slope is less than the response gradient threshold, thereby determining the start and end points of the illuminance peak fingerprint, and extracting the illuminance peak fingerprint.

7. The high-precision liquid level positioning method based on the peak illuminance method according to claim 1, characterized in that, The physical parameters of interfacial tension and wetting contact angle of the liquid under test, calculated by combining the capillary action model, specifically include: The response bandwidth and integral flux of the illuminance peak fingerprint are extracted, and the radius of curvature distribution of the meniscus at the container wall is analyzed in conjunction with the optical refractive index constant. The radius of curvature distribution is substituted into the Young-Laplace model to calculate the interfacial pressure difference, and the interfacial tension parameters of the liquid under test are derived by combining the hydrostatic equilibrium conditions. By analyzing the response asymmetry coefficient of the illuminance peak fingerprint, the physical parameters of the wetting contact angle of the liquid to the container wall and the interfacial tension of the liquid under test are obtained.

8. The high-precision liquid level positioning method based on the peak illuminance method according to claim 1, characterized in that, The generated interface dynamic correction vector specifically includes: The interfacial tension parameter and the wetting contact angle physical parameter are used as input variables, and the geometric radius of the container is used as a structural parameter, which are substituted into the capillary action model. The capillary action model calculates the capillary rise and fall heights formed by the liquid surface at the container wall by establishing an interfacial morphology evolution equation under the equilibrium relationship of liquid surface tension, wettability and gravity field. The output layer outputs the capillary rise and fall heights and establishes them as axial position correction values ​​for the interfacial anchor points, generating the interfacial dynamic correction vector. The capillary action model calculates and outputs the measurement deviation correction operator caused by the intrinsic properties of the material by establishing a mechanical evolution equation of liquid surface tension, wettability and gravity field.

9. A high-precision liquid level positioning method based on peak illuminance method according to claim 1, characterized in that, The output liquid level coordinates specifically include: Using the interface anchor point as the reference coordinate, the interface dynamic correction vector is used as the compensation operator to perform vector arithmetic operations to obtain the corrected center space coordinates; the corrected center travel coordinates are fused with the index position of the probe component in the planar scanning array to output the absolute boundary coordinates of the liquid under test in three-dimensional space.

Images