A quartz glass raw material impurity detection and separation system based on machine vision

Abstract

The present application relates to the technical field of mineral conveying, and particularly relates to a quartz glass raw material impurity detection and separation system based on machine vision, which acquires image data sets and synchronous environment parameters of quartz glass raw materials through a data acquisition module, extracts granularity distribution characteristics and light transmittance characteristics to determine raw material characteristic values through a raw material characteristic extraction module, divides the raw material into coarse-grained high-transmittance type or fine-grained low-transmittance type based on the raw material characteristic values, synchronous environment parameters and classification thresholds through a raw material characteristic classification module, applies polarization modulation light or wavelength selective modulation light based on the raw material type through an active optical modulation module, and collects modulation light images and analyzes and outputs impurity information and early warnings through a modulation analysis processing module. The present application realizes adaptive selection of optical modulation strategies according to the granularity and light transmittance characteristics of raw materials, overcomes the missed detection problem caused by optical interference, improves the impurity detection accuracy and system adaptability, and meets the technical requirements of continuous and stable production of high-purity quartz sand.

Description

Technical Field

[0001] This invention relates to the field of mineral conveying technology, and in particular to a machine vision-based system for detecting and separating impurities in quartz glass raw materials. Background Technology

[0002] Quartz glass, due to its excellent optical transmittance, extremely low coefficient of thermal expansion, and good chemical stability, is widely used in high-tech fields such as photolithography masks in semiconductor manufacturing, preforms in fiber optic communication, high-temperature resistant windows in aerospace, and optical components in high-energy laser systems. The purity of the quartz glass raw material directly determines the quality grade of the final product. If the raw material contains impurities such as iron, titanium, or chromium oxides, or fine gas-liquid inclusions formed during crystal growth, it can cause crystallization, bubble formation, or streak-like defects during subsequent melting, and in severe cases, even lead to the scrapping of an entire batch of high-value products. Therefore, efficient and accurate impurity detection and separation of quartz glass raw materials before the melting process has become a crucial step in ensuring the reliability of high-purity quartz products and the economic efficiency of production.

[0003] Chinese Patent Publication No. CN119513559A discloses a method for detecting impurities in quartz sand, comprising: acquiring spectral data of a quartz sand sample; determining the spectral interference degree of each characteristic peak based on its energy distribution; determining the matrix interference degree of each characteristic peak based on its own peak shape variation and its offset from the matrix characteristic peaks; further determining the interference intensity of each characteristic peak by combining its anti-interference degree with the interference intensity of the characteristic peaks; optimizing the chaos adjustment factor during the initialization of the chaotic mapping population; and finally, obtaining the characteristic band with the least interference using an optimization algorithm. Based on this, the purity and impurity content of the quartz sand are detected. This application can improve the accuracy of impurity detection in quartz sand.

[0004] However, the following problems still exist in the existing technology: First, existing technologies mostly use a single light source or spectral analysis to detect quartz raw materials. Because quartz has high transmittance and low reflectance, light penetrates the particles and undergoes multiple reflections. Surface scratches have low contrast with the matrix, and internal inclusions cannot penetrate to form an image. Surface and internal impurities are difficult to identify clearly at the same time, resulting in a high false negative rate and low detection accuracy.

[0005] Second, existing technologies employ a fixed detection strategy, using the same illumination and algorithm for all batches of raw materials. However, raw materials with different particle sizes and transmittances exhibit significant differences in optical response, making it impossible for the fixed strategy to adaptively adjust. This results in over-detection of some raw materials and under-detection of others, leading to low detection accuracy. Summary of the Invention

[0006] To address this, the present invention provides a machine vision-based system for detecting and separating impurities in quartz glass raw materials, which overcomes the problems of low detection accuracy and poor adaptability caused by the difficulty in simultaneously identifying surface and internal impurities due to optical interference in existing technologies, and the inability of a single detection strategy to adapt to differences in raw materials.

[0007] To achieve the above objectives, the present invention provides a machine vision-based system for detecting and separating impurities in quartz glass raw materials, comprising: The data acquisition module is used to acquire image datasets of quartz glass raw materials and synchronize environmental parameters; The raw material feature extraction module is used to preprocess the image dataset, extract the particle size distribution features and transmittance features of the quartz glass raw material, and determine the raw material feature values. The raw material characteristic classification module is used to classify the quartz glass raw material into coarse-grained high-transparency raw material or fine-grained low-transparency raw material based on the raw material characteristic value, the synchronous environmental parameters and the preset classification threshold. An active optical modulation module is used to determine the type of modulated light applied to the quartz glass material based on the material type; The modulation analysis and processing module is used to acquire the modulation light image of the quartz glass raw material under the modulation light irradiation, analyze the modulation light image to output impurity information and determine whether to issue a warning. The synchronized environmental parameters include ambient humidity data and ambient light intensity data.

[0008] Furthermore, the data acquisition module is used to acquire image datasets of quartz glass raw materials and synchronize environmental parameters, including... A continuous motion image sequence of the quartz glass raw material is acquired, and an image dataset is determined at a preset time interval; Collect the synchronization environment parameters corresponding to the image dataset.

[0009] Furthermore, the data acquisition module determines the image dataset to include, at preset time intervals. Obtain the timestamp of each frame in the continuous motion image sequence; The continuous motion image sequence is sampled according to the preset time interval, and the sampled image frames are used to construct the image dataset. The preset time interval is adjusted according to the movement speed of the quartz glass raw material.

[0010] Furthermore, the data acquisition module acquires the synchronization environment parameters corresponding to the image dataset, including: In response to the construction of the image dataset, environmental humidity data and ambient light intensity data corresponding to the image dataset are collected.

[0011] Furthermore, the raw material feature extraction module is used to preprocess the image dataset and extract the particle size distribution features and transmittance features of the quartz glass raw material, including... Perform grayscale statistics on the image data in the image dataset to obtain the mean grayscale value and grayscale variance of each image data. The overall grayscale distribution parameters of the image dataset are calculated based on the mean grayscale value and grayscale variance of each image to determine the granularity distribution characteristics. The light transmittance characteristics of the quartz glass raw material are determined based on the overall grayscale mean.

[0012] Furthermore, the raw material feature extraction module determines the raw material feature values ​​including, The ratio of the particle size distribution characteristics to the preset particle size priority value is determined as the particle size parameter; The ratio of the light transmittance characteristic to a preset light transmittance priority value is determined as the light transmittance parameter; The weighted sum of the particle size parameter and the transmittance parameter is determined as the characteristic value of the raw material.

[0013] Furthermore, the raw material characteristic classification module is used to classify the quartz glass raw material into coarse-grained high-transparency raw material or fine-grained low-transparency raw material based on the raw material characteristic values, the synchronous environmental parameters, and a preset classification threshold. The preset classification threshold is corrected based on the synchronous environment parameters to determine the corrected classification threshold; the raw material characteristic value is compared with the corrected classification threshold. If the raw material characteristic value is greater than or equal to the modified classification threshold, the quartz glass raw material is classified as a coarse-grained high-transparency raw material. If the raw material characteristic value is less than the modified classification threshold, the quartz glass raw material is classified as a fine-grained, low-transparency raw material.

[0014] Furthermore, the step of correcting the preset classification threshold based on the synchronization environment parameters to determine the corrected classification threshold includes: A first correction factor is determined based on the environmental humidity data; The second correction factor is determined based on the ambient light intensity data; The product of the preset classification threshold and the first correction coefficient and the second correction coefficient is used as the corrected classification threshold.

[0015] Furthermore, the modulation analysis and processing module acquires a modulation light image of the quartz glass raw material under the modulation light irradiation, and analyzes the modulation light image, including... If the quartz glass raw material is classified as a coarse-grained, high-transparency type, then a cross-polarization image of the quartz glass raw material under the polarization modulation light is acquired, and feature extraction is performed on the cross-polarization image to determine the impurity content information. If the quartz glass raw material is classified as a fine-grained, low-transmittance type, then the transmission spectrum image of the quartz glass raw material under wavelength selectively modulated light is collected, and grayscale analysis is performed on the transmission spectrum image to determine the impurity content information.

[0016] Furthermore, the modulation analysis and processing module outputs impurity information and determines whether to issue a warning, including: In response to the analysis result of the modulated light image, the content information of the impurities is output; If the impurity content exceeds the warning threshold, a warning signal will be issued.

[0017] Compared with existing technologies, this invention acquires image datasets and synchronous environmental parameters of quartz glass raw materials through a data acquisition module, extracts particle size distribution and transmittance features to determine raw material characteristic values ​​through a raw material feature extraction module, classifies raw materials into coarse-grained high-transmittance or fine-grained low-transmittance types based on raw material characteristic values, synchronous environmental parameters, and classification thresholds, and applies polarization-modulated light or wavelength-selective modulated light based on the raw material type through an active optical modulation module. The modulation analysis and processing module acquires the modulated light image and analyzes and outputs impurity information and early warnings. This invention achieves adaptive selection of optical modulation strategies based on the particle size and transmittance characteristics of the raw materials, overcomes the problem of missed detection caused by optical interference, improves the accuracy of impurity detection and system adaptability, and meets the technical requirements for continuous and stable production of high-purity quartz sand.

[0018] In particular, this invention focuses on the dynamic interference mechanism of synchronous environmental parameters on the classification threshold. In actual production environments, fluctuations in ambient humidity can alter the thickness of the adsorbed water film on the surface of quartz sand, thus affecting light scattering characteristics, while changes in ambient light intensity directly impact the stability of optical inspection. This invention uses a data acquisition module to obtain ambient humidity and ambient light intensity data as synchronous environmental parameters. The raw material feature classification module determines a first correction coefficient based on the ambient humidity data and a second correction coefficient based on the ambient light intensity data. The preset classification threshold is then multiplied by the first and second correction coefficients to obtain the corrected classification threshold. This achieves real-time environmental adaptive correction of the classification threshold, ensuring the stability and reliability of the classification of coarse-grained, high-transparency and fine-grained, low-transparency raw materials, and laying a data foundation for the accurate matching of subsequent optical modulation strategies.

[0019] In particular, this invention considers the impact of differences in raw material particle size and transmittance on optical detection results. In reality, quartz glass raw materials exhibit significant differences in transmittance due to variations in their formation and processing techniques. Coarse-grained, high-transmittance raw materials have long internal optical path lengths and weak scattering, making it difficult to simultaneously image surface and internal impurities. Fine-grained, low-transmittance raw materials have strong light absorption and shallow transmission, making it impossible for conventional light sources to effectively penetrate and identify internal impurities. This invention uses a raw material feature extraction module to extract particle size distribution and transmittance features to determine raw material characteristic values. A raw material feature classification module classifies raw materials into coarse-grained, high-transmittance or fine-grained, low-transmittance types based on these characteristic values ​​and synchronous environmental parameters. An active optical modulation module applies polarization-modulated light or wavelength-selective modulation light accordingly, achieving differentiated illumination strategy configurations for raw materials with different optical properties. This allows both surface reflection and internal transmission impurities of coarse-grained, high-transmittance raw materials to be effectively captured under cross-polarized light irradiation, while fine-grained, low-transmittance raw materials experience enhanced transmission at specific wavelengths under wavelength-selective modulation light to identify internal wall impurities. This overcomes the problem of both missed and over-detection caused by a single light source, improving the accuracy of impurity detection.

[0020] In particular, this invention optimizes the synergistic and complementary characteristics of polarization-modulated light and wavelength-selective modulation light. In practice, coarse-grained, high-transmittance raw materials have high surface smoothness and strong specular reflection. Under conventional light source illumination, surface scratches and reflective impurities such as metal particles have extremely low contrast with the substrate. Polarization-modulated light, through cross-polarization configuration, can effectively suppress specular reflection interference and significantly enhance the imaging clarity of surface defects. However, for fine-grained, low-transmittance raw materials, the particle surfaces are rough and densely packed. Incident light is repeatedly refracted and scattered between particle interfaces, and the polarization direction tends to become disordered, rendering the phase contrast mechanism of polarization-modulated light ineffective. Conversely, wavelength-selective modulation light suppresses short-wave scattering noise through narrow-band spectral filtering and utilizes long-wave penetration characteristics to achieve deep identification of absorbing impurities such as gas-liquid inclusions inside fine-grained, low-transmittance raw materials. However, for coarse-grained, high-transmittance raw materials, their spectral response curves are flat, and wavelength-selective modulation light lacks characteristic contrast, resulting in poor detection performance. This invention utilizes an active optical modulation module to selectively apply polarization-modulated light or wavelength-selective modulation light based on the raw material type. The modulation analysis and processing module then performs feature extraction by acquiring cross-polarized images or grayscale analysis by acquiring transmission spectrum images. This achieves complementary advantages and adaptability between the two types of modulation light, avoiding redundant configuration and mutual interference of optical resources. The system can maintain optimal detection status without hardware modification when switching between different raw material batches, thus enhancing the system's process adaptability and operational economy. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the structural connection of the machine vision-based impurity detection and separation system for quartz glass raw materials according to an embodiment of the present invention; Figure 2This is a logic block diagram illustrating how the quartz glass raw material is divided into coarse-grained high-transparency raw material or fine-grained low-transparency raw material according to an embodiment of the present invention. Figure 3 This is a logic diagram illustrating the determination of the modulated light image parsing method based on the raw material type in an embodiment of the present invention. Figure 4 This is a logic block diagram of issuing a warning signal according to an embodiment of the present invention. Detailed Implementation

[0022] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0023] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0024] It should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the term "connected" should be interpreted broadly. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0025] Please see Figure 1 The diagram shown is a structural connection schematic of a machine vision-based quartz glass raw material impurity detection and separation system according to an embodiment of the present invention. The machine vision-based quartz glass raw material impurity detection and separation system of the present invention includes: The data acquisition module is used to acquire image datasets of quartz glass raw materials and synchronize environmental parameters; The raw material feature extraction module, which is connected to the data acquisition module, is used to preprocess the image dataset, extract the particle size distribution features and transmittance features of the quartz glass raw material, and determine the raw material feature values. The raw material feature classification module, which is connected to the raw material feature extraction module, is used to classify the quartz glass raw material into coarse-grained high-transparency raw material or fine-grained low-transparency raw material based on the raw material feature value, the synchronous environmental parameters and the preset classification threshold. An active optical modulation module, which is connected to the raw material feature classification module, is used to determine the type of modulation light applied to the quartz glass raw material based on the raw material type; The modulation analysis and processing module is connected to the active optical modulation module and is used to acquire the modulation light image of the quartz glass raw material under the modulation light irradiation, analyze the modulation light image to output impurity information and determine whether to issue a warning. The synchronous environmental parameters include ambient humidity data and ambient light intensity data, and the modulation light type includes polarization modulation light and wavelength selective modulation light.

[0026] Specifically, the polarization-modulated light refers to the linearly polarized light output after being modulated by a linear polarizer. This can effectively suppress specular reflection light from the surface of the quartz glass raw material, allowing only depolarized scattered light generated by impurity particles to enter the camera, thereby highlighting surface scratches and attached impurities against a dark background.

[0027] Specifically, the wavelength-selective modulated light refers to narrowband infrared pulse light whose center wavelength matches the characteristic absorption peak of common impurities in quartz glass raw materials. It can penetrate quartz glass raw material particles and generate selective absorption or transmission differences at internal impurities, so that internal gas-liquid inclusions and other impurities appear as dark spots in the transmission image, thereby achieving effective identification of internal impurities.

[0028] Specifically, there are no restrictions on the specific forms of the data acquisition module, raw material feature extraction module, raw material feature classification module, active optical modulation module, and modulation analysis and processing module. They can be composed of logic components, including field-programmable processors, computers, or microprocessors within computers. Those skilled in the art can select appropriate models based on the on-site response requirements, which will not be elaborated further here.

[0029] Specifically, this invention synchronously acquires ambient humidity and ambient light intensity data through a data acquisition module. A raw material feature classification module determines a first correction coefficient based on ambient humidity and a second correction coefficient based on ambient light intensity, and then multiplies these coefficients with a preset classification threshold to obtain a corrected classification threshold. After comparing the raw material feature values ​​with the corrected classification threshold, the raw material is classified into coarse-grained high-transmittance or fine-grained low-transmittance types. An active optical modulation module selects polarization-modulated light or wavelength-selective modulation light accordingly. A modulation analysis and processing module acquires the corresponding modulation light image and analyzes the impurity content. Overall, this invention achieves closed-loop control by dynamically correcting the classification threshold based on environmental parameters and adaptively switching the illumination strategy according to the optical characteristics of the raw material. This effectively overcomes the problems of missed and over-detection caused by environmental fluctuations and raw material differences, improves the accuracy of impurity detection, and meets the stringent requirements for detection consistency in the continuous and stable production of high-purity quartz glass raw materials.

[0030] It is understood that this invention can be directly integrated into a continuous conveying production line for high-purity quartz glass raw materials, enabling real-time online detection and early warning of impurities on the surface and inside of quartz sand or quartz blocks, effectively ensuring the purity stability of raw materials in subsequent melting processes. Furthermore, the technical solution of this invention is not limited to quartz glass raw materials, but can also be extended to other non-metallic mineral materials with similar high light transmittance and large particle size distribution differences, such as the purification process of minerals like feldspar and calcite. Those skilled in the art can link the early warning signal with the automatic sorting device of the production line to achieve closed-loop automatic quality control, which will not be elaborated further here.

[0031] Specifically, the data acquisition module is used to acquire image datasets of quartz glass raw materials and synchronize environmental parameters, including: A continuous motion image sequence of the quartz glass raw material is acquired, and an image dataset is determined at a preset time interval; Collect the synchronization environment parameters corresponding to the image dataset.

[0032] Specifically, the data acquisition module determines the image dataset at preset time intervals, including: Obtain the timestamp of each frame in the continuous motion image sequence; The continuous motion image sequence is sampled according to the preset time interval, and the sampled image frames are used to construct the image dataset. The preset time interval is adjusted according to the movement speed of the quartz glass raw material.

[0033] Specifically, the quartz glass raw material moves continuously under the drive of the conveyor belt, and a camera deployed above the conveyor belt continuously acquires the continuous motion image sequence at a fixed frame rate. Since the raw material is in motion, there is spatial displacement between adjacent frames. By sampling the continuous motion image sequence at preset time intervals, data redundancy can be effectively reduced and the computational burden of subsequent processing can be reduced while ensuring sufficient coverage of the moving raw material.

[0034] Specifically, the preset time interval ranges from [0.5 seconds to 5 seconds], preferably from [1 second to 2 seconds]. An excessively long preset time interval may lead to missed detections, affecting the accuracy of impurity detection; conversely, an excessively short preset time interval results in small differences between adjacent image datasets, generating a large amount of redundant information and increasing the computational burden of subsequent processing. Setting the preset time interval within the range of [0.5 seconds to 5 seconds] ensures both the accuracy of impurity detection and avoids data redundancy, achieving a balance between detection accuracy and computational efficiency.

[0035] Specifically, the data acquisition module acquires the synchronization environment parameters corresponding to the image dataset, including: In response to the construction of the image dataset, environmental humidity data and ambient light intensity data corresponding to the image dataset are collected.

[0036] Specifically, the ambient humidity data is used to characterize the water vapor content in the air around the conveyor belt; fluctuations in ambient humidity will change the thickness of the water film adsorbed on the surface of the quartz glass raw material, thereby affecting the light scattering characteristics of the particle surface and causing the image grayscale distribution to drift.

[0037] Specifically, the ambient light intensity data is used to characterize the degree of interference from external ambient light on image acquisition. Changes in ambient light intensity directly affect the overall brightness and contrast of the image.

[0038] Specifically, the raw material feature extraction module is used to preprocess the image dataset and extract the particle size distribution and transmittance features of the quartz glass raw material, including... Perform grayscale statistics on the image data in the image dataset to obtain the mean grayscale value and grayscale variance of each image data. The overall grayscale distribution parameters of the image dataset are calculated based on the mean grayscale value and grayscale variance of each image to determine the granularity distribution characteristics. The light transmittance characteristics of the quartz glass raw material are determined based on the overall grayscale mean.

[0039] First, calculate the overall grayscale mean and overall grayscale variance of the image dataset. The overall grayscale mean refers to the average grayscale value of all pixels in all image frames, and the overall grayscale variance refers to the dispersion of the grayscale values ​​of all pixels relative to this average.

[0040] Then, the overall grayscale variance is divided by the overall grayscale mean, and the result is the overall grayscale distribution parameter. This parameter is directly used as the particle size distribution characteristic of the quartz glass raw material.

[0041] Specifically, when the raw material particles are coarse, the image has more shadows, lower brightness, and greater gray-level dispersion, resulting in a relatively large overall gray-level variance and a relatively small overall gray-level mean. Therefore, the overall gray-level distribution parameter has a larger value. Conversely, when the raw material particles are fine, the image has uniform brightness, fewer shadows, a relatively small overall gray-level variance, and a relatively large overall gray-level mean. Therefore, the overall gray-level distribution parameter has a smaller value. Thus, the magnitude of this parameter can be used to determine whether the particle size distribution of the raw material is biased towards coarse or fine particles.

[0042] Specifically, the raw material feature extraction module determines the raw material feature values ​​including: The ratio of the particle size distribution characteristics to the preset particle size priority value is determined as the particle size parameter; The ratio of the light transmittance characteristic to a preset light transmittance priority value is determined as the light transmittance parameter; The weighted sum of the particle size parameter and the transmittance parameter is determined as the characteristic value of the raw material.

[0043] Specifically, during the system trial operation, at least 30 sets of known qualified quartz glass raw material samples are collected, and the sample particle size distribution characteristic values ​​of each set of samples are extracted. The average value and standard deviation of the sample particle size distribution characteristic values ​​are calculated, and the result obtained by subtracting one standard deviation from the average value is used as the preset particle size priority value.

[0044] Specifically, during the system trial operation, at least 30 sets of known qualified quartz glass raw material samples are collected, and the light transmittance characteristic values ​​of each set of samples are extracted. The average value and standard deviation of the light transmittance characteristic values ​​are calculated, and the result obtained by subtracting one standard deviation from the average value is used as the preset light transmittance priority value.

[0045] Specifically, the weighting weight of the particle size parameter is 0.4, and the weighting weight of the transmittance parameter is 0.6. For impurity detection in quartz glass raw materials, transmittance directly determines the material's ability to transmit modulated light and the imaging quality, making it a core factor influencing the selection of subsequent optical modulation strategies. Particle size distribution characteristics, on the other hand, mainly affect the material's bulk density and surface scattering properties, contributing relatively less to the detection strategy. Assigning a higher weight to the transmittance parameter allows the material's characteristic values ​​to more sensitively reflect its light transmittance performance, thus prioritizing the dominant role of transmittance during classification and improving the accuracy of distinguishing between coarse-grained, high-transmittance and fine-grained, low-transmittance types.

[0046] Specifically, by sampling continuous motion image sequences and collecting synchronous environmental parameters, the representativeness of the input data and environmental consistency were ensured. Gray-scale statistics were used to determine particle size distribution and transmittance characteristics, achieving a quantitative characterization of the macroscopic physical properties of the raw materials. By introducing preset particle size and transmittance priority values ​​for feature normalization, the absolute dimensional differences between different batches of raw materials were eliminated. Weighted summation was used to obtain raw material feature values, and the core influencing factors were highlighted based on transmittance-dominated weight allocation. The method of determining priority values ​​by collecting qualified samples during trial operation enabled the system to possess self-learning and calibration capabilities. The synergistic effect of these steps ensured that the raw material feature classification module could obtain stable, accurate, and environmentally adaptive raw material feature values, thus laying a reliable data foundation for the subsequent active optical modulation module to accurately switch between polarization-modulated light and wavelength-selective modulation light according to the raw material type.

[0047] Please see Figure 2 As shown, Figure 2This is a logic block diagram illustrating how the quartz glass raw material is classified into coarse-grained high-transparency raw material or fine-grained low-transparency raw material according to an embodiment of the present invention. The raw material feature classification module is used to classify the quartz glass raw material into coarse-grained high-transparency raw material or fine-grained low-transparency raw material based on the raw material feature values, the synchronization environment parameters, and a preset classification threshold. The preset classification threshold is corrected based on the synchronous environment parameters to determine the corrected classification threshold; the raw material characteristic value is compared with the corrected classification threshold. If the raw material characteristic value is greater than or equal to the modified classification threshold, the quartz glass raw material is classified as a coarse-grained high-transparency raw material. If the raw material characteristic value is less than the modified classification threshold, the quartz glass raw material is classified as a fine-grained, low-transparency raw material.

[0048] Specifically, the method for determining the preset classification threshold is as follows: During the system trial operation, at least 50 groups of quartz glass raw material samples with known category labels are collected, of which at least 25 groups are coarse-grained high-transparency samples and at least 25 groups are fine-grained low-transparency samples; the raw material characteristic values ​​of each group of samples are calculated, and the distribution curves of the raw material characteristic values ​​of the two types of samples are plotted. The median value between the minimum value of the raw material characteristic value of the coarse-grained high-transparency sample and the maximum value of the raw material characteristic value of the fine-grained low-transparency sample is taken as the preset classification threshold.

[0049] Specifically, the step of correcting the preset classification threshold based on the synchronization environment parameters to determine the corrected classification threshold includes: A first correction factor is determined based on the environmental humidity data; The second correction factor is determined based on the ambient light intensity data; The product of the preset classification threshold and the first correction coefficient and the second correction coefficient is used as the corrected classification threshold.

[0050] Specifically, the first correction coefficient is determined based on the environmental humidity data as follows: when the relative humidity is less than or equal to 50%, the first correction coefficient is 1.0; when the relative humidity is greater than 50% and less than or equal to 75%, the first correction coefficient is 0.95; and when the relative humidity is greater than 75%, the first correction coefficient is 0.85. Increased humidity leads to a thicker adsorbed water film on the surface of quartz glass raw materials, enhanced light scattering, and a decrease in the overall grayscale of the image. This results in a lower calculated transmittance characteristic value, causing the raw material characteristic value to tend to be classified as fine-grained and low-transmittance. Therefore, compensation is achieved by lowering the classification threshold; the higher the humidity, the smaller the correction coefficient, to avoid misclassification of coarse-grained and high-transmittance raw materials.

[0051] Specifically, the second correction coefficient is determined based on the ambient light intensity data as follows: when the ambient light intensity is less than or equal to 500 lux, the second correction coefficient is 1.0; when the ambient light intensity is greater than 500 lux and less than or equal to 1000 lux, the second correction coefficient is 1.05; and when the ambient light intensity is greater than 1000 lux, the second correction coefficient is 1.10. Increased light intensity leads to enhanced interference from external ambient light, increased overall image brightness, higher grayscale mean, and a larger calculated transmittance characteristic value, causing the raw material characteristic value to tend to be classified as coarse-grained and high-transmittance. Therefore, compensation is achieved by increasing the classification threshold; the stronger the light, the larger the correction coefficient, to avoid over-detection of fine-grained and low-transmittance raw materials.

[0052] Please see Figure 3 As shown, Figure 3 This is a logical diagram illustrating the method for determining the modulation light image analysis based on the raw material type in this invention. The modulation analysis and processing module acquires the modulation light image of the quartz glass raw material under the modulation light irradiation, and analyzes the modulation light image, including... If the quartz glass raw material is classified as a coarse-grained, high-transparency type, then a cross-polarization image of the quartz glass raw material under the polarization modulation light is acquired, and feature extraction is performed on the cross-polarization image to determine the impurity content information. If the quartz glass raw material is classified as a fine-grained, low-transmittance type, then the transmission spectrum image of the quartz glass raw material under wavelength selectively modulated light is collected, and grayscale analysis is performed on the transmission spectrum image to determine the impurity content information.

[0053] Specifically, the step of extracting features from the cross-polarization image to determine impurity content information includes: The cross-polarized image is segmented, and connected regions with gray values ​​higher than a preset highlight threshold are extracted as candidate impurity regions. The area and cumulative pixel gray value of each candidate impurity region are calculated.

[0054] The sum of the areas of each candidate impurity region is divided by the total image area to obtain the proportion of polarization impurity area. Based on this proportion, the impurity content information is determined.

[0055] Specifically, the impurity content information is determined based on the area ratio of the polarization impurities.

[0056] Specifically, the method for determining the impurity content information based on the polarization impurity area ratio is as follows: multiple groups of quartz glass raw material samples with different impurity contents are collected in advance, and the polarization impurity area ratio of each group of samples is obtained. A linear fitting method is used to establish a mapping relationship between the polarization impurity area ratio and the impurity content. In actual testing, the currently calculated polarization impurity area ratio is substituted into this mapping relationship to determine the corresponding impurity content information, which is then output in the form of a mass percentage.

[0057] Specifically, performing grayscale analysis on the transmission spectrum image to determine impurity content information includes: The gray-level histogram of the transmission spectrum image is statistically analyzed to determine the range of the main gray-level peak in the background. Pixels with gray-level values ​​lower than the lower limit of the main background peak are identified as absorption-type impurity pixels.

[0058] The total area of ​​absorbing impurity pixels is statistically analyzed, and their proportion of the total image area is calculated to obtain the proportion of transmitted impurity area. Based on the proportion of transmitted impurity area, the impurity content information is estimated.

[0059] Specifically, multiple sets of standard quartz glass raw material samples with different impurity contents are collected in advance, and the proportion of the transmitted impurity area of ​​each set of samples is obtained. A linear fitting method is used to establish a mapping relationship between the proportion of the transmitted impurity area and the impurity content. In actual testing, the currently calculated proportion of the transmitted impurity area is substituted into this mapping relationship to estimate the corresponding impurity content information, which is then output in the form of mass percentage.

[0060] Specifically, the impurity content information is estimated based on the proportion of the area of ​​transmitted impurities.

[0061] Please see Figure 4 The above, Figure 4 This is a logic block diagram of issuing a warning signal according to an embodiment of the present invention. The modulation analysis and processing module outputs impurity information and determines whether to issue a warning, including: In response to the analysis result of the modulated light image, the content information of the impurities is output; If the impurity content exceeds the warning threshold, a warning signal will be issued.

[0062] Specifically, the method for determining the warning threshold is as follows: Several groups of quartz glass raw material samples with different impurity contents are collected in advance, and each group of samples is melted and molded under the same melting process conditions, and the actual impurity content of the products obtained in each group is detected; with the impurity content of the raw material as the horizontal axis and the impurity content of the product as the vertical axis, a corresponding relationship curve between the impurity content of the raw material and the impurity content of the product is established.

[0063] Specifically, based on the maximum permissible impurity content of quartz glass products in the downstream melting process, the corresponding raw material impurity content is retrieved from the correlation curve and determined as the warning threshold. When the system detects that the impurity content in the raw material exceeds this warning threshold, it indicates that the impurity content of the product obtained after melting the raw material will exceed the process allowable range, and a warning needs to be issued promptly.

[0064] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A machine vision-based system for detecting and separating impurities in quartz glass raw materials, characterized in that, include: The data acquisition module is used to acquire image datasets of quartz glass raw materials and synchronize environmental parameters; The raw material feature extraction module is used to preprocess the image dataset, extract the particle size distribution features and transmittance features of the quartz glass raw material, and determine the raw material feature values. The raw material characteristic classification module is used to classify the quartz glass raw material into coarse-grained high-transparency raw material or fine-grained low-transparency raw material based on the raw material characteristic value, the synchronous environmental parameters and the preset classification threshold. An active optical modulation module is used to determine the type of modulated light applied to the quartz glass material based on the material type; The modulation analysis and processing module is used to acquire the modulation light image of the quartz glass raw material under the modulation light irradiation, analyze the modulation light image to output impurity information and determine whether to issue a warning. The synchronous environmental parameters include ambient humidity data and ambient light intensity data, and the modulation light type includes polarization modulation light and wavelength selective modulation light.

2. The machine vision-based impurity detection and separation system for quartz glass raw materials according to claim 1, characterized in that, The data acquisition module is used to acquire image datasets of quartz glass raw materials and synchronize environmental parameters, including... A continuous motion image sequence of the quartz glass raw material is acquired, and an image dataset is determined at a preset time interval; Collect the synchronization environment parameters corresponding to the image dataset.

3. The machine vision-based impurity detection and separation system for quartz glass raw materials according to claim 2, characterized in that, The data acquisition module determines the image dataset at preset time intervals, including: Obtain the timestamp of each frame in the continuous motion image sequence; The continuous motion image sequence is sampled according to the preset time interval, and the sampled image frames are used to construct the image dataset. The preset time interval is adjusted according to the movement speed of the quartz glass raw material.

4. The machine vision-based impurity detection and separation system for quartz glass raw materials according to claim 2, characterized in that, The data acquisition module collects the synchronization environment parameters corresponding to the image dataset, including... In response to the construction of the image dataset, environmental humidity data and ambient light intensity data corresponding to the image dataset are collected.

5. The machine vision-based impurity detection and separation system for quartz glass raw materials according to claim 1, characterized in that, The raw material feature extraction module is used to preprocess the image dataset and extract the particle size distribution and transmittance features of the quartz glass raw material, including... Perform grayscale statistics on the image data in the image dataset to obtain the mean grayscale value and grayscale variance of each image data. The overall grayscale distribution parameters of the image dataset are calculated based on the mean grayscale value and grayscale variance of each image to determine the granularity distribution characteristics. The light transmittance characteristics of the quartz glass raw material are determined based on the overall grayscale mean.

6. The machine vision-based impurity detection and separation system for quartz glass raw materials according to claim 5, characterized in that, The raw material feature extraction module determines the raw material feature values, including: The ratio of the particle size distribution characteristics to the preset particle size priority value is determined as the particle size parameter; The ratio of the light transmittance characteristic to a preset light transmittance priority value is determined as the light transmittance parameter; The weighted sum of the particle size parameter and the transmittance parameter is determined as the characteristic value of the raw material; Specifically, the average value and standard deviation of the sample particle size distribution characteristic value are calculated, and the result obtained by subtracting one standard deviation from the average value is used as the preset particle size priority value. The average value and standard deviation of the sample transmittance characteristic value are calculated, and the result obtained by subtracting one standard deviation from the average value is used as the preset transmittance priority value.

7. The machine vision-based impurity detection and separation system for quartz glass raw materials according to claim 1, characterized in that, The raw material characteristic classification module is used to classify the quartz glass raw material into coarse-grained high-transparency raw material or fine-grained low-transparency raw material based on the raw material characteristic value, the synchronous environmental parameters, and a preset classification threshold. The preset classification threshold is corrected based on the synchronization environment parameters to determine the corrected classification threshold; The raw material characteristic value is compared with the corrected classification threshold; If the raw material characteristic value is greater than or equal to the modified classification threshold, the quartz glass raw material is classified as a coarse-grained high-transparency raw material. If the raw material characteristic value is less than the modified classification threshold, the quartz glass raw material is classified as a fine-grained, low-transparency raw material.

8. The machine vision-based impurity detection and separation system for quartz glass raw materials according to claim 7, characterized in that, The step of correcting the preset classification threshold based on the synchronization environment parameters, to determine the corrected classification threshold, includes: A first correction factor is determined based on the environmental humidity data; The second correction factor is determined based on the ambient light intensity data; The product of the preset classification threshold and the first correction coefficient and the second correction coefficient is used as the corrected classification threshold.

9. The machine vision-based impurity detection and separation system for quartz glass raw materials according to claim 1, characterized in that, The modulation analysis and processing module acquires a modulation light image of the quartz glass raw material under the modulation light irradiation, and analyzes the modulation light image, including... If the quartz glass raw material is classified as a coarse-grained, high-transparency type, then a cross-polarization image of the quartz glass raw material under the polarization modulation light is acquired, and feature extraction is performed on the cross-polarization image to determine the impurity content information. If the quartz glass raw material is classified as a fine-grained, low-transmittance type, then the transmission spectrum image of the quartz glass raw material under wavelength selectively modulated light is collected, and grayscale analysis is performed on the transmission spectrum image to determine the impurity content information.

10. The machine vision-based impurity detection and separation system for quartz glass raw materials according to claim 9, characterized in that, The modulation analysis and processing module outputs impurity information and determines whether to issue a warning, including: In response to the analysis result of the modulated light image, the content information of the impurities is output; If the impurity content exceeds the warning threshold, a warning signal will be issued.

Images