A coating thickness detection method and system based on visual monitoring

Abstract

The application relates to the technical field of optical three-dimensional measurement and machine vision, and discloses a coating thickness detection method and system based on visual monitoring, which comprises the following steps: controlling a test environment to induce transient evaporation of a liquid phase on a coating surface, and collecting an image sequence under structured light projection; calculating an evaporation gradient index field of the image, extracting a time stamp corresponding to a time sequence maximum value to construct an extreme value time field; performing a spatial gradient operation on the extreme value time field to generate an evaporation dynamics scalar field; fusing the scalar field with a structured light fringe modulation degree matrix to construct a quality map function; guiding a spatial phase unwrapping path by the quality map function, obtaining a full field of view absolute phase, and calculating coating thickness. The application uses phase transition space-time evolution rules to guide phase unwrapping, so that the algorithm path actively avoids a phase abnormal area caused by water droplet refraction, error global diffusion is inhibited, and the accuracy of coating thickness measurement under a dynamic alternating environment is improved.

Description

Technical Field

[0001] This invention relates to the field of optical three-dimensional measurement and machine vision technology, specifically to a coating thickness detection method and system based on visual monitoring. Background Technology

[0002] In the manufacturing process of electronic components such as printed circuit boards, the thickness of the surface coating directly affects the protective performance of the device. Structured light projection-based visual monitoring technology, due to its non-contact and full-field-of-view characteristics, is widely used to measure the three-dimensional morphology and thickness of coatings. This technology projects structured light stripes onto the surface of the object being measured, uses a camera to acquire images of the stripes highly modulated by the object's surface, and combines this with a spatial phase unfolding algorithm to calculate the absolute morphological data of the object.

[0003] Existing structured light measurement methods typically operate in stable, dry environments at room temperature. However, in environmental adaptability testing or complex industrial settings, coated samples are often exposed to fluctuating temperatures and humidity. In such environments, the coating surface is prone to gas-liquid phase transitions, resulting in condensation and the formation of discrete water droplets. When structured light is projected onto these droplets, it produces strong specular reflections, causing localized overexposure of the camera sensor and loss of interference fringe information in the corresponding areas. Simultaneously, light refraction at the droplet edges causes physical phase jumps in local image regions. Existing phase unwrapping algorithms primarily rely on fringe modulation as a path guide, but this evaluation mechanism cannot distinguish and isolate phase anomalies caused by droplet refraction. During de-overlay operations, the algorithm's unwrapping path directly passes through these refraction anomaly regions, generating initial calculation errors. Since spatial phase unwrapping algorithms are mathematically an error accumulation process, local phase calculation errors propagate along the calculation path to the surrounding valid data area, leading to global absolute phase data distortion and directly causing the reconstruction and detection of coating thickness to fail. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a coating thickness detection method and system based on visual monitoring, which solves the problem of overexposure caused by liquid condensation and discrete water droplets on the coating surface when measuring coating morphology and thickness in situ under alternating temperature and humidity conditions, resulting in light refraction and localized reflection.

[0005] To address the above problems, the present invention provides the following technical solution: The first aspect of this invention provides a coating thickness detection method based on visual monitoring, comprising the following steps: The test environment was controlled to induce transient liquid-phase evaporation on the coating surface, and image sequences of the coating surface were acquired under structured light projection. The evaporation gradient exponential field of the coating surface is calculated based on the image sequence, and the timestamps corresponding to the time series maxima of the evaporation gradient exponential field are extracted to construct the extreme value time field. Perform spatial gradient calculations on the extreme time field to generate a scalar field of evaporation kinetics; A mass map function is constructed by fusing the scalar field of evaporation kinetics with the modulation matrix of structured light stripes. The spatial phase expansion algorithm path is guided by the quality map function to obtain the full field-of-view absolute phase data of the coating surface, and the coating thickness is calculated based on the full field-of-view absolute phase data.

[0006] The time corresponding to the extreme value of pixel phase transition is converted into a spatial gradient field characterizing the liquid phase retreat feature, and this gradient field is used to construct a joint quality map function. This quality map function reduces the weight score of pixels at the dynamic evaporation edge. During spatial phase unwrapping, the algorithm guides the disassembly path according to this quality map function to avoid abnormal regions caused by abrupt changes in refractive index, suppressing the propagation of phase error to the external effective data area, thereby ensuring the accuracy of 3D topography measurement under dynamic alternating environment.

[0007] Furthermore, controlling the test environment conditions to induce transient liquid-phase evaporation on the coating surface includes: The temperature and humidity of the test environment are adjusted to approximate the target dew point temperature of the coating, and the global information entropy of the acquired image sequence is calculated in real time. When the absolute value of the first time derivative of the global information entropy of multiple consecutive frames is less than the set minimum judgment threshold, the coating surface is judged to have reached the gas-liquid phase transition saturation state. The target temperature of the test environment was then increased and dehumidification was activated to induce transient evaporation of liquid phase water.

[0008] Furthermore, during the transient evaporation of liquid-phase water, the method also includes: The state of intense evaporation is determined based on the spatial integral value of the evaporation gradient index within the effective area of ​​the entire field of view. When the spatial integral value reaches its peak, the spatial frequency of the stripes of the structured light projection is increased simultaneously, and the intensity of the projected light is reduced to avoid overexposure of the photosensitive target surface caused by discrete droplets.

[0009] Furthermore, the exponential field of the evaporation gradient on the coating surface is calculated based on the image sequence, including: Establish a time sliding window with a set frame length for the acquired image sequence; For each pixel in the image sequence, calculate the discrete variance of grayscale within the current time sliding window pixel by pixel; The grayscale discrete variance is used as the evaporation gradient exponent of the corresponding pixel at the current time to generate an evaporation gradient exponent field that includes both time and two-dimensional spatial dimensions.

[0010] Furthermore, the timestamps corresponding to the time series maxima of the evaporation gradient exponential field are extracted to construct the extreme value time field, including: An effective evaporation judgment threshold is set, which is calibrated based on the multiple of the variance of the background noise of the camera in static dark field. For any pixel in the evaporation gradient exponent field, determine whether the maximum evaporation gradient exponent of that pixel in the time series is greater than the effective evaporation judgment threshold. If the value is greater than the maximum evaporation gradient exponent, the timestamp corresponding to the maximum evaporation gradient exponent is extracted as the extreme time field value at that pixel. If the value is not greater than the maximum evaporation gradient exponent, it is determined that no gas-liquid phase transition has occurred at that pixel, and a set constant identifier value is assigned to that pixel. By using a threshold to isolate static regions where no phase transition has occurred, the background noise is prevented from introducing false spatial gradients in subsequent calculations.

[0011] Furthermore, spatial gradient calculations are performed on the extreme time field to generate a scalar field of evaporation kinetics, including: The filtering algorithm is called to perform smoothing preprocessing on the effective data region in the extreme time field, and the matrix containing discrete timestamps is transformed into a floating-point surface matrix with continuous transition. The spatial gradient magnitude of the extreme time field after smoothing preprocessing in the two-dimensional image pixel coordinate system is calculated using the difference operator, and the spatial gradient magnitude is mapped to the evaporation kinetic scalar field, which is used to characterize the spatial retreat propagation velocity of the gas-liquid interface.

[0012] Furthermore, the mass map function is constructed by fusing the evaporation kinetic scalar field with the structured light fringe modulation matrix. The numerical construction logic is as follows: The value of the quality graph function is formed by adding the first part of the data to the second part of the data; The first part of the data is the product of the normalized structured light stripe modulation matrix and the set first weight coefficient; The second part of the data is the product of the set attenuation penalty term and the set second weight coefficient. The attenuation penalty term is an exponential function constructed with the natural constant as the base and the negative of the product of the evaporation kinetic scalar field and the kinetic attenuation constant as the exponent. Furthermore, the sum of the first weighting coefficient and the second weighting coefficient is one.

[0013] Furthermore, the spatial phase expansion algorithm path guided by the quality map function includes: The quality map function is input into the minimum spanning tree algorithm. The pixel with the highest quality map function value is used as the starting seed point, and the minimum spanning tree path is generated in descending order of quality map function value. The unpacking operation of the wrapped phase is advanced pixel by pixel along the minimum spanning tree path. Guided by the prior values ​​of the quality map function, the low-quality region determined by the evaporation kinetic scalar field to be at the edge of dynamic evaporation retreat is bypassed while maintaining physical connectivity.

[0014] Furthermore, acquire the full field-of-view absolute phase data of the coated surface, including: For un-un-overlapped regions isolated by the minimum spanning tree path due to extremely low quality map function values, the un-un-overlapped regions are marked as invalid masks; The effective phase data from the periphery of the un-unoverlapped region is used for bilinear interpolation repair, and the output is the full field-of-view absolute phase data after removing phase jump interference.

[0015] A second aspect of the present invention provides a coating thickness detection system based on visual monitoring, comprising: An environmental test chamber is used to regulate the internal temperature and humidity to induce transient liquid-phase evaporation on the coating surface. Projector used to adaptively project structured light sequences onto the coated surface; Industrial cameras are used to continuously acquire image sequences of coated surfaces under structured light projection at a fixed frame rate; The measurement and control host is communicatively connected to the environmental test chamber, the projector, and the industrial camera; the measurement and control host is configured to control the environmental test chamber, the projector, and the industrial camera to execute the coating thickness detection method based on visual monitoring described in the first aspect above.

[0016] This invention provides a method and system for coating thickness detection based on visual monitoring. It has the following beneficial effects: 1. This invention dynamically adjusts the environmental test chamber state to induce transient evaporation by real-time monitoring of the global information entropy of the image sequence and the spatial integral value of the evaporation gradient exponent. During the intense evaporation phase, it simultaneously increases the spatial frequency of the structured light fringes and reduces the intensity of the projected light. This control mechanism effectively avoids overexposure of the photosensitive target surface caused by discrete water droplets on the coating surface under alternating temperature and humidity conditions, ensuring complete acquisition of interference fringe information.

[0017] 2. This invention utilizes a time-sliding window to extract the temporal extrema of the evaporation gradient index pixel by pixel to construct an extreme value time field. After removing background noise regions where no gas-liquid phase transition has occurred by passing a set threshold, differential operations are performed on the effective data to generate an evaporation kinetic scalar field. This data processing logic quantifies the recession process of liquid phase water in the time dimension into spatially distributed characteristic parameters, providing a physical prior basis for evaluating the quality of measurement data.

[0018] 3. This invention integrates the evaporation kinetic scalar field reflecting the retreat characteristics of the gas-liquid interface with a conventional fringe modulation matrix to construct a joint mass map function with an exponential decay penalty term. During spatial phase unwrapping, this function guides the disassembly path to actively avoid phase anomaly regions caused by abrupt changes in the refractive index of water droplets, suppressing the propagation of local calculation errors across the entire field of view, thereby improving the accuracy of absolute phase and thickness measurements of the coating under dynamic conditions. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the structure of a coating thickness detection system according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating the overall process of a coating thickness detection method according to an embodiment of the present invention. Figure 3 This is a timing diagram of the in-situ optical spatial reference calibration and initial field construction in one embodiment of the present invention; Figure 4 This is a processing timing diagram of environmental parameter induction and liquid phase saturation determination in one embodiment of the present invention; Figure 5 This is a timing diagram of phase transition transient triggering and optical frequency domain reconstruction in one embodiment of the present invention; Figure 6 This is a timing diagram of the joint phase de-overlay processing driven by the scalar field of evaporation kinetics in one embodiment of the present invention; Figure 7 This is a timing diagram of the compensation calculation for the relative thickness change of the coating in one embodiment of the present invention; Figure 8 This is a diagram illustrating the operational logic and data flow of a coating three-dimensional morphology measurement and control system according to a specific application embodiment of the present invention. Figure 9 The histogram compares the root mean square error of morphology reconstruction at different evaporation stages using a conventional method and the method of the present invention, as an embodiment of the present invention.

[0020] Among them, 10 is the environmental test chamber; 11 is the observation window; 20 is the projector; 30 is the industrial camera; 40 is the measurement and control host; 50 is the sample fixture; and 60 is the reference anchor point. Detailed Implementation

[0021] 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.

[0022] Reference Figure 1The coating thickness detection system provided by the present invention includes an environmental test chamber 10, a projector 20, an industrial camera 30, a measurement and control host 40, a sample fixture 50, and a reference anchor point 60.

[0023] The side wall of the environmental test chamber 10 is provided with an observation window 11. The projector 20 and the industrial camera 30 are arranged outside the environmental test chamber 10. The projection end of the projector 20 and the acquisition end of the industrial camera 30 are both facing the observation window 11. The projection light path formed by the projector 20 and the acquisition light path formed by the industrial camera 30 enter the internal space of the environmental test chamber 10 through the observation window 11 and cover the sample fixture 50, the reference anchor point 60 and the area where the coating sample to be tested is located.

[0024] The sample holder 50 is installed inside the environmental test chamber 10 to fix or support the coating sample to be tested. The reference anchor 60 is fixedly installed on the surface of the sample holder 50 and is located in the same measurement field of view of the industrial camera 30 as the coating sample. The reference anchor 60 is made of microcrystalline glass and can provide a dimensionally stable phase reference under varying temperature and humidity conditions.

[0025] In an optional embodiment, the reference anchor 60 includes one or more anchor units. When the reference anchor 60 includes multiple anchor units, the multiple anchor units are distributed at different positions of the sample fixture 50 and are together with the coating sample within the measurement field of view of the industrial camera 30. The measurement and control host 40 can fit the full field of view spatial drift correction field according to the phase change of each anchor unit.

[0026] Specifically, the process of fitting the full field-of-view spatial drift correction based on the phase changes of each anchor point unit by the telemetry and control host 40 includes the following steps: (1) Obtain the anchor point phase drift: Extract the phase drift set within the field of view. The center coordinates of each anchor unit ,in During the measurement process, the current phase value of each anchor point unit is acquired in real time. and the initial reference phase value By performing subtraction, the phase drift of each anchor point element is obtained. .

[0027] (2) Constructing a spatial drift surface model: using the image coordinates of each anchor point unit As the independent variable, with phase drift as the independent variable. Using a two-dimensional polynomial fitting algorithm (e.g., least squares method) as the dependent variable, a spatial drift model across the entire field of view is fitted. As a preferred embodiment, a second-order surface fitting model can be used, with the following calculation formula: ; The monitoring and control host uses the data from the NN anchor points to solve for the polynomial coefficients. ~ Thus, any pixel in the entire field of view can be obtained. Fitted phase shift .

[0028] (3) Global drift compensation: When calculating the global phase of the object under test, the measured phase of each pixel in the entire field of view is subtracted from the corresponding fitted phase drift. This enables high-precision full-field spatial drift correction, eliminating systematic errors caused by changes in ambient temperature or minor vibrations.

[0029] The measurement and control host 40 is communicatively connected to the environmental test chamber 10, the projector 20, and the industrial camera 30. The measurement and control host 40 sends control commands to the environmental test chamber 10, the projector 20, and the industrial camera 30, and receives environmental parameters, projection parameters, and image acquisition data to complete environmental control, optical acquisition, and image processing during the coating thickness detection process.

[0030] Reference Figure 2 The coating thickness detection method provided by the present invention includes the following steps: During the detection process, the measurement and control host 40 records the image acquisition time, ambient temperature, relative humidity, spatial frequency of projected fringes, intensity of projected light, camera exposure parameters, phase parameters of reference anchor points, and phase quality parameters in association with the corresponding image frames. The recorded data is used for subsequent phase calculation, morphology registration, thermal expansion compensation, refractive index compensation, and thickness change statistics to ensure the data correspondence between changes in environmental parameters, changes in projection parameters, and the gas-liquid phase transition process on the coating surface.

[0031] S10, the measurement and control host 40 controls the environmental test chamber 10 to maintain a normal temperature and humidity state, and controls the projector 20 to project multi-frequency phase-shift structured light into the internal space of the environmental test chamber 10; the industrial camera 30 synchronously acquires reflected image data. The measurement and control host 40 determines the initial anchor point reference phase based on the absolute phase data of the area where the reference anchor point 60 is located, and establishes the initial differential reference topography field of the coating surface accordingly.

[0032] S20, the measurement and control host 40 adjusts the internal environmental parameters of the environmental test chamber 10 to approach the dew point critical state of the coating sample; the industrial camera 30 acquires images of the coating surface, and the measurement and control host 40 calculates the global information entropy and its first-order time derivative of the image frame by frame. When the first-order time derivative meets the preset convergence condition, the measurement and control host 40 determines that the gas-liquid phase change on the coating surface has reached a saturated state of dynamic equilibrium, and records the current temperature parameters under this saturated state and the saturation time parameters experienced from the start of dew point approach to reaching the saturated state.

[0033] S30, after determining that saturation has been reached, or after triggering the timeout status flag and latching the alternative status parameters, the measurement and control host 40 switches the environmental parameters of the environmental test chamber 10 to the dehumidification state to induce liquid phase evaporation on the coating surface. The measurement and control host 40 calculates the evaporation gradient exponential field based on the grayscale variance data of the image pixels, and adjusts the spatial frequency and intensity of the projected light stripes of the projector 20 according to the numerical distribution of the evaporation gradient exponential field.

[0034] S40, during the liquid-phase evaporation process on the coating surface, the measurement and control host 40 extracts the temporal extrema of the evaporation gradient exponential field and constructs an extremum time field; then, it performs spatial gradient calculation on the extremum time field to obtain the evaporation kinetic scalar field. The measurement and control host 40 generates a weight matrix based on the evaporation kinetic scalar field and introduces this weight matrix into the quality map function of the structured light phase unrolling algorithm to obtain the full field-of-view absolute phase data.

[0035] S50, the measurement and control host 40 calculates the dynamic phase drift of the reference anchor point 60, and performs differential correction and phase height mapping on the absolute phase data of the entire field of view to generate the current three-dimensional topography data; subsequently, it registers the current three-dimensional topography data with the initial differential reference topography field to the same coordinate system. The measurement and control host 40 calculates the normal thermal expansion offset of the sample based on the measured temperature parameters at the phase acquisition time, the reference temperature parameters, and the thermal expansion coefficient of the substrate, and calculates the refractive index compensation characteristic term based on the saturation time parameter and the pre-calibrated refractive index compensation model. The measurement and control host 40 performs compensation calculations on the current three-dimensional topography data, the initial differential reference topography field, the normal thermal expansion offset, and the refractive index compensation characteristic term, and outputs the relative thickness change of the coating sample relative to the initial reference state.

[0036] Reference Figure 3 In step S10, the monitoring and control host 40 may further execute the following sub-steps S101 and S102.

[0037] S101, the measurement and control host 40 controls the projector 20 and the industrial camera 30 to perform multi-frequency phase shift acquisition and phase demodulation processing to obtain the initial anchor point reference phase of the reference anchor point 60.

[0038] The reference anchor point 60 is made of microcrystalline glass. This material has a low coefficient of thermal expansion and small dimensional changes under alternating temperature and humidity conditions, thus serving as a stable space reference inside the environmental test chamber 10.

[0039] The control unit 40 maintains the environmental test chamber 10 at a set normal temperature and humidity. For example, the temperature can be set to 20℃ to 25℃, and the relative humidity can be set to 40% to 50%. Performing initial data acquisition under this condition helps to reduce the effects of thermal stress and moisture absorption deformation caused by fluctuations in ambient temperature and humidity.

[0040] The control and measurement host 40 generates multiple sets of phase-shifted sinusoidal grating images with different spatial frequencies and drives the projector 20 to project these images into the environmental test chamber 10 through the observation window 11. The projected light covers the sample fixture 50, the reference anchor point 60, and the coating sample. Before acquisition, the reference anchor point 60 and the coating sample are both within the effective depth of field of the industrial camera 30.

[0041] Subsequently, the industrial camera 30 performs exposure according to the synchronous trigger signal of the measurement and control host 40, acquiring an image of deformed fringes formed by reflections from the reference anchor point 60, the sample fixture 50, and the surface of the coated sample. After receiving the deformed fringes image, the measurement and control host 40 performs phase calculation to obtain the initial absolute phase distribution data of the entire field of view. The specific processing steps for multi-frequency heterodyne phase unwrapping, wrapped phase extraction, and phase unwrapping can be achieved using structured light three-dimensional measurement methods commonly used in this field, and will not be elaborated here.

[0042] Obtain initial absolute phase distribution data for the entire field of view Subsequently, the telemetry and control host 40 determines the region corresponding to the reference anchor point 60 in the image pixel coordinate system, extracts the absolute phase value within this region, calculates its spatial average value, and uses this spatial average value as the initial anchor point reference phase. .

[0043] The pixel region corresponding to the reference anchor point 60 can be determined by either a high-contrast marked border on its surface or by a pre-calibrated pixel connected region mask. Its calculation model can be expressed as: ; in, This represents the spatially connected region occupied by the reference anchor point 60 in the camera pixel coordinate system. For the pixel coordinate space index of industrial camera 30, Representative area The number of effective pixels within.

[0044] S102, the measurement and control host 40 establishes the initial differential reference topography field of the coating surface based on the initial anchor point reference phase.

[0045] Both the projection light path of the projector 20 and the acquisition light path of the industrial camera 30 must pass through the observation window 11 of the environmental test chamber 10. The observation window 11 typically contains a multi-layered glass structure for anti-fogging and heat insulation. When the light beam passes through the multi-layered medium and the gas inside the chamber, optical path refraction distortion will occur. This distortion may include common-mode phase drift that changes with environmental conditions, or non-common-mode spatial distortion that changes with the field of view position.

[0046] To reduce the impact of the aforementioned distortions on the initial reference, the telemetry and control host 40 obtains the initial absolute phase distribution data from the entire field of view. The phase distribution data of the coated sample surface is extracted and compared with the initial anchor point reference phase. Perform differential calculations. By performing relative differential calculations under the same optical path, the common-mode phase drift caused by changes in the observation window 11, the gas state inside the chamber, and the installation attitude can be reduced.

[0047] For the remaining non-common-mode spatial distortion, the telemetry and control host 40 can be corrected by combining the pre-established phase height mapping function, multi-point reference calibration model or field distortion correction model.

[0048] Specifically, the process of the measurement and control host 40 making corrections based on the pre-established model includes: (1) Obtain depth information by combining phase height mapping function: In the system pre-calibration stage, the depth information of each pixel is established by moving the calibration plane with known height. The mapping relationship between the phase difference and the physical height. During actual measurement, the telemetry and control host calculates the current absolute phase. Phase with reference plane phase difference The actual height is then calculated by substituting it into a pre-established phase height mapping function. As a specific embodiment, the phase mapping function can adopt the following conventional nonlinear model: ; in, All of these are system phase mapping coefficient matrices obtained by fitting using the least squares method during the calibration phase.

[0049] (2) Eliminating lens distortion by combining field-of-view distortion correction model: To address the nonlinear deformation caused by the optical lens, the measurement and control host uses the camera intrinsic parameter matrix and distortion coefficients (including radial distortion coefficients) obtained in advance (e.g., using Zhang Zhengyou calibration method) to eliminate lens distortion. and tangential distortion coefficient The extracted two-dimensional image coordinates are resampled and corrected according to the distortion correction mathematical model, thereby eliminating stretching and deformation errors at the edge of the field of view. Finally, combined with the aforementioned height... The data is then fused into high-precision 3D point cloud data.

[0050] In one embodiment, the reference anchor point 60 includes one anchor point unit; in another embodiment, the reference anchor point 60 includes multiple anchor point units. When multiple anchor point units are provided, each anchor point unit is distributed at different positions on the sample fixture 50. The measurement and control host 40 calculates the phase drift of each anchor point unit and fits the spatial drift correction field of the measurement area of ​​the coated sample according to its spatial position. The fitting can be performed using plane fitting, polynomial fitting, radial basis function interpolation, or thin plate spline interpolation methods.

[0051] Specifically, the spatial drift correction field of the measurement area of ​​the coated sample is fitted according to its spatial position. The specific process is as follows: Get the coating sample around Spatial position of each valid reference point in the image coordinate system and its local drift (in Using the spatial location of the reference point, a two-dimensional spatial interpolation algorithm is employed to calculate all unknown pixels within the measurement area of ​​the coating sample. The correction value is used to generate a continuous two-dimensional matrix that matches the resolution of the measurement area, i.e., the spatial drift correction field. .

[0052] As a specific preferred embodiment, this scheme uses inverse distance weighted interpolation (IDW) to construct the spatial drift correction field, and any pixel point within the measurement area of ​​the coating sample... The formula for calculating the correction value is as follows: ; Among them, the weight term , Indicates the current pixel. To the benchmark points The Euclidean distance between them; The distance attenuation index is set, typically with a value of 2. Using the above formula, the spatial drift correction field of the seamlessly covered coating sample area can be accurately fitted based on the spatial positions of surrounding reference points.

[0053] After differential processing is completed, the telemetry and control host 40 uses system calibration parameters to map the differential phase data to a three-dimensional spatial coordinate system, generating an initial differential reference topography field. Its mapping model can be expressed as: ; in, Represents the initial differential reference topography field of the coating surface. This represents the phase height mapping function. The phase height mapping function and the system epipolar geometry calibration parameters can be obtained through camera-projector calibration; for example, the relevant intrinsic and extrinsic parameter matrices can be obtained using the Zhang Zhengyou calibration method, and the specific calibration process is a conventional technique in this field.

[0054] Initial differential reference topography field This represents the initial relative height distribution of the coated sample surface relative to the reference anchor point 60. The measurement and control host 40 can also calculate the local roughness parameters, regional roughness parameters, or initial surface texture parameters of the coated sample based on this initial differential reference topography field, and use them as input features for the subsequent refractive index compensation model.

[0055] Therefore, the initial differential reference topography field can be used as the zero-point reference for subsequent thickness variation calculations, and can to some extent reduce the influence of the refraction of the multilayer medium in the observation window 11 and the initial fixed optical path error on the detection results.

[0056] Reference Figure 4 In step S20, the monitoring and control host 40 may execute sub-steps S201 to S203 in sequence.

[0057] S201, the monitoring and control host 40 sends a temperature and humidity adjustment command to the environmental test chamber 10 to perform dew point approach control.

[0058] The monitoring and control host 40 pauses the original routine temperature and humidity alternation program of the environmental test chamber 10 and determines the target dew point temperature based on the pre-input physical parameters of the coating sample or the manually set reference value. At the same time, the monitoring and control host 40 acquires the real-time temperature and relative humidity fed back from inside the environmental test chamber 10.

[0059] The measurement and control host 40 also acquires the surface temperature parameters of the coating sample. These surface temperature parameters can be obtained by temperature sensors located on the sample holder 50, the back of the coating sample, or near the coating sample; they can also be estimated based on the air temperature, relative humidity, thermal conductivity of the sample holder 50, and thermal inertia parameters of the coating sample within the environmental test chamber 10.

[0060] Specifically, the process for estimating the thermal inertia parameters of the coating sample is as follows: based on the material properties and physical state of the coating sample, a thermodynamic characteristic model is constructed to estimate its thermal inertia parameters (Thermal Inertia, abbreviated as TIA). As a specific preferred embodiment, this thermal inertia parameter... The calculation formula is as follows: ; in, The thermal conductivity of the coated sample is given. The density of the coated sample, This represents the specific heat capacity of the coated sample.

[0061] In practice, the monitoring and control host can obtain the above-mentioned basic physical parameters by reading a pre-established database of coating material properties (table lookup method). Thus, the thermal inertia parameters can be directly calculated. Alternatively, an external temperature sensor can be used to acquire the transient temperature response curve of the coating under specific heat source excitation, and the thermal inertia parameter can be estimated by backfitting with known boundary conditions. This provides accurate thermodynamic prior data for subsequent compensation models.

[0062] The monitoring and control host 40 determines the dew point approximation control quantity based on the difference between the ambient dew point temperature and the surface temperature of the coating sample. It then adjusts the output power of the cooling, heating, and humidifying modules within the environmental test chamber 10 using a proportional-integral-derivative control algorithm, gradually bringing the environmental parameters inside the chamber closer to the critical dew point state of the coating sample. At this point, the coating surface possesses the thermodynamic conditions for condensation.

[0063] The parameter tuning, valve adjustment, and power control methods of the proportional-integral-derivative (PID) control algorithm can be implemented using conventional control methods in the field of environmental testing equipment, and will not be elaborated here.

[0064] S202, as the environmental parameters approach the dew point critical state, the monitoring and control host 40 calculates the global information entropy based on the image sequence acquired by the industrial camera 30.

[0065] Under alternating humid and hot conditions, coatings with different aging degrees typically exhibit different surface energy states. For coatings with strong hydrophilicity, condensation easily spreads on the surface to form a continuous water film; for coatings with strong hydrophobicity, condensation more readily forms discrete water droplets.

[0066] Regardless of whether the liquid phase exists in the form of a water film or water droplets, as the amount of condensation increases, the microscopic rough texture of the coating surface will be masked to varying degrees. This masking effect will cause the gray-level distribution in the camera image to tend to concentrate, thereby reducing the image texture richness. Compared with a single gray-level threshold determination method, the determination method based on global information entropy is not sensitive to changes in the reflective position of water droplets and can accommodate both continuous water films and discrete water droplets.

[0067] Industrial camera 30 continuously acquires a sequence of grayscale images of the coating surface at a set frame rate. Measurement and control host 40 reads the image data frame by frame, counts the frequency of pixel occurrence at each grayscale level within the entire image or a pre-defined effective measurement area, and calculates the probability value corresponding to each grayscale level, thereby obtaining the global information entropy of the current frame image. Its calculation model can be expressed as: ; in, Represents the moment The global entropy of the captured current frame image. This represents the total number of gray levels in an industrial camera's output image (30), within a typical eight-bit digital image. The value is 256. Represents a discrete grayscale index variable, with values ​​ranging from 0 to... . Representative moment The image has gray levels of The probability value of the number of pixels in a given image relative to the total number of valid pixels.

[0068] S203, the monitoring and control host 40 determines whether the gas-liquid phase change has reached dynamic equilibrium based on the first time derivative of the global information entropy, and locks the corresponding characteristic parameters.

[0069] As condensation continues, the micro-texture of the coating surface is gradually covered by the liquid phase, and the global information entropy St calculated by the monitoring and control host 40 typically shows a decreasing trend. When the condensation rate and evaporation rate tend to balance, the liquid phase coverage state on the coating surface no longer expands significantly, the image texture features tend to stabilize, and the corresponding global information entropy... It has entered a low and flat range.

[0070] 40 pairs of consecutive multi-frame global information entropy of the telemetry and control host Discrete-time differentiation is performed to obtain the first-order time derivative of the global information entropy. When this first-order time derivative satisfies the preset convergence condition, the measurement and control host 40 determines that the gas-liquid phase transition on the coating surface has reached saturation. The convergence condition can be expressed as: ; in, The first time derivative of the global information entropy. It is the absolute value of the first-order time derivative. This is the set minimum judgment threshold. Using the absolute value convergence criterion avoids misjudging a saturated state due to a negative derivative during the rapid decrease of global information entropy. The value can be determined based on the background image noise level of the industrial camera 30, for example, set to a constant between 0.01 and 0.05.

[0071] To prevent environmental disturbances from prolonging the judgment process, the monitoring and control host 40 can be configured with a timeout exit mechanism. When the dew point approach time exceeds the set maximum allowable time, the monitoring and control host 40 outputs a timeout status flag and latches the current temperature parameter, the current relative humidity parameter, and the ongoing approach time. In this case, the monitoring and control host 40 can proceed to the subsequent dehumidification process, but will not directly recognize this timeout state as a saturation state; the timeout status flag serves as an input parameter for the subsequent refractive index compensation model or phase quality evaluation process.

[0072] When the monitoring and control host 40 enters the subsequent dehumidification process due to the timeout status flag, the monitoring and control host 40 records the ambient temperature at the timeout moment as the equivalent current temperature parameter, records the time elapsed from the start of sending the temperature and humidity adjustment command to the timeout moment as the equivalent saturation time parameter, and writes the timeout status flag, the equivalent current temperature parameter, and the equivalent saturation time parameter into the data record.

[0073] Under normal convergence conditions, the telemetry and control host 40 outputs a saturation state confirmation signal only when the absolute values ​​of the first-order time derivatives of N consecutive frames are all less than the ϵ threshold. Furthermore, the telemetry and control host 40 can also combine the fluctuation amplitude of the global information entropy for joint judgment: when the consecutive... When the difference between the maximum and minimum values ​​of the intra-frame global information entropy is less than a preset stability threshold, and the convergence condition of the first-order time derivative is simultaneously met, the telemetry and control host 40 outputs a saturation state confirmation signal. This joint criterion can reduce the probability of misjudgment caused by occasional reflections, image noise, or local droplet disturbances. This is a threshold parameter for the number of consecutive frames, which can be set according to the actual frame rate, for example, from 10 to 30 frames.

[0074] After the control and measurement host 40 outputs a saturation state confirmation signal, it synchronously triggers data latching. The control and measurement host 40 reads the temperature value inside the environmental test chamber 10 at this time and records it as the current temperature parameter. Simultaneously, the monitoring and control host 40 retrieves internal timer data, calculates the time elapsed from the start of sending the temperature and humidity adjustment command to the output of the saturation state confirmation signal, and records this time as the saturation time parameter. .

[0075] If the monitoring and control host 40 enters the subsequent dehumidification process due to the timeout status flag, the temperature parameter latched at the timeout moment will be used as the equivalent current temperature parameter. The time elapsed from the start of sending the temperature and humidity adjustment command to the timeout period is used as the equivalent saturation time parameter. And set the timeout status flag to the equivalent current temperature parameter. Equivalent saturation time parameter Store them together.

[0076] During subsequent evaporation phase acquisition, the control host 40 also records the measured temperature parameters corresponding to the phase acquisition moment in real time. Current temperature parameters Thermodynamic boundary conditions used to characterize the coating when it reaches saturation or an equivalent high-humidity state; saturation time parameter Used to characterize the temporal characteristics of water vapor entering the micropores or surface structure of a coating; measuring temperature parameters. These parameters are used for subsequent calculations of normal thermal expansion offset. They are stored in the measurement and control host 40 as input data for subsequent thermomechanical coupling error stripping and hygroscopic optical path error compensation.

[0077] Reference Figure 5 In step S30, the monitoring and control host 40 may execute sub-steps S301 to S303 in sequence.

[0078] S301, the monitoring and control host 40 sends a state switching command to the environmental test chamber 10, causing the liquid phase moisture on the coating surface to enter the evaporation and shrinkage process.

[0079] The monitoring and control host 40 determines the dehumidification parameters according to the preset dehumidification control logic. For example, the monitoring and control host 40 can determine the dehumidification control amount of the environmental test chamber 10 based on the current temperature parameters, relative humidity parameters, and enthalpy-humidity relationship under normal atmospheric pressure.

[0080] Specifically, the process by which the monitoring and control host 40 determines the specific amount of humidity control in the environmental test chamber 10 includes the following steps: Calculate the current absolute moisture content: based on the current temperature parameter (Unit: °C) and relative humidity parameters (Percentage), combined with conventional atmospheric pressure The actual water vapor partial pressure of the current air was calculated using the Magnus empirical formula. and absolute moisture content (Unit: g / kg): ; ; ; in, This is the saturated vapor pressure at the current temperature (unit: hPa).

[0081] Determine the dehumidification control value: The monitoring and control host presets the target moisture content corresponding to the target dehumidification state. Calculate the current moisture content deviation Subsequently, a conventional proportional-integral (PI) control algorithm is used to convert this deviation into the dehumidification control quantity of the environmental test chamber's dehumidification actuator (such as a dehumidification impeller or exhaust valve). : ; in, , These are the preset proportional and integral coefficients, respectively. Through the above enthalpy-humidity calculation and closed-loop control, precise dynamic output of the dehumidification control quantity is achieved.

[0082] Subsequently, the control host 40 triggers the dehumidification hardware control valve of the environmental test chamber 10, controlling the dehumidification wheel, exhaust damper, and heating module to work together. In one embodiment, the internal environmental parameters of the environmental test chamber 10 switch from a high-humidity saturation state or a near-dew-point high-humidity substitution state to a high-temperature, low-humidity dehumidification state within a set time window of 5 to 10 minutes.

[0083] The aforementioned environmental change disrupts the original gas-liquid dynamic balance on the coating surface, causing liquid water adhering to the micropores, rough structures, or surface depressions of the coating to begin evaporation. During the evaporation process, the industrial camera 30 maintains a set high frame rate acquisition mode, continuously acquiring reflective images of the coating surface. The measurement and control host 40 establishes a time sliding window to temporarily store multiple consecutive frames of image data for subsequent grayscale fluctuation analysis.

[0084] S302, the measurement and control host 40 calculates the gray-scale variance based on the image data within the time sliding window and constructs the evaporation gradient exponential field.

[0085] To reduce the impact of structured light patterns, variations in projected light intensity, or variations in fringe spatial frequency on the calculation of grayscale variance, the measurement and control host 40 preferably uses uniform illumination images, fringe image sequences with the same phase shift number, or reflection intensity image sequences after normalization of the projection pattern to calculate the evaporation gradient exponential field.

[0086] When the liquid phase evaporates from the coating surface, its optical reflection state gradually transitions from directional reflection under the liquid phase to diffuse reflection from a dry, rough surface. This transition causes pixel grayscale to fluctuate over time. Based on this characteristic, the measurement and control host 40 extracts time-series grayscale changes at the pixel level to characterize the degree of evaporation activity at different locations on the coating surface.

[0087] The frame length of the time sliding window set by the monitoring and control host 40 is... The discrete variance of the grayscale values ​​within the window is calculated pixel by pixel, and this discrete variance is used as the evaporation gradient exponential field.

[0088] When the evaporation gradient exponential field is calculated based on the striped image sequence, the measurement and control host 40 selects image frames corresponding to the same phase shift step number for time dimension comparison; alternatively, the measurement and control host 40 first performs brightness normalization processing on the image grayscale under different projection parameters, and then calculates the evaporation gradient exponential field. If the projected light intensity of the projector 20 changes between adjacent phase shift sequences, the measurement and control host 40 normalizes and corrects the image grayscale based on the projected light intensity, exposure time, camera gain, or grayscale value of the reference white board area, so that the grayscale variance mainly reflects the change in reflection state caused by liquid phase evaporation.

[0089] Its mathematical calculation model can be expressed as: ; in, Representing pixels At any moment The evaporation gradient index. This represents the length of the time sliding window. Representing pixels At any moment The captured grayscale value. This represents the pixel at time t. At the time The average grayscale value within the time window.

[0090] The settings can be adjusted based on the acquisition frame rate, evaporation rate, and image noise level. In one embodiment, Select 5 to 15 frames to balance transient response and noise suppression capabilities.

[0091] Evaporation gradient exponential field This characterizes the intensity of evaporation state changes at various locations on the coating surface at the current moment. Grayscale fluctuations are smaller in regions where the liquid phase is stable or completely dry. Approaching zero; while in the region at the edge of the gas-liquid interface recession, due to high-frequency oscillations in the direction of light scattering, It exhibits local peak values.

[0092] S303, the measurement and control host 40 adjusts the fringe spatial frequency and projected light intensity of the projector 20 according to the evaporation gradient exponential field.

[0093] The adjustment is performed between two adjacent sets of complete phase-shifted image acquisition sequences. During the acquisition of the same set of multi-step phase-shifted images, the fringe spatial frequency, phase-shift step size, and projected light intensity of the projector 20 remain unchanged to ensure consistent phase-shift solution conditions.

[0094] In the initial stage of evaporation, residual liquid films or droplets on the coating surface are prone to specular reflection, leading to localized overexposure of the industrial camera and loss of phase information. In the later stage of evaporation, the coating surface gradually recovers its dry, rough morphology. At this point, structured light fringes with higher brightness and a preset reference measurement frequency are required to obtain morphology data with a higher signal-to-noise ratio. The preset reference measurement frequency is the structured light measurement frequency calibrated by the system under dry or near-dry surface conditions. It can be a medium to high spatial frequency and is not limited to low frequency.

[0095] The control and measurement host 40 performs spatial integration on the evaporation gradient exponential field Ex,y,t within the effective field of view to obtain a scalar index characterizing the current global evaporation state, and calculates the fringe spatial frequency and projected light intensity used in the next set of phase-shifted image acquisition sequences based on this index. Its control model can be expressed as: ; ; in, Representative moment The dynamic spatial frequency of the projector stripes. Representative moment The light intensity projected by the emitting surface of the projector 20. The effective integration region representing the entire field of view. The system's preset reference measurement frequency parameters, This represents the set frequency domain closed-loop modulation coefficient. This represents the maximum light intensity that the projector 20 hardware is allowed to output. This represents the set closed-loop modulation coefficient of the light intensity. and It represents a spatial integral element.

[0096] In discrete image processing, the above spatial integration can be implemented over the effective pixel region. The values ​​are summed. The larger the integral value of the evaporation gradient exponent field within the effective field of view, the stronger the current evaporation activity, and the lower the projected light intensity output by the measurement and control host 40, so as to reduce the risk of overexposure in the droplet or liquid film area; the smaller the integral value of the evaporation gradient exponent field within the effective field of view, the weaker the evaporation activity, and the closer the projected light intensity output by the measurement and control host 40 is to the preset high light intensity value.

[0097] It can be obtained through static calibration at room temperature, and can be determined based on the resolution of the projector 20, the sampling frequency of the industrial camera 30, the surface roughness of the coating, and the Nyquist sampling constraint. The value of is limited to the highest spatial frequency of the modulated stripe not exceeding the system's allowed sampling frequency. The value of is limited to ensuring that the intensity of the projected light is not lower than the minimum effective light sensitivity threshold of 30 for industrial cameras while suppressing overexposure.

[0098] The measurement and control host 40 will calculate the Convert the raster generation parameters of the projector 20, and... After being converted into a drive control signal for the projector 20, the signal is sent to the projector 20. At the same time, the measurement and control host 40 records the fringe spatial frequency, phase shift step size, and projected light intensity corresponding to each complete phase-shifted image acquisition sequence, and uses these as the acquisition metadata for that set of images.

[0099] When performing phase calculation and phase height mapping, the telemetry and control host 40 calls the phase height mapping function that matches the current fringe spatial frequency; or, the telemetry and control host 40 first converts the absolute phase data obtained at different fringe spatial frequencies into equivalent phase data in a unified projected coordinate system, and then performs phase height mapping.

[0100] To prevent the control quantity from exceeding the equipment's operating range, the monitoring and control host 40 performs a check before outputting the command. and Amplitude limiting is applied. Specifically, It is limited to the frequency range within which the projector 20 and the industrial camera 30 jointly meet the sampling requirements; It is limited to between the minimum effective light sensitivity threshold and the overexposure threshold of an industrial camera.

[0101] Through the aforementioned control method, the measurement and control host 40 reduces the projected light intensity and adjusts the fringe spatial frequency to improve the observability of the evaporation boundary region when evaporation activity intensifies; as evaporation nears its end, it restores high light intensity and the reference measurement frequency to obtain high signal-to-noise ratio morphological data of the dry region. The adjustment of the fringe spatial frequency is used to improve the image observability of the evaporation boundary, but does not indicate that this region has a higher unfolding priority in the subsequent phase unfolding process; during the phase unfolding stage, the measurement and control host 40 still reduces the unfolding weight of the dynamic droplet edge region according to the quality map function.

[0102] Reference Figure 6 In step S40, the monitoring and control host 40 may execute sub-steps S401 to S403 in sequence.

[0103] S401, the measurement and control host 40 extracts the temporal extrema of the evaporation gradient exponential field, constructs the extremum time field, and generates the evaporation kinetic scalar field based on the extremum time field.

[0104] During the liquid-phase evaporation and shrinkage process, the gas-liquid interface on the coating surface sequentially passes through different pixel positions. When the gas-liquid interface passes through a certain pixel, the reflection and scattering states at that pixel change significantly, corresponding to the evaporation gradient index. It typically forms a peak in the time series.

[0105] The monitoring and control host 40 determines the threshold based on effective evaporation. Pixels that have undergone effective evaporation phase transition are selected. A time sliding window is used to calculate the evaporation gradient exponential field at the current moment in real time. For the construction of the extreme value time field, the measurement and control host 40 continuously records the historical maximum value of the evaporation gradient exponent and its corresponding timestamp for each pixel throughout the entire liquid phase evaporation observation period. When the newly calculated evaporation gradient exponent is greater than the historical maximum value recorded for that pixel, the measurement and control host 40 updates the historical maximum value and corresponding timestamp for that pixel.

[0106] When the liquid phase evaporation observation period ends, or when the preset stop condition is met, the measurement and control host 40 generates an extreme value time field based on the timestamp corresponding to the final historical maximum value recorded by each pixel. Specifically, only when the maximum value of the evaporation gradient exponent of a pixel in the time series is greater than the effective evaporation judgment threshold... When the maximum value is less than or equal to the effective evaporation judgment threshold, the monitoring and control host 40 extracts the timestamp corresponding to the maximum value and writes it into the extreme value time field; when the maximum value is less than or equal to the effective evaporation judgment threshold, the host 40 extracts the timestamp corresponding to the maximum value and writes it into the extreme value time field. At that time, the monitoring and control host 40 marks the pixel as an invalid pixel or a static pixel.

[0107] Its computational model can be expressed as: ; in, Representing pixels Extreme time field distribution at the location; The time variable representing the time at which the evaporation gradient exponent reaches its maximum value. ; The effective evaporation judgment threshold can be determined based on 3 to 5 times the variance of the background noise of static images acquired by an industrial camera 30 in a darkroom environment. This represents the flag value assigned to invalid or static pixels, used to isolate regions where no effective phase transition has occurred in subsequent spatial gradient calculations. (The above...) The time range corresponding to the operation is the preset liquid phase evaporation observation period, or the effective analysis period from the start of the self-dehumidification state until the overall evaporation gradient exponential field is lower than the preset stability threshold, rather than a single short time sliding window.

[0108] After constructing the extreme time field, the measurement and control host 40 first performs measurements on the extreme time field. Smoothing preprocessing is performed to reduce the impact of isolated noise points on spatial gradient calculations. Smoothing preprocessing can employ Gaussian filtering, median filtering, or other local smoothing methods. For those labeled as... Invalid or static pixels are excluded by the measurement and control host 40 as mask areas during filtering and gradient calculation, or they are used to perform local interpolation using the extreme time data of surrounding valid pixels before being included in the calculation.

[0109] Subsequently, the telemetry and control host 40 calculates the spatial gradient magnitude of the extreme time field in the two-dimensional image coordinate system, obtaining the scalar field of evaporation kinetics. Its calculation model can be expressed as: ; in, Represents the scalar field of evaporation dynamics. and These represent the extreme time fields at... direction and The first-order partial derivative of the direction in the discrete pixel coordinate system can be approximated by the Sobel operator, the central difference operator, or other gradient operators.

[0110] Evaporation kinetics scalar field Used to characterize the extreme time variation gradient, propagation hysteresis, and propagation non-uniformity of the liquid phase retreat process. Because Characterized by the spatial gradient amplitude of the extreme time field, a larger value indicates a greater variation in extreme time within a unit spatial distance, and a slower or more uneven recession process at the gas-liquid interface in that region; conversely, a smaller value indicates a smaller variation in extreme time within a unit spatial distance, and a faster or more uniform recession process. Therefore, It does not directly represent the liquid phase retreat velocity itself, but is used to characterize the spatial hysteresis characteristics of the liquid phase retreat process; when an approximate value of the propagation velocity is needed, the measurement and control host 40 can perform the conversion based on the reciprocal of the spatial gradient of the extreme time field or the calibrated conversion relationship.

[0111] S402, the measurement and control host 40 constructs a phase expansion mass diagram function based on the scalar field of evaporation kinetics.

[0112] In conventional structured light 3D measurements, the reliability of phase data is typically evaluated by relying on fringe modulation. However, during liquid-phase evaporation, spurious high-modulation regions may form at droplet edges or water film boundaries due to abrupt changes in refractive index. If the phase unwrapping path directly passes through such regions, local phase jumps can easily be diffused into global unwrapping errors.

[0113] To this end, the measurement and control host 40 introduces an evaporation kinetic scalar field. As physical prior weights, surface phase transition dynamics information and fringe modulation information are used together for phase quality evaluation. The quality map function can be expressed as: ; in, A quality graph function representing the fused spatiotemporal characteristics. This represents the conventional fringe modulation matrix calculated from a multi-step phase-shifted image, and its value can be normalized to the range of 0 to 1. For evaporation kinetics scalar field.

[0114] Before importing the mass map function, the measurement and control host 40 can perform scalar field analysis on the evaporation kinetics. Normalization is performed to map the values ​​to the range of 0 to 1; alternatively, it can be based on... The actual dimension is set to the dynamic decay constant λ, so that the exponential term is kept within the preset quality evaluation range.

[0115] and This represents the set balance weight coefficient, with the following constraints: The weighting coefficients can be pre-calibrated using standard test specimens or calibration samples, for example, in a water vapor evaporation scenario. 0.4 is acceptable. 0.6 is acceptable. Represents the kinetic decay constant, used to adjust the penalty intensity of the evaporation kinetic scalar field on the quality evaluation results; According to The distribution range can be set, for example, from 0.1 to 1.0.

[0116] By using the aforementioned quality graph function, the measurement and control host 40 reduces the quality score of the high-dynamic evaporation edge region when evaluating phase reliability, thus preventing false high-profile regions from being misjudged as priority deployment regions.

[0117] S403, the telemetry and control host 40 guides the phase unfolding path according to the quality map function to obtain the absolute phase data of the entire field of view.

[0118] The measurement and control host 40 will display the quality graph function. The input spatial phase unfolding algorithm uses pixels with higher quality evaluation scores as seed points for unfolding and generates phase unfolding paths in descending order of quality evaluation scores. In one implementation, the measurement and control host 40 generates a minimum spanning tree path based on a quality graph function and advances the wrapping phase unfolding along this path.

[0119] During the unfolding process, the path prioritizes covering areas where the surface is already dry and stable, and avoids low-quality areas at the edge of dynamic evaporation and shrinkage, thereby reducing the impact of droplet refraction dead angles, dynamic water film boundaries, or local strong reflective areas on the global phase unfolding results.

[0120] For pixel regions with quality evaluation scores below a preset threshold, the measurement and control host 40 can adopt a delayed unfolding strategy: first, complete the phase unfolding of the surrounding high-quality regions, and then combine the multi-frequency heterodyne phase, the phase continuity constraint of adjacent unfolded regions, the local smoothing constraint, or the time series phase prediction results to perform secondary unfolding, interpolation repair, or confidence marking on the low-quality regions.

[0121] When the number of effective evaporation pixels is lower than a preset ratio, or when the evaporation kinetic scalar field of a certain region cannot be reliably generated, the measurement and control host 40 can set the kinetic weight term of that region to a neutral weight value, or degrade it by using a conventional stripe modulation matrix. This serves as the primary input to the quality map function. Therefore, the phase unfolding process can continue even when evaporation characteristics are not obvious or local data is insufficient.

[0122] The queue management, path optimization, and phase accumulation processes of the minimum spanning tree phase expansion algorithm can be implemented using graph theory phase expansion methods commonly used in this field, and will not be elaborated here.

[0123] After the phase unfolding process guided by the above path, the telemetry and control host 40 outputs full-field absolute phase data. For pixel regions that have undergone delay unwrapping, interpolation repair, or low-confidence labeling, the measurement and control host 40 synchronously outputs a phase quality label matrix. When subsequently calculating the relative thickness change, the measurement and control host 40 can use this phase quality label matrix to deweight, remove, or individually label low-confidence regions.

[0124] Full field of view absolute phase data As input data for step S50, it is used to calculate the relative thickness change of the coating sample relative to the initial reference state by combining the current anchor point reference phase, the initial differential reference topography field, the thermal expansion compensation amount, and the refractive index compensation characteristic term.

[0125] Reference Figure 7 In step S50, the monitoring and control host 40 may execute sub-steps S501 to S506 in sequence.

[0126] In the following steps, unless otherwise stated, Indicates the current phase acquisition time. This represents the pixel coordinate index of industrial camera 30.

[0127] S501, the telemetry and control host 40 extracts the current anchor point reference phase and calculates the dynamic phase drift.

[0128] After outputting the full field-of-view absolute phase data in step S40, the telemetry and control host 40 extracts the current absolute phase value in the pixel area of ​​the reference anchor point 60 in the current image, calculates its spatial average value, and obtains the reference phase of the current anchor point.

[0129] The formula is: ; in, Representative moment The current anchor point reference phase at lower reference anchor point 60; Representative moment The area of ​​the lower reference anchor point 60 in the 30-pixel coordinate system of the industrial camera; Representative area The total number of valid pixels within; Representative moment Next pixel The full field of view absolute phase data.

[0130] The control host 40 compares the current anchor point reference phase with the initial anchor point reference phase established in step S10 to obtain the dynamic phase drift of the reference anchor point 60.

[0131] The formula is: ; in, The amount of dynamic phase drift representing the reference anchor point 60; This represents the initial anchor point reference phase established in step S10.

[0132] The dynamic phase drift is used to characterize the gas state inside the observation window 11 and the environmental test chamber 10, the minute displacement of the sample clamp 50, and the common-mode phase change in the same optical path caused by thermal drift of the optical system. The dynamic phase drift can be used as a monitoring quantity, alarm quantity, or quality evaluation quantity of the system drift state, and can also be equivalently involved in the calculation of the current differential phase data.

[0133] When the reference anchor point 60 includes multiple anchor point units, the telemetry and control host 40 calculates the dynamic phase drift of each anchor point unit and fits the spatial drift field according to the spatial position of the multiple anchor point units in the field of view; the spatial drift field is used to perform position correlation correction on the current absolute phase data.

[0134] In practical implementation, the measurement and control host 40 can determine the pixel area of ​​the current reference anchor point 60 through a high-contrast marked border on the surface of the reference anchor point 60, a preset connected component mask, or a feature point recognition algorithm. To reduce the impact of abnormal reflections, blemishes, or partial occlusions on anchor point phase extraction, the measurement and control host 40 can... After outlier removal, median filtering, or weighted averaging of the phase data, the current anchor point reference phase is calculated.

[0135] S502, the measurement and control host 40 performs dynamic differential correction and phase height mapping to generate the current differential three-dimensional topography field.

[0136] Due to the current anchor point reference phase Includes initial anchor reference phase and dynamic phase drift Therefore, the telemetry and control host 40 will compare the full field-of-view absolute phase data with the current anchor point reference phase. By performing differential calculations, the anchor point reference phase and dynamic phase drift component can be simultaneously subtracted under the current measurement state. The telemetry and control host 40 performs differential calculations between the full-field-of-view absolute phase data and the current anchor point reference phase to obtain the current differential phase data relative to the reference anchor point 60. Equivalently, the telemetry and control host 40 can also first subtract the initial anchor point reference phase from the full-field-of-view absolute phase data, and then further subtract the dynamic phase drift amount to obtain the current differential phase data. The calculation model is as follows: ; in, Representing pixels At any moment The current differential phase data relative to the reference anchor point 60.

[0137] Equivalently, the telemetry and control host 40 can also first subtract the initial anchor point reference phase from the full field of view absolute phase data, and then further subtract the dynamic phase drift to obtain the current differential phase data.

[0138] Subsequently, the telemetry and control host 40 uses a pre-calibrated phase height mapping function to map the current differential phase data into the current differential three-dimensional topographic field. The calculation model is as follows: ; in, Represents the current differential three-dimensional topographic field; This represents the phase height mapping function. When the projector 20 uses different fringe spatial frequencies at different times, the phase height mapping function is further correlated with the current fringe spatial frequency. The measurement and control host 40 calls the corresponding frequency-related phase height mapping function according to the fringe spatial frequency ft corresponding to the current acquisition sequence, or converts the current absolute phase data into equivalent phase data under a unified frequency scale before performing phase height mapping.

[0139] The current differential three-dimensional topography field characterizes the normal height distribution of the coated sample surface relative to the reference anchor point 60 under the current environmental conditions.

[0140] Before performing the differential operation, the measurement and control host 40 can perform rigid registration, affine registration, or perspective registration between the current image and the initial image based on the marked border of the reference anchor point 60, the sample edge contour, preset calibration points, or image feature points, so that the current differential three-dimensional topography field Compared with the initial differential reference topography field They are in the same pixel coordinate system or the same physical coordinate system.

[0141] The measurement and control host 40, based on the coordinate transformation relationship obtained from registration, resamples the current differential 3D topography field to the unified coordinate grid where the initial differential reference topography field is located, or resamples the initial differential reference topography field to the current measurement coordinate grid before performing differential operations. For scale changes caused by thermal expansion or small deformations of the fixture, the measurement and control host 40 can use affine registration, perspective registration, or non-rigid registration methods that include local deformation constraints to reduce misalignment differential errors.

[0142] Specifically, to achieve high-precision pixel-level alignment, the measurement and control host 40 can employ affine registration, perspective registration, or non-rigid registration methods. As a preferred embodiment, this solution uses a perspective registration model to eliminate misalignment differential errors.

[0143] Set the homogeneous coordinates of the pixels in the image to be registered as follows: The homogeneous coordinates of the corresponding pixel in the reference image are: The perspective registration mathematical transformation model used by the telemetry and control host is as follows: ; in, The scaling factor is the 3×3 matrix in the middle. This is the perspective transformation matrix (homography matrix).

[0144] In the specific calculation, the telemetry and control host extracts corresponding feature points (such as the aforementioned anchor points) from the image and solves for the transformation matrix using the Direct Linear Transform (DLT) algorithm or the Random Sample Consensus (RANSAC) algorithm. The various coefficients are then used to perform global resampling and spatial interpolation on the target image, accurately aligning the image to the reference coordinate system, thereby effectively eliminating misalignment errors caused by environmental vibrations or sample thermal expansion.

[0145] This registration process avoids misalignment errors caused by minute displacement of the sample fixture, thermal deformation of the sample, or camera field of view drift.

[0146] S503, the measurement and control host 40 calculates the uncompensated surface normal change.

[0147] The measurement and control host 40 performs algebraic difference between the current differential 3D topography field and the initial differential reference topography field established in step S10 to obtain the uncompensated surface normal variation. Its calculation model is as follows: ; in, This represents the uncompensated change in surface normal. Represents the current differential three-dimensional topographic field; This represents the initial differential reference topography field established in step S10.

[0148] The uncompensated surface normal variation includes the actual coating thickness variation, the overall normal displacement caused by substrate thermal expansion, the refractive index shift caused by liquid phase residue or pore moisture absorption, and the residual optical path drift component.

[0149] S504, the measurement and control host 40 calculates the normal thermal expansion offset.

[0150] The measurement and control host 40 reads the reference temperature parameters corresponding to the establishment of the initial reference in step S10, and reads the measurement temperature parameters corresponding to the phase acquisition in step S40. The measurement and control host 40 calculates the normal thermal expansion offset of the sample based on the thermal expansion coefficient of the substrate, the normal reference dimension of the substrate, and the temperature difference.

[0151] Its computational model can be expressed as: ; in, Represents the offset of normal thermal expansion; Represents the coefficient of thermal expansion of the substrate; Represents the normal reference dimension of the substrate; The measured temperature parameter represents the moment of phase acquisition; The reference temperature parameter represents the initial reference temperature parameter established in step S10.

[0152] In a preferred embodiment, when the thermal expansion of the coating material has a significant impact on the measurement results, the measurement and control host 40 can further incorporate the coating's thermal expansion coefficient and initial coating thickness to establish a substrate-coating composite thermal expansion compensation model. When the reference anchor point 60 is fixedly installed on the sample fixture 50 rather than on the surface of the coated sample substrate, or when there is a normal thermal displacement difference between the sample fixture 50 and the coated sample, the measurement and control host 40 can also calculate the fixture thermal expansion compensation term based on the thermal expansion coefficient of the sample fixture 50, the normal reference distance from the anchor point installation position to the sample bearing surface, and the temperature difference, and incorporate this fixture thermal expansion compensation term into the normal thermal expansion offset. The calculation model can be expressed as: ; in, Represents the coefficient of thermal expansion of the coating; This represents the initial coating thickness at the corresponding pixel location of the coated sample in its initial state. (Initial coating thickness) It can be obtained from pre-input sample nominal thickness, contact thickness gauge test results, cross-sectional microscopic test results, standard sample calibration data, or other known thickness measurement results.

[0153] When there is no need to consider the thermal expansion of the coating itself, or when the thermal expansion of the coating is much smaller than the normal thermal expansion of the substrate, the measurement and control host 40 can use only the substrate thermal expansion compensation item.

[0154] S505, the measurement and control host 40 calculates the refractive index compensation characteristic term.

[0155] The measurement and control host 40, based on at least one of the following parameters recorded in step S20: saturation time parameter, saturation temperature parameter, measurement temperature parameter at the phase acquisition time, relative humidity parameter, and coating material type, calls a pre-calibrated refractive index compensation model to calculate the refractive index compensation characteristic terms. The relative humidity parameter is obtained in real-time from the humidity sensor within the environmental test chamber 10; the coating material type parameter can be obtained through manual input, sample code reading, or by calling the material database built into the measurement and control host 40; the initial surface roughness parameter of the coating sample can be calculated based on the initial differential reference topography field obtained in step S10, or it can be obtained from an external roughness meter, a white light interferometer, or pre-input sample detection data.

[0156] Its general expression can be written as: ; in, Represents the refractive index compensation characteristic term; This represents a pre-calibrated refractive index compensation model; Represents the saturation time parameter; The temperature parameter represents the saturation state. The measured temperature parameter represents the moment of phase acquisition; Represents the relative humidity parameter; Parameters representing the coating material category; This represents the initial surface roughness parameter of the coated sample.

[0157] This refractive index compensation feature is used to characterize the offset caused by liquid phase residue, pore moisture absorption, water film refractive index changes, or local water content on the coating surface to the high mapping result of the structured light phase.

[0158] In practical implementation, the refractive index compensation model can be a standard sample calibration curve, a lookup table, a linear regression model, a polynomial fitting model, or a machine learning regression model. The refractive index compensation model can be pre-established using standard coated samples. Specifically, the calibration conditions of the standard coated samples cover a preset temperature range, relative humidity range, saturation time range, coating material type range, and surface roughness range. The actual height change or actual thickness change of the standard coated samples can be obtained using a contact thickness gauge, white light interferometer, laser displacement sensor, standard displacement stage, or a standard sample with known thickness. The measurement and control host 40 uses the difference between the apparent height offset measured by structured light and the actual height change as the refractive index compensation value, and establishes a lookup table, fitting curve, or regression model based on this refractive index compensation value. Specifically, the measurement and control host 40 collects structured light phase data under different saturation times, temperatures, humidity levels, and material types on standard samples with known coating thickness changes or known surface height changes, and uses the difference between the apparent height offset measured by structured light and the actual height change of the standard sample as the refractive index compensation value for fitting or tabulation.

[0159] For example, in a simplified implementation, the refractive index compensation characteristic term can be expressed using a linear model as follows: ; in, These are the model coefficients obtained by pre-calibrating using standard coated samples.

[0160] In another implementation, the refractive index compensation characteristic term can also be expressed using a lookup table model as follows: ; in, This represents a pre-established refractive index compensation lookup table.

[0161] Therefore, the saturation time parameter does not determine the refractive index compensation characteristic term alone, but rather participates in the compensation calculation as one of the input characteristics characterizing the coating's moisture absorption and water vapor permeability.

[0162] S506, the measurement and control host 40 outputs the relative thickness change of the coating sample.

[0163] Before calculating the relative thickness change, the measurement and control host 40 can determine the effective measurement area based on the evaporation gradient exponent field, phase quality marker matrix, or image grayscale stability parameters. For areas where the evaporation gradient exponent is lower than a preset stability threshold and the phase quality is higher than a preset quality threshold, the measurement and control host 40 uses these areas as the effective thickness calculation area. For areas that still have liquid phase coverage, strong reflection, phase jumps, or low-confidence quality markers, the measurement and control host 40 performs delayed calculations, weight reduction processing, interpolation repair, or removal processing. The measurement and control host 40 will calculate the uncompensated surface normal change. Normal thermal expansion offset and refractive index compensation characteristic term Algebraic operations are performed to obtain the relative thickness change of the coating sample at each pixel location. The relative thickness change refers to the normal change of the coating surface relative to the initial reference state after deducting the thermal expansion of the substrate, the thermal drift of the fixture, and the hygroscopic refractive shift. Under the condition that the back of the coating sample substrate is fixed, the normal displacement of the substrate can be characterized by the thermal expansion model, or the normal displacement of the substrate-fixture system has been deducted through the reference anchor point 60 and the compensation model, this normal change is used to characterize the change in coating thickness relative to the initial state. Specifically, the measurement and control host 40 uses the uncompensated surface normal change... Subtracting the normal thermal expansion offset and refractive index compensation characteristic term This is to obtain the relative thickness change of the coating after peeling thermal expansion error and moisture absorption refractive error. When the measurement and control host 40 identifies that the phase quality of a certain area is lower than the preset threshold, the refractive index compensation model is outside the calibration range, or the registration residual is greater than the preset threshold, the measurement and control host 40 can add a low confidence mark to the relative thickness change of that area, or remove that area from the statistical process of average thickness change, maximum thickness change, minimum thickness change, and uniformity index.

[0164] The monitoring and control host 40 can further monitor... Perform spatial filtering, outlier removal, and regional statistical processing to output the full field-of-view distribution map of the relative thickness change of the coating sample, the average relative thickness change, the maximum relative thickness change, the minimum relative thickness change, and the thickness change uniformity index.

[0165] See attached document Figure 8Taking the polyurethane conformal coating test on the surface of a printed circuit board as an example, the measurement and control host 40 sets the initial state of the environmental test chamber 10 to a temperature of 25 degrees Celsius and a relative humidity of 50%. Based on the pre-input physical parameters of the polyurethane coating, the measurement and control host 40 calculates the target dew point temperature as 13.9 degrees Celsius. The measurement and control host 40 sends control commands to adjust the output power of the cooling and humidification modules inside the environmental test chamber 10, making the internal environmental parameters approach this dew point temperature. During this process, the industrial camera 30 continuously captures images at a fixed frame rate of 100 frames per second. The measurement and control host 40 calculates the global information entropy of the current frame image in real time. When the absolute value of the first-order time derivative of the global information entropy of 20 consecutive frames is less than the set minimum judgment threshold, (In this embodiment) When the value is 0.02, the control host 40 outputs a confirmation signal, indicating that the coating surface has reached the gas-liquid phase transition saturation state. At this time, the system latches the data and records the saturation time parameter. 125 seconds, current temperature parameter The temperature is 14.1 degrees Celsius.

[0166] The control and measurement host 40 then sends a state switching command to the environmental test chamber 10, activating the internal dehumidification wheel and raising the target temperature to 40 degrees Celsius, inducing transient evaporation of the liquid phase moisture adhering to the coating surface. The control and measurement host 40 establishes a length... Using a 10-frame time sliding window, the discrete variance of grayscale values ​​within the window is calculated pixel-by-pixel to generate an exponential evaporation gradient field. During the evaporation retreat phase, the system effectively covers the entire field of view. The evaporation gradient exponent within the space is subjected to double spatial integration. When the integral value reaches its peak, it indicates that the evaporation activity is at its most intense. Based on the feedforward control model, the control unit 40 projects the fringe spatial frequency of the projector 20... The basic low-frequency parameters are synchronously increased to the edge of the Nyquist sampling limit frequency allowed by the system, and the projected light intensity is... Suppress to maximum hardware light intensity 40% of the image is used to avoid overexposure dead zones on the photosensitive target surface caused by the specular reflection of discrete water droplets.

[0167] After completing the dynamic sequence acquisition of the transient evaporation process, the monitoring and control host 40 configures the effective evaporation judgment threshold. It is four times the variance of the camera's static dark-field noise floor. (This is for items that meet the following criteria.) The measurement and control host 40 extracts the timestamps corresponding to the maximum values ​​of the pixels to construct the extreme value time field. The monitoring and control host 40 calls the median filtering algorithm to... After smoothing preprocessing, the spatial two-dimensional gradient magnitude is calculated using central difference operator convolution, and then mapped to generate an evaporation kinetic scalar field. .

[0168] During the de-aliasing operation phase, the telemetry and control host 40, in conjunction with the conventional stripe modulation matrix, performs the operation. With scalar field of evaporation kinetics Constructing a quality map function In this embodiment, the balancing weight coefficient Set to 0.35. Set to 0.65, kinetic decay constant The value is set to 0.5. The telemetry and control host 40 executes the spatial phase unfolding algorithm to... The pixel with the highest score is used as the starting seed point to generate the minimum spanning tree path. Calculation feedback shows that pixels located at the edges of discrete water droplets... The numerical values ​​exhibit abrupt changes, and their quality evaluation scores are severely penalized by the exponential function. The algorithm automatically bypasses these optical refraction blind spots during its progression. For the very few un-unoverlapped regions that are ultimately isolated, the telemetry and control host 40 marks them as invalid masks and retrieves valid phase data from the periphery to perform bilinear interpolation repair, outputting continuous, unaffected droplet edge jumps across the entire field of view absolute phase data.

[0169] See attached document Figure 9 To verify the effectiveness of the phase transition transient triggering and optical frequency domain adaptive reconstruction technology and the combined phase de-overlap algorithm driven by evaporation kinetics provided in this invention, a comparative experiment was conducted on five groups of polyurethane coating samples from the same batch on the same hardware platform. The traditional static single-frequency structured light combined with modulation-guided phase unfolding method (control group) and the scheme of this invention (experimental group) were used. The root mean square error was calculated based on high-precision offline calibration data of the coating's three-dimensional morphology obtained by laser confocal microscopy under extremely dry conditions at room temperature.

[0170] Experimental data show that in the initial stage of pure drying without condensation, the measurement accuracy of the experimental and control groups is basically the same, with the root mean square error (RMSE) of morphology reconstruction remaining at around 1.5 μm. When the environment enters a high-humidity condensation state, a continuous or semi-continuous water film forms on the coating surface. In the control group, due to localized strong reflection and overexposure, some interference fringe phase information is completely lost, and the RMSE rises to 6.8 μm. The experimental group, by significantly suppressing the transmitted light intensity through a feedforward closed loop, stabilizes the RMSE at 2.4 μm. In the most intense stage triggering transient evaporation, a large number of discrete, receding water droplets form on the coating surface, and the refraction at the droplet edges causes a violent phase jump. In the control group, the unscrambling path, misled by spurious modulation, directly penetrates the droplet edges, causing the global phase error to spread rapidly, and the RMSE of morphology reconstruction surges to 12.5 μm, resulting in severe spatial topological distortion in the test results. The experimental group relies on the scalar field of evaporation kinetics. The constructed penalty weighting mechanism successfully cuts off the propagation path of errors across the refraction dead angle in terms of mathematical logic, and the root mean square error of topography reconstruction is effectively converged to 2.1 micrometers.

[0171] Through numerical calculations and experimental comparisons of the above specific embodiments, this solution effectively overcomes the optical interferometry measurement bottlenecks caused by overexposure of water droplet reflection and refraction dead angles in complex temperature and humidity alternating environments and drastic gas-liquid phase transition transient processes. Those skilled in the art can determine, based on the above numerical boundaries and experimental verification logic, that this solution significantly suppresses in-situ online measurement errors of the coating's microscopic three-dimensional morphology under dynamic alternating environments, and possesses sufficient engineering feasibility.

[0172] 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 method for detecting coating thickness based on visual monitoring, characterized in that, Includes the following steps: The test environment was controlled to induce transient liquid-phase evaporation on the coating surface, and an image sequence of the coating surface was acquired under structured light projection. The evaporation gradient exponential field of the coating surface is calculated based on the image sequence, and the timestamps corresponding to the time series maxima of the evaporation gradient exponential field are extracted to construct the extreme value time field. Perform spatial gradient calculations on the extreme time field to generate a scalar field of evaporation kinetics; The mass map function is constructed by fusing the evaporation kinetic scalar field with the structured light fringe modulation matrix. The spatial phase unfolding algorithm path is guided by the quality map function to obtain the full field-of-view absolute phase data of the coating surface, and the coating thickness is calculated based on the full field-of-view absolute phase data.

2. The coating thickness detection method based on visual monitoring according to claim 1, characterized in that, The controlled test environment conditions to induce transient liquid-phase evaporation on the coating surface include: The temperature and humidity of the test environment are adjusted to approximate the target dew point temperature of the coating, and the global information entropy of the acquired image sequence is calculated in real time. When the absolute value of the first time derivative of the global information entropy of multiple consecutive frames is less than the set minimum judgment threshold, the coating surface is determined to have reached the gas-liquid phase transition saturation state. The target temperature of the test environment was then increased and dehumidification was activated to induce transient evaporation of liquid phase water.

3. The coating thickness detection method based on visual monitoring according to claim 2, characterized in that, During the transient evaporation of the liquid phase water, the method further includes: The state of intense evaporation is determined based on the spatial integral value of the evaporation gradient index within the effective area of ​​the entire field of view. When the spatial integral value reaches its peak, the spatial frequency of the stripes of the structured light projection is simultaneously increased, and the intensity of the projected light is reduced to avoid overexposure of the photosensitive target surface caused by discrete droplets.

4. The coating thickness detection method based on visual monitoring according to claim 1, characterized in that, The calculation of the evaporation gradient exponential field of the coating surface based on the image sequence includes: Establish a time sliding window with a set frame length for the acquired image sequence; For each pixel in the image sequence, the discrete variance of grayscale within the current time sliding window is calculated pixel by pixel; The grayscale discrete variance is used as the evaporation gradient exponent of the corresponding pixel at the current time to generate the evaporation gradient exponent field containing the time dimension and the two-dimensional spatial dimension.

5. The coating thickness detection method based on visual monitoring according to claim 1, characterized in that, The step of extracting the timestamps corresponding to the time series maxima of the evaporation gradient exponential field to construct the extreme value time field includes: An effective evaporation determination threshold is set, which is calibrated based on a multiple of the variance of the camera's static dark field background noise. For any pixel in the evaporation gradient exponent field, determine whether the maximum evaporation gradient exponent of the pixel in the time series is greater than the effective evaporation determination threshold. If the value is greater than the maximum evaporation gradient index, the timestamp corresponding to the maximum evaporation gradient index is extracted as the extreme time field value at that pixel. If the value is not greater than the maximum evaporation gradient index, it is determined that no gas-liquid phase transition has occurred at that pixel, and the pixel is assigned a set constant identifier value.

6. The coating thickness detection method based on visual monitoring according to claim 1, characterized in that, The step of performing spatial gradient calculations on the extreme time field to generate an evaporation kinetic scalar field includes: The filtering algorithm is called to perform smoothing preprocessing on the effective data region in the extreme time field, and the matrix containing discrete timestamps is transformed into a floating-point surface matrix with continuous transition. The spatial gradient magnitude of the extreme time field after smoothing preprocessing in the two-dimensional image pixel coordinate system is calculated using a difference operator, and the spatial gradient magnitude is mapped to the evaporation kinetic scalar field, which is used to characterize the spatial retreat propagation velocity characteristics of the gas-liquid interface.

7. The coating thickness detection method based on visual monitoring according to claim 1, characterized in that, The numerical construction logic for fusing the evaporation kinetic scalar field with the structured light fringe modulation matrix to construct the mass map function is as follows: The value of the quality map function is formed by adding the first part of the data and the second part of the data; The first part of the data is the product of the normalized structured light stripe modulation index matrix and the set first weight coefficient; The second part of the data is the product of the set attenuation penalty term and the set second weight coefficient, wherein the attenuation penalty term is an exponential function constructed with the natural constant as the base and the negative of the product of the evaporation kinetic scalar field and the kinetic attenuation constant as the exponent; Furthermore, the sum of the first weighting coefficient and the second weighting coefficient is one.

8. The coating thickness detection method based on visual monitoring according to claim 1, characterized in that, The spatial phase unfolding algorithm path guided by the quality map function includes: The quality map function is input into the minimum spanning tree algorithm. The pixel with the highest quality map function value is used as the starting seed point, and the minimum spanning tree path is generated in descending order of quality map function value. The un-overlap operation of the wrapped phase is advanced pixel by pixel along the minimum spanning tree path. Guided by the prior values ​​of the quality map function, the low-quality region determined by the evaporation kinetic scalar field to be at the edge of dynamic evaporation retreat is bypassed while maintaining physical connectivity.

9. The coating thickness detection method based on visual monitoring according to claim 8, characterized in that, The acquisition of the full field-of-view absolute phase data of the coating surface includes: For un-un ... The effective phase data from the periphery of the un-de-overlapped region is used for bilinear interpolation repair, and the full-field absolute phase data with phase jump interference removed is output.

10. A coating thickness detection system based on visual monitoring, characterized in that, include: An environmental test chamber is used to regulate the internal temperature and humidity to induce transient liquid-phase evaporation on the coating surface. A projector is used to adaptively project a sequence of structured light onto the coating surface; An industrial camera is used to continuously acquire image sequences of the coated surface under structured light projection at a fixed frame rate; The measurement and control host is communicatively connected to the environmental test chamber, the projector, and the industrial camera, respectively; the measurement and control host is configured to control the environmental test chamber, the projector, and the industrial camera to perform the coating thickness detection method based on visual monitoring as described in any one of claims 1 to 9.

Images