A multi-channel spectrum-based thin film thickness online detection method and system

Abstract

The application belongs to the technical field of thin film thickness detection, and discloses a kind of based on multi-channel spectrum's thin film thickness online detection method and system, the method includes: using multiple beams with continuous spectrum light beam irradiation thin film surface multiple different positions, and the reflection light formed by the collection of multiple-channel spectral image of thin film surface;Each channel of the multi-channel spectral image corresponds to each irradiated position of the thin film surface;The multi-channel spectral image is segmented according to channel, and the single-channel spectral image of each channel is extracted;According to each single-channel spectral image, the film thickness analysis algorithm is used to calculate the film thickness of each irradiated position;Film thickness calculation result is presented on the display interface;So as to improve the accuracy and reliability of density detection;So as to realize the synchronous online detection of multiple positions of thin film, significantly improve the detection efficiency and spatial resolution, and can improve the accuracy and stability of measurement.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of thin film thickness detection, in particular to a thin film thickness online detection method and system based on multi-channel spectrum. BACKGROUND

[0002] In the manufacturing process of OLED devices, the quality of the encapsulation thin film is directly related to the barrier performance, service life and light emission uniformity of the device, so it is very important to accurately control the film forming process. Especially in the thin film deposition stage, real-time online monitoring of the thin film thickness and its distribution uniformity has become a key link to improve production yield and product reliability. However, the existing film thickness detection methods have various limitations, which are difficult to meet the requirements of high stability, high spatial resolution and full-field measurement in modern high-performance OLED mass production process monitoring.

[0003] At present, the traditional technologies applied to thin film thickness measurement mainly include contact or offline measurement methods such as ellipsometry and step profiler. Such methods usually need to move the sample to a specific measurement environment, or analyze the thickness after the film forming process is completely finished, so they cannot realize real-time thickness feedback and regulation during the deposition process. In addition, the above methods generally have the problems of slow measurement speed, inability to reflect the dynamic changes of film thickness, and limited applicability to transparent or micro-nano scale thin films, especially they are difficult to evaluate the thickness uniformity of the thin film in the whole substrate range, thereby limiting their practical application in continuous production lines.

[0004] Another commonly used measurement method is to measure by optical film thickness meter based on the principle of spectral interference. The typical structure of such optical film thickness meter includes a wide-spectrum light source, an optical measurement probe and a spectrum detector, which can obtain interference signals through transmission or reflection mode, and inversely calculate the thickness of the thin film. Although such equipment has the advantages of non-contact and fast measurement speed, it still has obvious shortcomings: first, most of the existing systems are designed for single-point measurement, which can only obtain the thickness information of a local position on the substrate at a time, and cannot simultaneously realize multi-point or area scanning measurement of the large-area film forming area, which makes it difficult to comprehensively evaluate the thickness uniformity and defect distribution of the thin film; second, such instruments are easily affected by external environment and hardware stability, especially the fluctuation of light source output intensity and spectral drift will significantly reduce the repeatability and accuracy of the measurement results, so frequent manual calibration and correction are needed. This feature not only increases the maintenance cost, but also seriously affects the stability and reliability of the equipment in long-time continuous operation, which is not conducive to the integration into the full-automatic production line to realize uninterrupted quality monitoring.

[0005] In view of the above problems, the existing technology needs to be improved. SUMMARY

[0006] The application aims to provide a multi-channel spectrum-based thin film thickness online detection method and system, which can realize synchronous online detection of multiple positions of a thin film, significantly improve detection efficiency and spatial resolution, and improve measurement accuracy and stability.

[0007] In a first aspect, the application provides a multi-channel spectrum-based thin film thickness online detection method, which comprises:

[0008] A1. Irradiating multiple different positions on the surface of a thin film with multiple light beams having continuous spectra, and collecting multi-channel spectrum images formed by reflected light from the surface of the thin film; each channel of the multi-channel spectrum images corresponds to each irradiated position on the surface of the thin film;

[0009] A2. Dividing the multi-channel spectrum images by channels, and extracting single-channel spectrum images of each channel;

[0010] A3. Calculating the film thickness of each irradiated position by using a film thickness analysis algorithm according to each single-channel spectrum image;

[0011] A4. Presenting the film thickness calculation results on a display interface.

[0012] In a second aspect, the application provides a multi-channel spectrum-based thin film thickness online detection system, which comprises a broadband light source, an optical collimation and filtering system, a light splitting unit, an imaging objective, a multi-channel probe, a fiber array, an imaging spectrometer, a camera acquisition system, and an upper computer.

[0013] The broadband light source is used to generate an illumination light beam with a continuous spectrum; the optical collimation and filtering system is used to collimate and filter the illumination light beam; the light splitting unit is used to divide the collimated and filtered illumination light beam into multiple sub-beams, which irradiate multiple different positions on the surface of a thin film after passing through the imaging objective and the multi-channel probe, and are reflected to form multiple reflected light beams carrying thickness interference information; the reflected light beams are coupled to the fiber array after passing through the imaging objective and the light splitting unit again, and are transmitted to the imaging spectrometer for generation of multi-channel spectrum images; the camera acquisition system is used to collect the multi-channel spectrum images and send them to the upper computer.

[0014] The upper computer is used to divide the multi-channel spectrum images by channels, extract single-channel spectrum images of each channel, calculate the film thickness of each irradiated position by using a film thickness analysis algorithm according to each single-channel spectrum image, and present the film thickness calculation results on a display interface.

[0015] Beneficial effects: The application provides a kind of film thickness online detection method and system based on multi-channel spectrum, by utilizing multiple light beams with continuous spectrum to irradiate multiple different positions of film surface, and multiple-channel spectral image is formed by collecting reflected light, then the image is segmented to extract single-channel spectral image, and the film thickness of each irradiated position is calculated based on film thickness analysis algorithm, and finally the film thickness calculation result is presented on the display interface.The method realizes the synchronous detection of multiple positions on the film surface by collecting multi-channel spectral image, significantly improves the detection efficiency and spatial resolution, and can comprehensively evaluate the thickness distribution uniformity of the film.Compared with traditional contact or offline measurement method, the method has the advantages of non-contact, fast measurement speed, can reflect the dynamic change of film thickness, etc., overcomes the disadvantages that traditional method is limited in application in continuous production line.In addition, through the fine film thickness analysis algorithm, the method can more accurately invert the film thickness, improve the accuracy and stability of the measurement result, reduce the dependence on manual calibration and correction, thereby reducing the maintenance cost and improving the reliability of the equipment in long-time continuous operation. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A flow chart of a film thickness online detection method based on multi-channel spectrum provided by the application.

[0017] Figure 2 A schematic diagram of a film thickness online detection system based on multi-channel spectrum provided by the application.

[0018] Figure 3 A schematic diagram of a multi-channel spectral image.

[0019] Label explanation: 1, broadband light source; 2, optical collimation and filtering system; 3, light splitting unit; 4, imaging objective; 5, multi-channel probe; 6, fiber array; 7, imaging spectrometer; 8, camera acquisition system; 9, upper computer. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. The components of the embodiments of the application described and shown in the drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the application provided in the drawings is not intended to limit the scope of the claimed application, but only represents selected embodiments of the application. Based on the embodiments of the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0021] It should be noted that similar reference numerals and letters refer to like items in the accompanying drawings, and that, once an item is defined in one drawing, it is not necessary to further define and explain it in subsequent drawings. Meanwhile, in the description of the present application, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0022] Please refer to Figure 1 In some embodiments of the present application, a multi-channel spectrum-based thin film thickness online detection method includes:

[0023] A1. A plurality of light beams with continuous spectrum are used to irradiate a plurality of different positions on the surface of a thin film (such as the packaging film of an OLED device), and a multi-channel spectrum image formed by the reflected light from the surface of the thin film is collected; each channel of the multi-channel spectrum image corresponds to each irradiated position on the surface of the thin film;

[0024] A2. The multi-channel spectrum image is segmented by channel, and a single-channel spectrum image of each channel is extracted;

[0025] A3. According to each single-channel spectrum image, a film thickness analysis algorithm is used to calculate the film thickness of each irradiated position;

[0026] A4. The film thickness calculation result is presented on a display interface.

[0027] The present application aims to provide an efficient, accurate and stable thin film thickness online detection scheme. Through the collection and processing of multi-channel spectrum images, the synchronous film thickness measurement of multiple positions on the surface of the thin film is realized, effectively overcoming the limitations of single-point measurement, poor real-time performance and insufficient stability in the prior art, thereby improving the quality control level in the thin film production process.

[0028] The multi-channel spectrum-based thin film thickness online detection method proposed in the present application is characterized by using multi-channel spectrum images to accurately calculate the film thickness. The "multi-channel spectrum image" refers to an image composed of multiple independent spectral channels, each channel recording the spectral information of the reflected light from different positions on the surface of the thin film (for example Figure 3 The multi-channel spectrum image shown in the figure, each white band in the figure represents the spectrum of one channel). These channels can be understood as spatially independent measurement regions, each region corresponding to an irradiated position. In this way, the spectral data of multiple points on the surface of the thin film can be obtained simultaneously, laying a foundation for subsequent parallel processing. The "single-channel spectrum image" refers to the spectral data corresponding to a single irradiated position separated from the multi-channel spectrum image.

[0029] In a specific implementation environment, the method can be implemented based on the system shown in Figure 2 .

[0030] The detection method of the present application first involves step A1, i.e. illuminating multiple different positions on the film surface with multiple beams of light having continuous spectrum, and collecting the multi-channel spectral image formed by the reflected light from the film surface. Each channel of the multi-channel spectral image corresponds to each illuminated position on the film surface. In actual operation, multiple ways can be used to achieve the illumination of multiple beams. For example, a broadband light source can be used in combination with a beam splitter array to divide a beam of broadband light into multiple sub-beams, and then guide these sub-beams to illuminate different positions on the film surface through independent optical paths. Another way is to use multiple independent broadband light sources, each corresponding to an illumination position. The collection of reflected light to form a multi-channel spectral image can be completed by an imaging spectrometer in combination with a camera acquisition system. The imaging spectrometer can separate the reflected light from different spatial positions in the spectral dimension, and the camera acquisition system can capture to form a multi-channel spectral image. For example, a push-broom imaging spectrometer can be used to obtain spectral data row by row or point by point by scanning the film surface, and finally synthesize a multi-channel spectral image.

[0031] Next is step A2, i.e. dividing the multi-channel spectral image by channel to extract the single-channel spectral image of each channel. The purpose of this step is to separate the spectral data of different illumination positions for independent film thickness calculation. One implementation is to run image processing software in the host computer, which is pre-configured with the area range of each channel in the multi-channel spectral image. When receiving the multi-channel spectral image, the software performs cropping or masking operation according to these pre-set areas, thereby extracting the single-channel spectral image corresponding to each channel. For example, if the multi-channel spectral image is composed of M channels arranged side by side, the software can divide the image into M independent sub-images according to the pixel range of each channel, and each sub-image is a single-channel spectral image.

[0032] Subsequently, step A3 is performed, i.e. calculating the film thickness of each irradiated position according to each single-channel spectral image by using a film thickness analysis algorithm. The film thickness analysis algorithm is the core of the present method, and its goal is to inverse the film thickness from the spectral data. For example, the WaveNumberFFT algorithm, the LombScargle algorithm, the LombScargleWithFFTFilter algorithm, the ComplexMethod algorithm, the NonlinearLeastSquare algorithm or the Phase algorithm can be used for film thickness calculation (all of these algorithms are prior art, and will not be described in detail here); in fact, these algorithms can be integrated in the system, and when actual detection is performed, the film material characteristics, signal quality and precision requirements can be flexibly selected, specifically, the user can select independently, or the system can automatically select based on a pre-set recommendation table (the recommendation table records the applicable scenarios of each algorithm, and the system selects the corresponding algorithm according to the scenario matched by the film material characteristics, signal quality and precision requirements).

[0033] Finally, step A4 is performed, i.e. presenting the film thickness calculation results on the display interface. This step aims to intuitively show the user the detection results of the film thickness. One implementation is that the display module in the host computer can receive the film thickness calculation results, and display them on the display in the form of numbers, charts or images. For example, the specific film thickness value of each irradiated position can be displayed, or a film thickness distribution map can be generated, which intuitively shows the thickness difference of different positions on the film surface in the form of color coding or contour lines, helping the user to quickly identify the uniformity of the film. In addition, the original multi-channel spectral image, single-channel spectral image and other related detection parameters can also be displayed together, so that the user can conduct comprehensive analysis and monitoring.

[0034] Compared with the prior art, the film thickness online detection method based on multi-channel spectrum of the present application has significant advantages and innovations. Traditional optical film thickness meters mostly use single-point measurement mode, which needs to scan or move the sample to obtain thickness information at different positions. This not only has slow measurement speed, but also is difficult to realize real-time online monitoring, especially on a continuous production line. The present application fundamentally solves the limitations of single-point measurement by using multiple light beams with continuous spectrum to simultaneously irradiate multiple different positions on the film surface and collecting multi-channel spectral images. This multi-channel parallel collection method enables the system to obtain thickness information of multiple positions on the film at one time, greatly improving the detection efficiency and spatial resolution. In addition, by segmenting the multi-channel spectral image by channel and independently analyzing the film thickness of each single-channel spectral image, the present application can comprehensively evaluate the film thickness distribution, which is crucial for ensuring the uniformity of the packaging film of OLED and other devices. Finally, the film thickness calculation results are presented on the display interface, providing intuitive and real-time feedback to the operator, which helps to adjust the process parameters in a timely manner and improve the product yield. Therefore, the method of the present application is superior to the prior art in terms of real-time performance, efficiency and comprehensiveness, and provides a more advanced and reliable solution for online quality control of the film manufacturing process.

[0035] In actual online detection applications, if each processing step (such as image acquisition, channel segmentation and film thickness calculation) is executed in series, it may result in low data processing efficiency and difficulty in meeting real-time requirements, especially in scenarios with large amounts of data or requiring fast response, the system throughput will be limited. If the above problems are not solved, the detection results may lag behind, and the film thickness changes in the production process cannot be fed back in a timely manner, thereby affecting product quality control and production efficiency. To this end, the present application further proposes an optimization scheme by introducing a parallel processing mechanism to improve data processing efficiency and real-time response capability of the system.

[0036] Specifically, before step A1, it further includes the following steps:

[0037] A0. Start multiple parallel threads and multiple queues; the multiple parallel threads include an image acquisition thread, multiple channel segmentation threads and multiple film thickness calculation threads; the multiple queues include a channel segmentation queue and a thickness calculation queue;

[0038] In step A1, the multi-channel spectral image is acquired in real time based on the image acquisition thread, and the multi-channel spectral image is sequentially added to the channel segmentation queue;

[0039] In step A2, the multi-channel spectral image is segmented based on the multiple channel segmentation threads, and each single-channel spectral image extracted is sequentially added to the thickness calculation queue;

[0040] In step A3, the multiple film thickness calculation threads sequentially obtain single-channel spectral images from the thickness calculation queue for film thickness calculation.

[0041] The multiple parallel threads refer to program execution units created at the operating system level or the application software level, which can independently execute tasks. The image acquisition thread is responsible for interacting with the camera acquisition system and other hardware, and real-time acquisition of multi-channel spectral images. The multiple channel segmentation threads take images from the queue for segmentation in parallel to speed up the process. The multiple film thickness calculation threads also take single-channel spectral images from the queue in parallel to execute complex film thickness analysis algorithms. The multiple queues, such as the channel segmentation queue and the thickness calculation queue, are data structures for inter-thread communication and data buffering. The channel segmentation queue is used to store multi-channel spectral images to be segmented, serving as a data bridge between the image acquisition thread and the channel segmentation thread. The thickness calculation queue is used to store segmented single-channel spectral images, serving as a data bridge between the channel segmentation thread and the film thickness calculation thread. These queues usually use a first-in, first-out (FIFO) mechanism to ensure the orderliness and integrity of data processing.

[0042] The scheme of the present application effectively solves the efficiency bottleneck that may exist in the traditional serial processing mode in online detection by introducing the multi-thread and multi-queue mechanism. Specifically, the image acquisition thread runs independently, is responsible for uninterrupted real-time acquisition of multi-channel spectral images, and quickly puts them into the channel segmentation queue, thereby ensuring the continuous supply of data sources and avoiding the waiting of subsequent processing caused by slow acquisition speed. Then, multiple channel segmentation threads are started, which can take multi-channel spectral images to be processed from the channel segmentation queue in parallel and perform channel segmentation operations. Since the segmentation tasks are distributed to multiple threads for parallel execution, the total image segmentation time is greatly shortened. After segmentation, the extracted single-channel spectral images are sequentially added to the thickness calculation queue. Finally, multiple film thickness calculation threads also take single-channel spectral images from the thickness calculation queue in parallel to independently execute complex film thickness analysis algorithms. This parallel processing mode significantly improves the throughput of film thickness calculation. Through the queue as a data buffer, the various processing stages (acquisition, segmentation, calculation) are effectively decoupled, so that even if the processing speed of a certain stage is temporarily slower than that of other stages, it will not immediately block the entire process, but will be buffered through the queue, thereby ensuring smooth operation and high efficiency of the entire system.

[0043] By the technical solution, the overall processing efficiency and real-time performance of the film thickness online detection method can be significantly improved. Compared with the traditional serial processing mode, the parallel thread and queue mechanism are introduced, so that time-consuming operations such as image acquisition, channel segmentation and film thickness calculation can be executed concurrently, effectively avoiding the single processing link from becoming a system bottleneck. Therefore, the system can process continuous input multi-channel spectral images with higher throughput, ensuring the rapid response and data update frequency of online detection, thereby better meeting the needs of real-time monitoring and rapid feedback on the industrial production line. In addition, the modular parallel design also improves the scalability and robustness of the system, enabling the system to adapt to different processing loads and hardware configurations.

[0044] In step A3, in addition to using existing methods such as WaveNumberFFT algorithm, LombScargle algorithm, LombScargleWithFFTFilter algorithm, ComplexMethod algorithm, NonlinearLeastSquare algorithm or Phase algorithm for film thickness calculation, the following method can also be used for calculation. Specifically, in step A3, for each single-channel spectral image, the film thickness of the irradiated position corresponding to the single-channel spectral image is calculated using the following film thickness analysis algorithm:

[0045] A301. Within the spatial frequency range of the single-channel spectral image in the wave number domain, a plurality of spatial frequency points are determined with a preset frequency step, and the weights of each spatial frequency point are initialized;

[0046] A302. According to the weights and the light intensity of each spatial frequency point, a reference constant is calculated;

[0047] A303. In turn, each spatial frequency point is taken as a target spatial frequency point, and the following steps are executed:

[0048] B1. Initialize the cumulative variable to zero;

[0049] B2. Divide all spatial frequency points into a plurality of spatial frequency point groups;

[0050] B3. In turn, for each spatial frequency point group, according to the reference constant corresponding to each spatial frequency point and the target spatial frequency point of the spatial frequency point group, the cumulative variable is iteratively updated;

[0051] B4. After completing the iterative update of the cumulative variable for each spatial frequency point group, the power spectrum value corresponding to the spatial frequency point group is calculated according to the updated cumulative variable and the reference constant;

[0052] B5. According to the power spectrum value corresponding to each spatial frequency point group, the optimal spatial frequency point update operation is performed;

[0053] A304. According to the optimal spatial frequency point and the refractive index of the film, the film thickness of the irradiated position corresponding to the single-channel spectral image is calculated.

[0054] Specifically, in step A301, first, the spatial frequency range of the single-channel spectral image in the wave number domain needs to be determined. This range is usually set according to the material properties and the expected thickness range of the film to be measured. Within this range, a series of discrete spatial frequency points are determined at a preset frequency step (for spatial frequency step), for example, 0.001 nm -1 or 0.0001 nm -1 . These spatial frequency points are the basis for subsequent film thickness calculation. At the same time, the weights corresponding to each spatial frequency point need to be initialized. These weights can be obtained from empirical values or through calibration data, and their purpose is to distinguish the importance of different spatial frequency points in the calculation process.

[0055] In step A302, a series of reference constants are calculated according to the weights of the determined spatial frequency points and the light intensity in the single-channel spectral image corresponding to these spatial frequency points. These reference constants are the key input for subsequent iterative calculation and power spectrum value calculation, and they contain spectral data and weight information, providing basic data for film thickness analysis.

[0056] In step A303, the core of the film thickness analysis algorithm is to iteratively process each spatial frequency point. Specifically, this step takes each spatial frequency point as the target spatial frequency point in turn and performs a series of sub-steps. First, in sub-step B1, initialize the cumulative variable to zero to prepare for subsequent iterative accumulation operations. Next, in sub-step B2, divide all spatial frequency points into multiple spatial frequency point groups, for example, 1000 spatial frequency points for a group. This grouping process helps improve computational efficiency, especially when dealing with a large number of spatial frequency points. In sub-step B3, the algorithm iteratively updates the cumulative variable for each spatial frequency point group according to the spatial frequency points in the group and the part of the reference constant corresponding to the target spatial frequency point calculated earlier. This iterative process is the core of the film thickness analysis algorithm, which gradually approaches the true film thickness information through continuous accumulation and updating. After completing the iterative update of the cumulative variable for each spatial frequency point group, in sub-step B4, the power spectrum value corresponding to the spatial frequency point group is calculated according to the updated cumulative variable and the reference constant. The power spectrum value reflects the energy distribution of the spectral signal at a specific spatial frequency and is an important basis for judging the film thickness. Finally, in sub-step B5, the optimal spatial frequency point is updated according to the power spectrum values corresponding to each spatial frequency point group. This operation aims to find the optimal spatial frequency point that best matches the actual film thickness from all spatial frequency points.

[0057] In step A304, once the optimal spatial frequency point is determined, the film thickness of the irradiated position corresponding to the single-channel spectral image can be calculated in combination with the refractive index of the film. The refractive index of the film is an inherent optical parameter of the material and is crucial for accurately calculating the film thickness.

[0058] The scheme of the present application refines the film thickness analysis algorithm into multiple specific steps, and introduces the concepts of spatial frequency points, weights, reference constants, accumulated variables, and power spectrum values, to achieve accurate calculation of the film thickness. First, by determining the spatial frequency points in the wave number domain and initializing the weights, a discrete data basis and weighting processing capability are provided for subsequent calculations. Second, by calculating the reference constants, the spectral light intensity and weight information are integrated to provide a unified input for the iteration process. Further, by dividing the spatial frequency points into groups and iteratively updating the accumulated variables, the calculation efficiency is effectively improved, and the feature information related to the film thickness is gradually extracted. By calculating the power spectrum values and updating the optimal spatial frequency point, the spatial frequency directly related to the film thickness can be accurately identified from the complex interference spectrum. Finally, in combination with the refractive index of the film, the thickness of the film can be directly and accurately calculated.

[0059] Through the above technical scheme, the present application can provide a more detailed and accurate film thickness analysis algorithm, thereby improving the accuracy and reliability of the online detection of the film thickness. The algorithm can more effectively process spectral data, reduce noise interference, and more accurately identify feature information related to the film thickness by introducing the weights of the spatial frequency points, the reference constants, and the method of step-by-step iterative calculation of the power spectrum values. This refined analysis process enables the detection accuracy to remain high when facing films of different material types and thickness ranges, thereby improving the robustness and applicability of the entire detection method.

[0060] In some preferred embodiments, in step A301, the weights of the spatial frequency points are matched in the preset weight database according to the actual scene parameters; the weight database records the combinations of weights applicable under different scene parameter conditions determined through debugging in advance; the scene parameters include the material type and the target thickness range of the film.

[0061] Specifically, the "actual scene parameters" refer to the specific condition information related to the film to be detected and its detection environment during the online detection of film thickness. These parameters are key factors that affect the performance of the film thickness analysis algorithm, such as the "material type" and "target thickness range" of the film. The "material type" of the film can include but is not limited to silicon oxide, silicon nitride, metal film, etc. Different materials have different optical properties, which will affect the characteristics of the spectral signal. The "target thickness range" refers to the expected thickness interval of the film to be detected, such as 100nm-500nm, 500nm-1000nm, etc. Different thickness ranges require different weight configurations to optimize the sensitivity and accuracy of the algorithm.

[0062] The "preset weight database" is a data set that stores a variety of weight combinations, which are obtained by experiments, simulations or expert experience debugging and optimization in advance for different scene parameters. This database can be implemented as a file, table or special database system stored in the host computer. In actual application, when the detection system obtains the current actual scene parameters, it will find the "weight combination" that best matches the current scene parameters from the "preset weight database" through query or matching algorithm. "Matching" means that the system will retrieve and select the weight set most suitable for the current detection conditions in the database according to the current film material type and target thickness range, etc., for subsequent spatial frequency point weight initialization. This approach ensures that the film thickness analysis algorithm can dynamically select the optimal weight configuration according to the specific detection environment and film characteristics, thereby improving the accuracy and robustness of the calculation.

[0063] The scheme of the present application solves the limitations of traditional weight initialization methods when faced with diversified film detection scenarios by introducing a matching mechanism between actual scene parameters and a preset weight database. When the detection system obtains the material type and target thickness range of the film, etc. actual scene parameters, the system can intelligently retrieve and load the weight combination most suitable for the current scene from the pre-established weight database. These weight combinations are determined after a large number of debugging and optimization, and can more accurately reflect the characteristics of the spectral signal and the needs of the film thickness analysis algorithm under different scenarios. Thus, in step A301, the weights of each spatial frequency point are no longer fixed, but are dynamically adjusted and optimized according to the actual situation, thereby providing a more accurate and reliable basis for subsequent reference constant calculation, cumulative variable iterative update and power spectrum value calculation.

[0064] By the technical solution, the precision and adaptability of the online detection of the film thickness can be improved. Since the weights are matched and optimized according to the actual scene parameters (such as the film material type and the target thickness range), the film thickness analysis algorithm can obtain better performance when processing different types of films or detection tasks of different thickness ranges. This avoids the calculation errors or convergence problems caused by improper weight setting, so that the system can maintain high detection accuracy and stability when facing complex and variable industrial production environments, thereby improving the robustness and practical value of the entire detection method.

[0065] Specifically, the reference constants include weighted spectral intensity, spectral deviation sum, spectral deviation square sum, frequency step cosine factor, frequency step sine factor, cosine weight constant, and sine weight constant.

[0066] In step A302, the reference constants are calculated according to the following formula:

[0067] ;

[0068] ;

[0069] ;

[0070] ;

[0071] ;

[0072] ;

[0073] ;

[0074] wherein N is the total number of spatial frequency points, is the weighted spectral intensity corresponding to the i-th spatial frequency point, is the weight of the i-th spatial frequency point, is the light intensity in the single-channel spectral image corresponding to the i-th spatial frequency point, is the spectral deviation sum, is the spectral deviation square sum, is the frequency step cosine factor corresponding to the i-th spatial frequency point, is the frequency step sine factor corresponding to the i-th spatial frequency point, is the preset frequency step, is the wavelength corresponding to the i-th spatial frequency point, is the cosine weight constant corresponding to the i-th spatial frequency point, is the sine weight constant corresponding to the i-th spatial frequency point.

[0075] The scheme of the present application defines and calculates a series of reference constants, which lay a solid foundation for spectral analysis and power spectrum calculation in film thickness analysis algorithm. These constants can more accurately represent the characteristics of single-channel spectral images, such as weighted spectral intensity and spectral deviation for evaluating the quality and fitting degree of spectral data, which helps to extract effective information in complex spectral background. At the same time, the frequency step cosine factor and the frequency step sine factor directly participate in the calculation related to Fourier transform, ensuring the accuracy of spectral analysis. The cosine weight constant and the sine weight constant play a key weighting role in power spectrum calculation, which helps to highlight the spectral features related to film thickness, thereby improving the accuracy of film thickness calculation.

[0076] Through the above technical scheme, the present application can provide a set of standardized and accurate reference constant calculation method for film thickness analysis algorithm. This helps to improve the accuracy and stability of film thickness calculation, especially when dealing with complex spectral data, it can more effectively extract the feature information related to film thickness, thereby improving the overall performance and reliability of online detection of film thickness.

[0077] Further, the accumulation variables include cosine accumulation, sine accumulation, cosine square accumulation, sine square accumulation, and sine-cosine cross accumulation.

[0078] In step B3, the following steps are performed for each spatial frequency point group in turn:

[0079] B301. Initialize the sine function and cosine function according to the following formula:

[0080] ;

[0081] ;

[0082] wherein, is the sine function, is the cosine function, is the wavelength corresponding to the first spatial frequency point in the current spatial frequency point group;

[0083] B302. Update the cosine accumulation, sine accumulation, cosine square accumulation, and sine-cosine cross accumulation according to the following formula:

[0084] ;

[0085] ;

[0086] ;

[0087] ;

[0088] wherein, is a cosine accumulation, denotes the cosine accumulation before update, is a sine accumulation, denotes the sine accumulation before update, is a cosine square accumulation, denotes the cosine square accumulation before update, is a sine-cosine cross accumulation, denotes the sine-cosine cross accumulation before update;

[0089] B303. If all spatial frequency points of the current spatial frequency point group have not been traversed, update the sine function and the cosine function according to the following formulas, and return to step B302:

[0090] ;

[0091] ;

[0092] wherein, denotes the sine function before update, denotes the cosine function before update, j is the ranking sequence number of the current target spatial frequency point among all spatial frequency points, is a frequency step sine factor corresponding to the current target spatial frequency point, is a frequency step cosine factor corresponding to the current target spatial frequency point;

[0093] B304. If all spatial frequency points of the current spatial frequency point group have been traversed, update the cosine square accumulation, the sine square accumulation and the sine-cosine cross accumulation according to the following formulas, and end the iteration of the current spatial frequency point group:

[0094] ;

[0095] ;

[0096] ;

[0097] wherein, is the sine square accumulation.

[0098] Specifically, the aforementioned accumulated variables include cosine accumulated value c, sine accumulated value s, cosine square accumulated value cc, sine square accumulated value ss, and sine-cosine cross accumulated value cs. These accumulated variables are designed to systematically accumulate spectral information from multi-channel spectral images for subsequent calculation of power spectral values, which are key indicators for film thickness resolution.

[0099] In step B301, the sine function Sum and cosine functions Initialized. Here. This represents the wavelength corresponding to the first spatial frequency point in the current spatial frequency point group. By performing sine and cosine operations on the wavelength of the first spatial frequency point, an initial phase reference is set for the accumulation process of this spatial frequency point group.

[0100] In practical applications, in step B302, the cosine accumulator c, the sine accumulator s, the cosine square accumulator cc, and the sine-cosine cross accumulator cs are updated. When traversing each spatial frequency point in the spatial frequency point group, the current sine function... Sum and cosine functions The value of c is accumulated into the corresponding variable. Specifically, c is accumulated. The value of s is accumulated. The value of cc is accumulated. The squared value, cs accumulation and The product of these operations ensures that the contribution of each spatial frequency point to the overall spectral information is effectively recorded.

[0101] Furthermore, in step B303, if the current spatial frequency point group has not yet been traversed, then the sine function... Sum and cosine functions It updates according to a specific formula. This update utilizes the frequency step size sine factor. and frequency step size cosine factor and the previous spatial frequency point and The value. This iterative update method avoids repeatedly performing time-consuming trigonometric function calculations for each spatial frequency point, thus significantly improving computational efficiency. Here, j represents the sorting index of the current target spatial frequency point among all spatial frequency points.

[0102] Furthermore, in step B304, after traversing all spatial frequency points in the current spatial frequency point group, the cosine square accumulator cc, the sine square accumulator ss, and the sine-cosine cross accumulator cs will undergo final adjustments. These adjustments are typically used to normalize or center the accumulated results to eliminate potential biases and provide more stable and accurate input for subsequent power spectrum calculations.

[0103] The scheme of the present application defines the composition of the accumulation variables (cosine accumulation c, sine accumulation s, cosine square accumulation cc, sine square accumulation ss, and cross-sine-cosine accumulation cs) precisely, and provides a set of systematic initialization and iterative updating formulas, ensuring that the spectral information extracted from the multi-channel spectral image can be accurately and efficiently accumulated. Specifically, the initialization of the sine function and the cosine function in step B301 sets a clear starting point for the accumulation process of each spatial frequency point group. Subsequently, the incremental updating mechanism in step B302 enables the continuous and accurate aggregation of the spectral contribution of each spatial frequency point. Importantly, the iterative updating of the sine function and the cosine function in step B303 cleverly utilizes the frequency step factor, avoiding repeated trigonometric function calculations for each spatial frequency point, thereby significantly improving computational efficiency. Finally, the adjustment of the accumulation variables in step B304 further optimizes the numerical stability of the accumulation results, making them more suitable for subsequent power spectrum value calculations, thereby laying a solid foundation for the accuracy and robustness of the film thickness analysis algorithm.

[0104] Through the above technical scheme, the calculation accuracy of the accumulation variables is significantly improved, as each component (cosine accumulation, sine accumulation, cosine square accumulation, sine square accumulation, and cross-sine-cosine accumulation) is defined and updated through explicit mathematical formulas. This high precision directly contributes to a more accurate representation of spectral characteristics. In addition, within each spatial frequency point group, the iterative updating of the sine function and the cosine function, as described in step B303, significantly improves computational efficiency by minimizing repeated trigonometric function calculations. The final adjustment of the square and cross accumulation variables further ensures the numerical stability and robustness of the entire film thickness analysis algorithm. Therefore, the method of the present application can achieve more reliable and accurate film thickness measurement, effectively reducing potential errors that may occur in traditional methods due to unclear or inefficient accumulation process definitions.

[0105] Further, in step B4, the following steps are performed for each spatial frequency point group in turn:

[0106] B401. After completing the iterative updating of the accumulation variables for the current spatial frequency point group, calculate the discriminant value according to the following formula:

[0107] ;

[0108] wherein, is the discriminant value;

[0109] B402. The power spectrum value corresponding to the current spatial frequency point group is calculated according to the following formula:

[0110] ;

[0111] wherein, is the power spectrum value corresponding to the current spatial frequency point group, is the sine weight constant corresponding to the current target spatial frequency point, is the cosine weight constant corresponding to the current target spatial frequency point.

[0112] wherein, the discriminant value d is calculated based on the sine square accumulation ss, the cosine square accumulation cc and the sine-cosine cross accumulation cs obtained by iterative updating in the above step B3, and reflects the mutual relationship between these accumulation variables. The sine weight constant and the cosine weight constant are determined according to the reference constant calculated in the above step A302, in combination with the information of the current target spatial frequency point. The calculation of the power spectrum value p comprehensively utilizes these accumulation variables, weight constants and spectral deviation square sum , aiming to quantify the spectral feature intensity related to the film thickness of the current spatial frequency point group.

[0113] The scheme of the present application can more accurately evaluate the linear correlation between the accumulation variables by introducing the calculation of the discriminant value d, thereby providing a correction factor for the calculation of the power spectrum value. The calculation formula of the power spectrum value p is based on the consideration of multiple reference constants such as the weighted spectral intensity, the spectral deviation and the frequency step cosine factor and the sine factor, and combines the comprehensive information of the accumulation variables. This calculation method can effectively extract the periodic information related to the film thickness from complex multi-channel spectral images and quantify it in the form of power spectrum, making the subsequent optimal spatial frequency point selection more reliable and thus improving the accuracy of film thickness calculation.

[0114] Through the above technical scheme, the present application can provide a more accurate and robust power spectrum calculation method. By introducing the discriminant value d, the calculation error caused by the high correlation between the accumulation variables under certain specific conditions can be effectively avoided, thereby improving the stability of the power spectrum calculation. In addition, the power spectrum calculation formula which comprehensively utilizes multiple reference constants and accumulation variables makes the obtained power spectrum value more comprehensively reflect the interference spectral characteristics of the film, providing a solid foundation for the accuracy of the subsequent film thickness analysis algorithm, especially when dealing with films with complex spectral characteristics, the advantage is more significant.

[0115] Preferably, step B5 can include:

[0116] B501. Extract the maximum value in the power spectrum values corresponding to each spatial frequency point group as the current maximum power spectrum value;

[0117] B502. If the current target spatial frequency point is the first spatial frequency point, then take the current target spatial frequency point as the optimal spatial frequency point, and take the current maximum power spectrum value as the global maximum power spectrum value;

[0118] B503. If the current target spatial frequency point is not the first spatial frequency point, then compare the current maximum power spectrum value with the global maximum power spectrum value. If the current maximum power spectrum value is greater than the global maximum power spectrum value, then update the optimal spatial frequency point to the current target spatial frequency point, and update the global maximum power spectrum value to the current maximum power spectrum value. Otherwise, keep the optimal spatial frequency point and the global maximum power spectrum value unchanged.

[0119] Specifically, in step B501, after completing the iterative update of the accumulation variable for each spatial frequency point group and calculating the power spectrum value corresponding to the spatial frequency point group, the maximum value needs to be identified and extracted from it. This maximum value is temporarily marked as the "current maximum power spectrum value", which represents the power spectrum intensity generated by the film thickness value most likely to correspond to the spatial frequency point group being processed.

[0120] Step B502 is a special processing for the initialization stage of the algorithm. When the algorithm first processes the first target spatial frequency point, since there is no historical comparison data, the current target spatial frequency point is directly set as the initial "optimal spatial frequency point", and the "current maximum power spectrum value" corresponding to it is set as the "global maximum power spectrum value". This provides a benchmark for subsequent comparison and update operations.

[0121] In practical applications, step B503 is the core comparison and update logic. When processing subsequent target spatial frequency points, the system compares the "current maximum power spectrum value" calculated with the "global maximum power spectrum value" stored previously. If the current power spectrum value is greater, it means that a better potential film thickness solution has been found. At this time, the "optimal spatial frequency point" is updated to the current target spatial frequency point, and the "global maximum power spectrum value" is also updated to the current larger value. Conversely, if the current maximum power spectrum value is not greater than the global maximum power spectrum value, it means that the current target spatial frequency point is not a better solution. At this time, the "optimal spatial frequency point" and the "global maximum power spectrum value" remain unchanged to preserve the best result found previously.

[0122] By the technical solution, the robustness and accuracy of the film thickness analysis algorithm in the film thickness online detection method can be significantly improved. Compared with only performing a simple optimal spatial frequency point updating operation, the application introduces a global maximum power spectrum value comparison and updating mechanism, effectively avoiding the risk of the algorithm falling into a local optimal solution, ensuring that the optimal spatial frequency point finally determined is the best match in the global range. Therefore, the film thickness data calculated will be more accurate and reliable, reducing the measurement error caused by algorithm defects, and is especially suitable for thin film production and quality control scenarios with high precision requirements.

[0123] Specifically, in step A304, the film thickness of the irradiated position corresponding to the single-channel spectral image is calculated according to the following formula:

[0124] ;

[0125] Where t is the film thickness of the irradiated position corresponding to the single-channel spectral image, is the optimal spatial frequency point, and n is the refractive index of the thin film.

[0126] Specifically, the film thickness t refers to the actual physical thickness of the thin film at the irradiated position. The optimal spatial frequency point is determined after the optimal spatial frequency point updating operation on the power spectrum values corresponding to each spatial frequency point group in step A303 of the film thickness analysis algorithm, and it represents the main frequency component of the thin film interference spectrum in the wave number domain. The refractive index n of the thin film is an inherent optical parameter of the thin film material, which can usually be obtained by consulting a material manual or by experimental measurement. The refractive index n reflects the change of the speed of light in the thin film material, and is a key parameter for calculating the optical path difference and the actual thickness.

[0127] By the technical solution, the calculation process of the film thickness is standardized and refined, ensuring that the conversion from the spectral data to the final film thickness result has high accuracy and reliability. The explicit application of the formula enables step A304 of the film thickness analysis algorithm to provide a direct and physically meaningful film thickness calculation method, thereby improving the precision and stability of the entire film thickness online detection method, and providing a solid data foundation for subsequent film thickness distribution presentation and quality control.

[0128] In practical applications, merely presenting the calculation results may not intuitively reflect the overall thickness distribution of the film surface, and it is also difficult to quickly locate the abnormal thickness area, thereby affecting the efficiency and accuracy of the user's judgment of the film quality. If the above problems are not solved, the operator may need to spend more time analyzing and comparing data, reducing the real-time and practicality of online detection. To this end, the present application further proposes a scheme for optimizing the presentation of film thickness calculation results, aiming to improve the readability and analysis efficiency of the detection results through more intuitive and comprehensive information display. Specifically, step A4 includes:

[0129] A401. generating a thickness distribution map according to the calculation results of the film thickness of each irradiated position;

[0130] A402. displaying detection information on the display interface; the detection information includes the calculated film thickness data, the thickness distribution map, and the multi-channel spectral image.

[0131] Wherein, generating a thickness distribution map means visualizing the film thickness calculation results of each irradiated position through graphical means. Specifically, it can be to map the multiple different positions of the film surface irradiated in two-dimensional space, and according to the film thickness value calculated at each position, encode it with different colors, gray scales or textures, thereby forming an image that intuitively reflects the spatial distribution of film thickness. The purpose is to enable the user to observe the uniformity of the film thickness, the local thickness change trend and the potential defect area at a glance.

[0132] In practical applications, the detection information displayed on the display interface can be understood as a comprehensive detection report provided to the user. Specifically, the detection information includes the calculated film thickness data, i.e. the accurate film thickness value of each irradiated position; the thickness distribution map, which provides a macro overview of the thickness distribution; and the multi-channel spectral image, which provides the original image information to facilitate the user to trace and compare data. The purpose is to provide a comprehensive and multi-dimensional information view, so that the user can evaluate the quality of the film from multiple angles such as numerical value, spatial distribution and original data.

[0133] The scheme of the present application converts discrete film thickness calculation results into continuous and intuitive visual information by generating a thickness distribution map. This graphical presentation makes the spatial variation pattern of the film thickness, such as thickness uniformity, gradient change or local abnormal points, quickly identifiable through color or gray scale differences. At the same time, by integrating the film thickness data, thickness distribution map and original multi-channel spectral image on the display interface, the present application provides a comprehensive information platform. The operator can not only obtain accurate numerical values, but also quickly grasp the overall trend through the thickness distribution map, and cross-verify through the original spectral image, thereby ensuring a comprehensive understanding and accurate judgment of the test results. This coordinated display of multi-dimensional information significantly improves the efficiency and depth of the user's evaluation of the film quality.

[0134] Through the above technical scheme, the present application can provide a more intuitive, efficient and comprehensive film thickness detection result presentation method. The generation of the thickness distribution map visualizes the spatial distribution characteristics of the film thickness, and the user can quickly identify the thickness uniformity, local thickness abnormality or defect area without comparing the numerical values one by one, greatly improving the interpretation efficiency of the detection results. In addition, the integration of the calculated film thickness data, thickness distribution map and original multi-channel spectral image for display provides the user with multi-level and multi-angle detection information, which helps the operator to make more in-depth analysis and decision, thereby effectively improving the quality control level and response speed in the film production process.

[0135] Reference Figure 2 The present application provides a multi-channel spectral-based online film thickness detection system, which comprises a broadband light source 1, an optical collimation and filtering system 2, a light splitting unit 3, an imaging objective 4, a multi-channel probe 5, a fiber array 6, an imaging spectrometer 7, a camera acquisition system 8 and an upper computer 9.

[0136] The broadband light source 1 is used to generate an illumination light beam with a continuous spectrum; the optical collimation and filtering system 2 is used to collimate and filter the illumination light beam; the light splitting unit 3 is used to split the collimated and filtered illumination light beam into multiple sub-beams, which, after passing through the imaging objective 4 and the multi-channel probe 5, irradiate multiple different positions on the film surface and are reflected to form multiple beams of reflected light carrying thickness interference information; these reflected light beams pass through the imaging objective 4 and the light splitting unit 3 again, are coupled to the fiber array 6, and are transmitted to the imaging spectrometer 7 for generation of a multi-channel spectral image; the camera acquisition system 8 is used to acquire the multi-channel spectral image and send it to the upper computer 9.

[0137] The host computer 9 is configured to segment the multi-channel spectral image by channel, and extract a single-channel spectral image of each channel (for details, refer to the step A2 in the foregoing); calculate the film thickness of each irradiated position according to the single-channel spectral image (for details, refer to the step A3 in the foregoing); and present the film thickness calculation result on the display interface (for details, refer to the step A4 in the foregoing).

[0138] Specifically, the broadband light source 1 in the system is configured to generate an illumination beam with a continuous spectrum. As a preferred embodiment, the broadband light source 1 can adopt a halogen lamp, which has the advantages of low cost and wide spectral range. In other embodiments, an LED, a xenon lamp or a supercontinuum laser can also be used to provide a higher brightness or a wider spectral range of the illumination beam.

[0139] The optical collimation and filtering system 2 can be composed of a series of lenses, collimators and filters. For example, the collimator is used to convert the divergent beam emitted by the broadband light source 1 into a parallel beam, and the filter is used to select a specific wavelength range of light to adapt to the detection needs of different thin film materials. In some implementations, a tunable filter can also be used to provide greater flexibility.

[0140] The light splitting unit 3 can adopt a beam splitter or a microlens array to accurately split one collimated illumination beam into multiple sub-beams.

[0141] The imaging objective 4 usually adopts a high numerical aperture lens group, which is used to focus the split sub-beams onto the thin film surface and simultaneously collect the reflected light carrying thickness interference information from the thin film surface. For example, an achromatic objective can be used to reduce chromatic aberration and ensure imaging quality. In some applications, different magnification objectives can also be selected according to the thickness range of the thin film and the measurement accuracy requirements.

[0142] The multi-channel probe 5 can be composed of multiple independent optical probes, each corresponding to an irradiated position. These probes can be accurately arranged to ensure that the sub-beams can irradiate the preset multiple different positions on the thin film surface. The reflected light is collected through these probes. For example, the probes can be designed to have a specific working distance and spot size to adapt to different detection scenarios.

[0143] The fiber array 6 is used to couple the multiple reflected beams reflected from the thin film surface and passing through the imaging objective 4 and the light splitting unit 3 to the imaging spectrometer 7. For example, one-dimensional or two-dimensional fiber bundles can be used, with one end aligned with the output end of the light splitting unit 3 and the other end connected to the entrance slit of the imaging spectrometer 7. The arrangement of the fiber array 6 needs to match the input interface of the imaging spectrometer 7 to ensure effective transmission of spectral information.

[0144] The imaging spectrometer 7 is the core component for dispersing the incident light beam in the spectral dimension and detecting the light intensity distribution, thereby generating a multi-channel spectral image. For example, a grating-based imaging spectrometer can be employed, which separates different wavelengths of light through a grating and projects them to different spatial locations. In other implementations, prism-based dispersion elements can also be employed. The performance parameters of the imaging spectrometer 7, such as spectral resolution and spectral range, will directly affect the accuracy and applicability of the film thickness detection.

[0145] The camera acquisition system 8 is used to capture the multi-channel spectral image generated by the imaging spectrometer 7. For example, a high-frame-rate, high-resolution CCD camera or CMOS camera can be employed to ensure the real-time and accuracy of image acquisition. The acquired image data is transmitted to the host computer 9 through a high-speed interface for subsequent processing. The selection of the camera acquisition system 8 should consider its quantum efficiency, noise level, and dynamic range to optimize the acquisition quality of the spectral signal.

[0146] The host computer 9 serves as the control and data processing center of the entire system, and its functions include receiving the multi-channel spectral image sent by the camera acquisition system 8, executing image processing algorithms, running film thickness analysis algorithms, and presenting the results on the display interface. The host computer 9 is usually equipped with a high-performance processor and sufficient memory to support complex parallel computing tasks. The specific content of splitting the multi-channel spectral image by channel, extracting the single-channel spectral image of each channel, calculating the film thickness of each irradiated location using the film thickness analysis algorithm according to each single-channel spectral image, and presenting the film thickness calculation results on the display interface has been described in the above embodiments and will not be repeated here. It is emphasized that the host computer 9 realizes these functions through software programs, for example, a special graphical user interface (GUI) software can be developed, which integrates image segmentation modules, film thickness calculation modules, and result display modules to provide a friendly operation experience and efficient data processing capability.

[0147] The multi-channel spectral-based thin film thickness online detection system proposed in this application provides an innovative solution to the problems of single-point measurement, poor real-time performance, and insufficient stability in the prior art. Traditional optical film thickness meters are mostly designed for single-point measurement, making it difficult to simultaneously acquire multi-point or area scanning data in a large-area film formation area, resulting in the inability to comprehensively evaluate the thickness uniformity of the thin film. By configuring a broadband light source 1, a light splitting unit 3, and a multi-channel probe 5, the system can simultaneously emit light beams to multiple different positions on the thin film surface and collect reflected light, thereby achieving simultaneous detection of multiple points on the thin film, significantly improving the detection efficiency and the evaluation capability of the thin film uniformity.

[0148] In addition, the existing instrument is susceptible to the influence of external environment such as light source fluctuation and spectral drift, and the stability of its own hardware, resulting in the decrease of repeatability and accuracy of measurement results. The system acquires multi-channel spectral images through the precise imaging spectrometer 7 and camera acquisition system 8, and processes data by the host computer 9, which can compensate or correct these disturbances to a certain extent through algorithms, thereby improving the stability and reliability of measurement. For example, through cross-validation or statistical analysis of multi-channel data, the influence of single-channel measurement error on the overall result can be effectively reduced. In summary, through hardware integration and software optimization, the system realizes real-time, multi-point and high-precision online detection of film thickness, and provides a more efficient and reliable quality control method for the production of high-performance films.

[0149] The above only describes the embodiments of the present application and is not used to limit the protection scope of the present application. For those skilled in the art, the present application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A multi-channel spectrum-based thin film thickness online detection method, characterized in that, The method comprises: A1. Irradiating a plurality of different positions on the surface of a thin film with a plurality of light beams having continuous spectra, and collecting a multi-channel spectral image formed by reflected light of the surface of the thin film; each channel of the multi-channel spectral image corresponds to each irradiated position on the surface of the thin film; A2. Dividing the multi-channel spectral image by channel, and extracting a single-channel spectral image of each channel; A3. Calculating the film thickness of each irradiated position by using a film thickness analysis algorithm according to each single-channel spectral image; A4. Presenting the film thickness calculation result on a display interface; In step A3, for each single-channel spectral image, the film thickness of the irradiated position corresponding to the single-channel spectral image is calculated by using the following film thickness analysis algorithm: A301. Within the spatial frequency range of the single-channel spectral image in the wave number domain, a plurality of spatial frequency points are determined at a preset frequency step, and the weight of each spatial frequency point is initialized; A302. The reference constant is calculated according to the weight and the light intensity of each spatial frequency point; A303. Each spatial frequency point is sequentially taken as a target spatial frequency point, and the following steps are performed: B1. Initialize the cumulative variable to zero; B2. Divide all the spatial frequency points into a plurality of spatial frequency point groups; B3. For each spatial frequency point group, iteratively update the cumulative variable according to the spatial frequency points of the spatial frequency point group and the reference constant corresponding to the target spatial frequency point; B4. After the iterative update of the cumulative variable for each spatial frequency point group is completed, the power spectrum value corresponding to the spatial frequency point group is calculated according to the updated cumulative variable and the reference constant; B5. The optimal spatial frequency point is updated according to the power spectrum value corresponding to each spatial frequency point group; A304. The film thickness of the irradiated position corresponding to the single-channel spectral image is calculated according to the optimal spatial frequency point and the refractive index of the thin film.

2. The multi-channel spectroscopy based thin film thickness on-line detection method according to claim 1, wherein, Before step A1, the following step is further included: A0. Start a plurality of parallel threads and a plurality of queues; the plurality of parallel threads include an image acquisition thread, a plurality of channel division threads and a plurality of film thickness calculation threads; the plurality of queues include a channel division queue and a thickness calculation queue; In step A1, the multi-channel spectral image is acquired in real time based on the image acquisition thread, and the multi-channel spectral image is sequentially added to the channel division queue; In step A2, the multi-channel spectral image is obtained from the channel division queue for division processing based on the plurality of channel division threads, and each single-channel spectral image extracted is sequentially added to the thickness calculation queue; In step A3, the single-channel spectral image is obtained from the thickness calculation queue for film thickness calculation based on the plurality of film thickness calculation threads.

3. The multi-channel spectroscopy based thin film thickness on-line detection method according to claim 1, wherein, The reference constant includes weighted spectral intensity, spectral deviation sum, spectral deviation square sum, frequency step cosine factor, frequency step sine factor, cosine weight constant and sine weight constant; In step A302, the reference constant is calculated according to the following formula: ； ； ； ； ； ； ； wherein N is the total number of spatial frequency points, is a weighted spectral intensity corresponding to the i-th spatial frequency point, is a weight of the i-th spatial frequency point, is an optical intensity corresponding to the i-th spatial frequency point in the single-channel spectral image, is the spectral deviation sum, is the spectral deviation square sum, is a frequency step cosine factor corresponding to the i-th spatial frequency point, is a frequency step sine factor corresponding to the i-th spatial frequency point, is the preset frequency step, is a wavelength corresponding to the i-th spatial frequency point, is a cosine weight constant corresponding to the i-th spatial frequency point, is a sine weight constant corresponding to the i-th spatial frequency point.

4. The method for on-line measurement of film thickness based on multi-channel spectrum according to claim 3, characterized in that, The accumulated variables include cosine accumulated amount, sine accumulated amount, cosine square accumulated amount, sine square accumulated amount, and sine-cosine cross accumulated amount; In step B3, the following steps are performed sequentially for each of the spatial frequency point groups: B301. Initialize the sine and cosine functions according to the following formulas: ； ； wherein is a sine function, is a cosine function, is a wavelength corresponding to a first spatial frequency point in the current spatial frequency point group. B302. Update the cosine cumulative sum, the sine cumulative sum, the cosine square cumulative sum, and the sine-cosine cross cumulative sum according to the following formula: ； ； ； ； wherein is the cosine accumulation quantity, denotes the cosine accumulation quantity before update, is the sine accumulation quantity, denotes the sine accumulation quantity before update, is the cosine square accumulation quantity, denotes the cosine square accumulation quantity before update, is the sine-cosine cross accumulation quantity, denotes the sine-cosine cross accumulation quantity before update; B303. If not all spatial frequency points of the current spatial frequency point group have been traversed, update the sine and cosine functions according to the following formulas, and return to step B302: ； ； wherein, denotes a sine function before update, denotes a cosine function before update, j is the ranking serial number of the current target spatial frequency point in all spatial frequency points, is a frequency step sine factor corresponding to the current target spatial frequency point, is a frequency step cosine factor corresponding to the current target spatial frequency point; B304. If all spatial frequency points of the current spatial frequency point group have been traversed, then update the cosine square accumulation, the sine square accumulation, and the sine-cosine cross accumulation according to the following formulas, and end the iteration of the current spatial frequency point group: ； ； ； wherein is the sinusoidal square accumulation quantity.

5. The method for on-line measurement of film thickness based on multi-channel spectrum according to claim 4, characterized in that, In step B4, the following steps are performed sequentially for each of the spatial frequency point groups: B401. After iteratively updating the accumulated variables for the current spatial frequency point group, calculate the discriminant value according to the following formula: ； wherein is the discriminant value; B402. Calculate the power spectral density value corresponding to the current spatial frequency point group according to the following formula: ； wherein, is a power spectrum value corresponding to the current spatial frequency point group, is a sine weight constant corresponding to the current target spatial frequency point, is a cosine weight constant corresponding to the current target spatial frequency point.

6. The multi-channel spectroscopy based thin film thickness on-line monitoring method according to claim 1, wherein, Step B5 includes: B501. Extract the maximum value of the power spectrum value corresponding to each spatial frequency point group, and use it as the current maximum power spectrum value; B502. If the current target spatial frequency point is the first spatial frequency point, then the current target spatial frequency point is taken as the optimal spatial frequency point, and the current maximum power spectrum value is taken as the global maximum power spectrum value. B503. If the current target spatial frequency point is not the first spatial frequency point, compare the current maximum power spectrum value with the global maximum power spectrum value. If the current maximum power spectrum value is greater than the global maximum power spectrum value, update the optimal spatial frequency point to the current target spatial frequency point and update the global maximum power spectrum value to the current maximum power spectrum value. Otherwise, keep the optimal spatial frequency point and the global maximum power spectrum value unchanged.

7. The multi-channel spectroscopy based thin film thickness on-line monitoring method according to claim 1, wherein, In step A304, the film thickness at the irradiated location corresponding to the single-channel spectral image is calculated according to the following formula: ； wherein t is the film thickness of the irradiated position corresponding to the single-channel spectral image, is the optimal spatial frequency point, n is the refractive index of the thin film.

8. The multi-channel spectroscopy based thin film thickness on-line monitoring method according to claim 1, wherein, Step A4 includes: A401. Generate a thickness distribution map based on the calculated film thickness at each irradiated location; A402. The display interface shows the detection information, which includes the calculated film thickness data, the thickness distribution map, and the multi-channel spectral image.

9. A multi-channel spectroscopy based thin film thickness on-line measurement system, characterized in that, The system includes a broadband light source, an optical collimation and filtering system, a beam splitting unit, an imaging objective, a multi-channel probe, a fiber optic array, an imaging spectrometer, a camera acquisition system, and a host computer. The broadband light source is used to generate an illumination light beam with a continuous spectrum; the optical collimation and filtering system is used to collimate and filter the illumination light beam; the light splitting unit is used to split the collimated and filtered illumination light beam into multiple sub-beams, which, after passing through the imaging objective and the multi-channel probe, irradiate multiple different positions on the film surface and are reflected to form multiple beams of reflected light carrying thickness interference information; The reflected light is coupled to the fiber array after passing through the imaging objective and the light splitting unit again and is transmitted to the imaging spectrometer for generation of a multi-channel spectral image, and the camera acquisition system is used to acquire the multi-channel spectral image and send it to the upper computer; The upper computer is used to split the multi-channel spectral image by channel, extract a single-channel spectral image for each channel, calculate the film thickness at each irradiated position using a film thickness analysis algorithm based on each single-channel spectral image, and present the film thickness calculation results on a display interface; When the upper computer calculates the film thickness at each irradiated position using a film thickness analysis algorithm based on each single-channel spectral image, it calculates the film thickness at the irradiated position corresponding to each single-channel spectral image using the following film thickness analysis algorithm: A301. Determine multiple spatial frequency points in the spatial frequency range of the single-channel spectral image in the wave number domain with a preset frequency step, and initialize the weight of each spatial frequency point; A302. Calculate a reference constant based on the weight and the light intensity of each spatial frequency point; A303. Perform the following steps in turn with each spatial frequency point as a target spatial frequency point: B1. Initialize an accumulation variable to zero; B2. Divide all the spatial frequency points into multiple spatial frequency point groups; B3. In turn for each spatial frequency point group, iteratively update the accumulation variable based on the spatial frequency points of the spatial frequency point group and the reference constant corresponding to the target spatial frequency point; B4. After completing the iterative update of the accumulation variable for each spatial frequency point group, calculate the power spectrum value corresponding to the spatial frequency point group based on the updated accumulation variable and the reference constant; B5. Update the optimal spatial frequency point based on the power spectrum values corresponding to each spatial frequency point group; A304. Calculate the film thickness at the irradiated position corresponding to the single-channel spectral image based on the optimal spatial frequency point and the refractive index of the film.

Images