Method and system for detecting appearance defects of casting products based on machine vision

Abstract

The invention discloses a method and system for detecting appearance defects of casting products based on machine vision, and specifically relates to the technical field of defect detection of casting products. The invention classifies various casting materials of a production line into different categories, identifies casting materials with abnormal absorption rate, performs terahertz wave scanning on them, obtains scattering images and evaluates the stability of scattering point density distribution. The invention uses a terahertz time-domain spectrometer to scan, analyzes the amplitude fluctuation state of the time-domain signal and evaluates the degree of waveform distortion. The image quality captured by the imaging system is evaluated after comprehensive analysis, and is divided into different levels. Defect detection is directly performed on high-quality images, and the accuracy of medium-quality images is analyzed and the emission power and scanning speed of the terahertz imaging device are dynamically adjusted to improve the imaging quality, thereby significantly improving the accuracy and efficiency of detection, ensuring the quality and safety of casting products, and effectively avoiding product quality problems and potential equipment failures caused by untimely defect detection.

Description

Technical Field

[0001] The present invention relates to the technical field of casting product defect detection, and in particular to a casting product appearance defect detection method and system based on machine vision. Background Art

[0002] Machine vision-based casting product appearance defect detection refers to the use of machine vision technology to automatically detect the surface of casting products to identify and classify surface defects. This technology uses a high-resolution camera to collect images of the casting surface, and uses image processing and machine learning algorithms to analyze the images, which can accurately detect defects such as pores, cracks, shrinkage holes, etc. on the casting surface.

[0003] In the prior art, when using a terahertz imaging system to detect the appearance defects of casting products, the terahertz imaging device transmits terahertz waves to penetrate the casting products, and uses a receiver to capture the terahertz wave signals that penetrate the castings, generating high-resolution three-dimensional images, which can accurately detect defects such as tiny pores, cracks, and inclusions on the surface and inside of the castings. Due to the high penetration and non-destructive testing characteristics of terahertz waves, terahertz imaging systems are particularly suitable for comprehensive testing of castings with complex geometries, providing more sophisticated defect recognition capabilities than traditional X-ray or ultrasonic testing, thereby significantly improving the quality control level of products.

[0004] The prior art has the following deficiencies:

[0005] In the prior art, when using a terahertz imaging system to detect appearance defects of casting products, different materials may have different absorption rates of terahertz waves. For example, certain metals and high-density alloys may highly absorb terahertz waves, resulting in a significant reduction in the intensity of the terahertz waves transmitted to the terahertz imaging device, so that the imaging system may only be able to capture blurry or incomplete images. At the same time, if the image quality is low, the detection system may not be able to accurately identify and classify small defects on the surface of the casting, such as microcracks or pores, which will not only cause defective castings to go undetected and enter the market, leading to product quality and safety issues, but may also cause serious equipment failures and accidents due to the inability to detect and deal with defects early. Summary of the invention

[0006] The purpose of the present invention is to provide a method and system for detecting appearance defects of casting products based on machine vision to solve the shortcomings of the background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a method for detecting appearance defects of casting products based on machine vision, comprising the following steps:

[0008] S1: According to the characteristic information of each casting material on the production line, it is divided into different material categories, and terahertz wave detection is performed on different categories of casting materials to obtain their absorption rate in real time. The real-time absorption rate of different casting materials is analyzed to identify casting materials with abnormal absorption rate;

[0009] S2: Perform terahertz wave scanning on casting materials with abnormal absorption rate, obtain scattering images, determine the abnormal changes in the scattering point density distribution area in the scattering image, and evaluate the stability of the scattering point density distribution;

[0010] S3: Scan the casting material with abnormal absorption rate through the terahertz time-domain spectrometer to obtain the time domain signal when the terahertz wave passes through the casting material, analyze the amplitude fluctuation state of the time domain signal, and evaluate the distortion degree of the time domain signal waveform;

[0011] S4: After comprehensively analyzing the stability of the scattering point density distribution and the distortion degree of the time domain signal waveform, the quality of the casting product appearance image captured by the imaging system is evaluated;

[0012] S5: Divide the quality of the casting product appearance images captured by the imaging system into high-quality images, medium-quality images and low-quality images. For high-quality images, directly perform defect detection and analysis, and for low-quality images, exclude them;

[0013] S6: For medium-quality images, analyze the accuracy of the imaging system in capturing the appearance image of the casting product within a fixed time period. Based on the analysis results, dynamically adjust the emission power and scanning speed of the terahertz imaging device to improve the imaging quality.

[0014] In a preferred embodiment, the terahertz wave absorptivity data acquired in real time is analyzed by a random forest algorithm to identify casting materials with abnormal absorptivity, specifically:

[0015] Obtain the absorption rate data of casting materials in real time from the terahertz wave detection equipment;

[0016] Using historical data, normal samples are marked as 0 and abnormal samples are marked as 1 to form a labeled data set;

[0017] Divide the dataset into training and testing sets;

[0018] Use the training set to train the random forest model and use the test set to evaluate the model performance;

[0019] Input the terahertz wave absorption rate data acquired in real time into the trained random forest model;

[0020] Extract features from real-time data and input the extracted feature data into the random forest model for classification and judgment;

[0021] If the model output is 1, the absorption rate of the casting material is abnormal, and it is marked as a casting material with abnormal absorption rate; if the model output is 0, the absorption rate of the casting material is normal, and it is marked as a casting material with normal absorption rate.

[0022] In a preferred embodiment, a scattering point density distribution anomaly index is generated according to the abnormal changes in the scattering point density distribution area in the scattering map to evaluate the stability of the scattering point density distribution. The method for obtaining the scattering point density distribution anomaly index is:

[0023] Get the scatter point density distribution data set, the data set contains N samples, each sample has d features, ;in, Represents the eigenvector of the i-th sample, standardizes the data, and calculates the covariance matrix Σ of the standardized data set. ; In the formula, is the mean vector of the standardized data, is the i-th sample after standardization, T is the matrix transpose; perform eigenvalue decomposition on the covariance matrix Σ to obtain eigenvalues ​​λ1,λ2,…,λd and corresponding eigenvectors v1,v2,…,vd; ; , d is a positive integer greater than 0; select the matrix V composed of the first k eigenvectors as the principal component matrix V=[v1,v2,…,vk] according to the size of the eigenvalue; project the standardized data X′ into the principal component space to obtain a low-dimensional representation Z, ; Where Z is an N×k matrix, representing the low-dimensional data projected into the principal component space, and the reconstruction error of each sample is calculated, that is, the difference between the original data and the reconstructed data. in, is the reconstructed data, and the scattering point density distribution anomaly index is calculated. The square sum of the reconstruction error is used as the scattering point density distribution anomaly index. The specific calculation expression is: ; In the formula, is the anomaly index of scatter point density distribution.

[0024] In a preferred embodiment, the acquired scattering point density distribution anomaly index is compared with a preset scattering point density distribution anomaly index reference threshold. If the scattering point density distribution anomaly index is greater than or equal to the scattering point density distribution anomaly index reference threshold, the more obvious the abnormal change of the scattering point density distribution area in the scattering map, the lower the stability of the scattering point density distribution, and a scattering point density distribution anomaly signal is generated at this time; if the scattering point density distribution anomaly index is less than the scattering point density distribution anomaly index reference threshold, the less obvious the abnormal change of the scattering point density distribution area in the scattering map, the higher the stability of the scattering point density distribution, and a scattering point density distribution normal signal is generated at this time.

[0025] In a preferred embodiment, characteristic information is extracted from the processed time domain signal, the amplitude fluctuation state of the time domain signal is analyzed, and a peak amplitude fluctuation index is generated according to the amplitude change of the terahertz wave to evaluate the degree of distortion of the time domain signal waveform. The peak amplitude fluctuation index is obtained as follows:

[0026] The time domain signal Decomposed into several intrinsic mode functions IMF and a residual term, the decomposition expression is: ; In the formula, is the kth eigenmode function, r(t) is the residual term;

[0027] Perform Hilbert transform on each IMF, calculate the Hilbert transform of each IMF, and obtain its instantaneous amplitude and instantaneous phase. ; is the Hilbert transform of the kth IMF, is the circumference of a circle, s is the adjacent moment of time t; calculate the instantaneous amplitude of the kth IMF and instantaneous phase , the specific calculation expression is: ; In the formula, is the Hilbert transform of the kth IMF, is the instantaneous amplitude of the kth IMF; ; In the formula, is the instantaneous phase of the kth IMF;

[0028] Calculate the peak amplitude of each IMF , ; Calculate the standard deviation of the peak amplitude of each IMF, and calculate the peak amplitude fluctuation index after summing up the standard deviations of the peak amplitudes of all IMFs.

[0029] In a preferred embodiment, the acquired peak amplitude fluctuation index is compared with a preset peak amplitude fluctuation index reference threshold. If the peak amplitude fluctuation index is greater than or equal to the peak amplitude fluctuation index reference threshold, the more obvious the distortion of the time domain signal waveform is, the more an amplitude fluctuation abnormal signal is generated; if the peak amplitude fluctuation index is less than the peak amplitude fluctuation index reference threshold, the less obvious the distortion of the time domain signal waveform is, the more an amplitude fluctuation normal signal is generated.

[0030] In a preferred embodiment, the scattering point density distribution anomaly index and the peak amplitude fluctuation index are normalized, and the quality assessment coefficient of the casting product appearance image captured by the imaging system is calculated by the normalized scattering point density distribution anomaly index and the peak amplitude fluctuation index;

[0031] Compare the quality assessment coefficient of the casting product appearance image captured by the imaging system with the gradient standard threshold, the gradient standard threshold includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold, and compare the quality assessment coefficient of the casting product appearance image captured by the imaging system with the first standard threshold and the second standard threshold respectively;

[0032] If the quality assessment coefficient of the casting product appearance image captured by the imaging system is greater than the second standard threshold, the corresponding casting product appearance image is classified as a high-quality image, and defect detection analysis is directly performed on the high-quality image; if the quality assessment coefficient of the casting product appearance image captured by the imaging system is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, the corresponding casting product appearance image is classified as a medium-quality image, and further analysis is performed on the medium-quality image; if the quality assessment coefficient of the casting product appearance image captured by the imaging system is less than the first standard threshold, the corresponding casting product appearance image is classified as a low-quality image, and the low-quality image is excluded.

[0033] In a preferred embodiment, when the appearance image of the casting product captured by the imaging system is classified as a medium quality image, that is, the quality assessment coefficient generated within a fixed time period is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, the quality assessment coefficients generated within a subsequent fixed time period that are greater than or equal to the first standard threshold and less than or equal to the second standard threshold are collected, and a corresponding data set is established, and the mean and standard deviation of the data set are calculated, and the mean and standard deviation of the quality assessment coefficients in the data set are further analyzed to determine the accuracy of the appearance image of the casting product captured by the imaging system within the subsequent fixed time period, divide the accuracy of the appearance image of the casting product into different categories, divide the accuracy of the appearance image of the casting product into accurate images and inaccurate images, and dynamically adjust the emission power and scanning speed of the terahertz imaging device for the inaccurate images to improve the imaging quality.

[0034] In a preferred embodiment, if the mean of the quality assessment coefficients in the data set is greater than or equal to the reference threshold of the mean of the quality assessment coefficients, and the standard deviation of the quality assessment coefficients is less than the reference threshold of the standard deviation of the quality assessment coefficients, it is marked as an accuracy image; if the mean of the quality assessment coefficients is greater than or equal to the reference threshold of the mean of the quality assessment coefficients, and the standard deviation of the quality assessment coefficients is greater than or equal to the reference threshold of the standard deviation of the quality assessment coefficients, it is marked as an inaccuracy image; if the mean of the quality assessment coefficients is less than the reference threshold of the mean of the quality assessment coefficients, and the standard deviation of the quality assessment coefficients is greater than or equal to the reference threshold of the standard deviation of the quality assessment coefficients, it is marked as an inaccuracy image; if the mean of the quality assessment coefficients is less than the reference threshold of the mean of the quality assessment coefficients, and the standard deviation of the quality assessment coefficients is less than the reference threshold of the standard deviation of the quality assessment coefficients, it is marked as an inaccuracy image.

[0035] The present invention also provides a casting product appearance defect detection system based on machine vision, a casting material analysis module, a scattering point density analysis module, a time domain signal waveform evaluation module, a comprehensive analysis module, an image quality classification module and an imaging quality adjustment module;

[0036] Casting material analysis module: According to the characteristic information of each casting material on the production line, it is divided into different material categories, and terahertz wave detection is performed on different categories of casting materials to obtain their absorption rate in real time. The real-time absorption rate of different casting materials is analyzed to identify casting materials with abnormal absorption rate;

[0037] Scattering point density analysis module: performs terahertz wave scanning on casting materials with abnormal absorption rate, obtains scattering images, determines abnormal changes in the scattering point density distribution area in the scattering image, and evaluates the stability of the scattering point density distribution;

[0038] Time domain signal waveform evaluation module: Scan the casting material with abnormal absorption rate through the terahertz time domain spectrometer, obtain the time domain signal when the terahertz wave passes through the casting material, analyze the amplitude fluctuation state of the time domain signal, and evaluate the distortion degree of the time domain signal waveform;

[0039] Comprehensive analysis module: After comprehensively analyzing the stability of the scattering point density distribution and the degree of distortion of the time domain signal waveform, the quality of the appearance image of the casting product captured by the imaging system is evaluated;

[0040] Image quality classification module: the quality of the casting product appearance images captured by the imaging system is classified into high-quality images, medium-quality images and low-quality images. For high-quality images, defect detection and analysis are directly performed, and for low-quality images, defects are excluded.

[0041] Imaging quality adjustment module: For medium-quality images, the accuracy of the imaging system in capturing the appearance image of the casting product within a fixed time period is analyzed. Based on the analysis results, the emission power and scanning speed of the terahertz imaging device are dynamically adjusted to improve the imaging quality.

[0042] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0043] 1. The present invention realizes high-precision detection of appearance defects of casting products by using machine vision and terahertz imaging technology. The specific method includes classifying casting materials, acquiring and analyzing their absorption rates in real time, identifying and scanning materials with abnormal absorption rates, and evaluating the stability of the scattering point density distribution and the degree of distortion of the time domain signal waveform. Based on the results of the comprehensive analysis, the images captured by the imaging system are divided into quality categories, and the medium-quality images are further analyzed. The imaging quality is improved by dynamically adjusting the emission power and scanning speed of the terahertz imaging device.

[0044] 2. The present invention can not only significantly improve the accuracy and efficiency of casting product appearance defect detection, but also ensure product quality and safety. By dynamically adjusting imaging parameters, the imaging effect of medium-quality images is optimized, quality and safety hazards caused by detection errors are avoided, equipment failures and accidents are effectively prevented, and the reliability and production efficiency of the production line are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0046] Figure 1The figure is a flow chart of the method of the present invention.

[0047] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Example 1

[0049] See also Figure 1 As shown, the casting product appearance defect detection method based on machine vision described in this embodiment includes the following steps:

[0050] S1: According to the characteristic information of each casting material on the production line, it is divided into different material categories, and terahertz wave detection is performed on different categories of casting materials to obtain their absorption rate in real time. The real-time absorption rate of different casting materials is analyzed to identify casting materials with abnormal absorption rate;

[0051] S2: Perform terahertz wave scanning on casting materials with abnormal absorption rate, obtain scattering images, determine the abnormal changes in the scattering point density distribution area in the scattering image, and evaluate the stability of the scattering point density distribution;

[0052] S3: Scan the casting material with abnormal absorption rate through the terahertz time-domain spectrometer to obtain the time domain signal when the terahertz wave passes through the casting material, analyze the amplitude fluctuation state of the time domain signal, and evaluate the distortion degree of the time domain signal waveform;

[0053] S4: After comprehensively analyzing the stability of the scattering point density distribution and the distortion degree of the time domain signal waveform, the quality of the casting product appearance image captured by the imaging system is evaluated;

[0054] S5: Divide the quality of the casting product appearance images captured by the imaging system into high-quality images, medium-quality images and low-quality images. For high-quality images, directly perform defect detection and analysis, and for low-quality images, exclude them;

[0055] S6: For medium-quality images, analyze the accuracy of the imaging system in capturing the appearance image of the casting product within a fixed time period. Based on the analysis results, dynamically adjust the emission power and scanning speed of the terahertz imaging device to improve the imaging quality.

[0056] Among them, in S1, according to the characteristic information of each casting material of the production line, they are divided into different material categories, and terahertz wave detection is performed on the casting materials of different categories to obtain their absorption rates in real time. The real-time absorption rates of different casting materials are analyzed to identify casting materials with abnormal absorption rates;

[0057] According to the characteristic information of each casting material on the production line, the casting materials are first classified, including physical properties such as chemical composition, density, hardness, etc. By analyzing these characteristic information, the casting materials are divided into different categories, such as cast iron, cast steel, aluminum alloy, copper alloy, etc. A material classification database is established to quickly identify and process different categories of casting materials in subsequent inspections.

[0058] Terahertz wave detection is performed on castings of each material category, and the absorption rate of the casting material is obtained in real time by utilizing the high penetration and non-destructive detection characteristics of terahertz waves. The casting material is scanned by a terahertz wave source and detector, the attenuation of the terahertz wave when passing through the material is recorded, and the absorption rate of each type of material is calculated. This process ensures that the characteristic changes of different materials are accurately recorded during the detection process.

[0059] The random forest algorithm is used to analyze the real-time terahertz wave absorption rate data to identify casting materials with abnormal absorption rates, specifically:

[0060] The absorption rate data of the casting material is obtained in real time from the terahertz wave detection device. The absorption rate data includes multiple features, each feature representing the absorption rate at different frequencies.

[0061] Use methods such as mean filling or interpolation to handle missing values ​​to ensure data integrity. Use statistical methods (such as the 3σ rule) to identify and handle outliers to ensure data accuracy.

[0062] Statistical features such as mean, variance, maximum, minimum, etc. are extracted from the raw absorption rate data. These features will be used for subsequent model training and real-time analysis.

[0063] Using historical data, normal samples are marked as 0 and abnormal samples are marked as 1 to form a labeled data set;

[0064] Divide the data set into training set and test set, generally in the ratio of 70% training set and 30% test set;

[0065] Train a random forest model using the training set.

[0066] Set the hyperparameters of the random forest, such as the number of trees (n_estimators) and the maximum depth of the trees (max_depth).

[0067] Construct multiple decision trees, each tree is randomly sampled from the training set (Bootstrap sampling), and feature selection and splitting are performed on each tree.

[0068] Use the test set to evaluate the model performance and calculate indicators such as accuracy, precision, and recall to verify the effectiveness of the model.

[0069] The terahertz wave absorption rate data acquired in real time is input into the trained random forest model.

[0070] Perform feature extraction on real-time data, such as calculating statistical features such as mean and variance.

[0071] The extracted feature data is input into the random forest model for classification judgment. The model output of 1 indicates abnormal absorption rate, and the output of 0 indicates normal absorption rate.

[0072] According to the results of the model output, the casting material with abnormal absorption rate is identified. If the model output is 1, the casting material absorption rate is abnormal, and it is marked as a casting material with abnormal absorption rate; if the model output is 0, the casting material absorption rate is normal, and it is marked as a casting material with normal absorption rate.

[0073] S2: Perform terahertz wave scanning on casting materials with abnormal absorption rate, obtain scattering images, determine the abnormal changes in the scattering point density distribution area in the scattering image, and evaluate the stability of the scattering point density distribution.

[0074] Use the terahertz wave transmitter and receiver to scan the casting material. Set appropriate scanning parameters (such as frequency, power, scanning speed, etc.) to ensure that a high-quality scattering image is obtained. Perform a comprehensive scan of the casting material with abnormal absorption rate and record the scattering of the terahertz wave in the material. Capture the scattering signal of the terahertz wave in the material through the receiver and generate the corresponding scattering image. Save the generated scattering image for subsequent analysis.

[0075] According to the abnormal changes in the scattering point density distribution area in the scattering map, the scattering point density distribution anomaly index is generated to evaluate the stability of the scattering point density distribution. The method for obtaining the scattering point density distribution anomaly index is as follows:

[0076] Get the scatter point density distribution data set, the data set contains N samples, each sample has d features, ;in, Represents the eigenvector of the i-th sample. The data is standardized so that the mean of each feature is 0 and the variance is 1. The covariance matrix Σ of the standardized data set is calculated. ; Where EX is the mean vector of the standardized data, is the i-th sample after standardization, T is the matrix transpose; perform eigenvalue decomposition on the covariance matrix Σ to obtain eigenvalues ​​λ1,λ2,…,λd and corresponding eigenvectors v1,v2,…,vd; ; , d is a positive integer greater than 0; according to the size of the eigenvalue, the matrix V composed of the first k main eigenvectors is selected as the principal component matrix ; Project the standardized data X′ into the principal component space to obtain a low-dimensional representation Z, ; Where Z is an N×k matrix, representing the low-dimensional data projected into the principal component space, and the reconstruction error of each sample is calculated, that is, the difference between the original data and the reconstructed data. in, is the reconstructed data, and the scattering point density distribution anomaly index is calculated. The square sum of the reconstruction error is used as the scattering point density distribution anomaly index. The specific calculation expression is: ; Where AS is the anomaly index of scattering point density distribution.

[0077] The obtained scattering point density distribution anomaly index is compared with the preset scattering point density distribution anomaly index reference threshold. If the scattering point density distribution anomaly index is greater than or equal to the scattering point density distribution anomaly index reference threshold, it means that the more obvious the abnormal change of the scattering point density distribution area in the scattering map, the lower the stability of the scattering point density distribution, and a scattering point density distribution anomaly signal is generated at this time; if the scattering point density distribution anomaly index is less than the scattering point density distribution anomaly index reference threshold, it means that the less obvious the abnormal change of the scattering point density distribution area in the scattering map, the higher the stability of the scattering point density distribution, and a scattering point density distribution normal signal is generated at this time.

[0078] The larger the anomaly index of the scattering point density distribution, the worse the stability of the scattering point density distribution. A high anomaly index indicates that the scattering point density distribution deviates significantly from the normal distribution and may have significant fluctuations or abnormal changes. This instability usually means that there may be defects inside or on the surface of the material, such as cracks, voids, or uneven material distribution.

[0079] A high scattering point density distribution anomaly index indicates that the scattering point density varies greatly in different areas and fails to maintain a uniform and consistent distribution pattern. This phenomenon is particularly important in terahertz wave detection because it reveals potential problems in the internal structure of the material. These problems may affect the mechanical properties and service life of the casting, leading to product quality and safety hazards.

[0080] S3: Use a terahertz time-domain spectrometer to scan the casting material with abnormal absorption rate, obtain the time domain signal when the terahertz wave passes through the casting material, analyze the amplitude fluctuation state of the time domain signal, and evaluate the degree of distortion of the time domain signal waveform.

[0081] Ensure that the terahertz time-domain spectrometer equipment (including terahertz wave emitter, detector, control system, etc.) has been properly installed and calibrated.

[0082] Clean the surface of casting materials with abnormal absorption rates to ensure there are no surface contaminants or impurities that may affect the measurement.

[0083] Set scanning parameters, such as the emission frequency range of the terahertz wave, scanning speed, resolution, etc.

[0084] Place the casting material at the scanning position of the terahertz time-domain spectrometer, ensuring that the material is on the path of the terahertz wave.

[0085] Start the scanning program, the terahertz wave transmitter starts to emit terahertz pulses, and the detector receives the terahertz waves after passing through the material.

[0086] During the scanning process, the detector captures the propagation and reflection signals of terahertz waves in the material in real time and records time domain data.

[0087] The acquired time domain signal usually includes time and amplitude information, reflecting the propagation characteristics of the terahertz wave in the material.

[0088] The acquired time domain signal is preprocessed, including denoising, smoothing and other processing, to improve the signal quality. Gaussian filtering or wavelet denoising technology can be used for denoising, and moving average filtering can be used for smoothing.

[0089] Extract characteristic information from the processed time domain signal, analyze the amplitude fluctuation state of the time domain signal, generate the peak amplitude fluctuation index according to the amplitude change of the terahertz wave, and evaluate the distortion degree of the time domain signal waveform. The method for obtaining the peak amplitude fluctuation index is as follows:

[0090] The time domain signal m(t) is decomposed into several intrinsic mode functions IMF and a residual term. The decomposition expression is: ; In the formula, is the kth eigenmode function, r(t) is the residual term;

[0091] Perform Hilbert transform on each IMF, calculate the Hilbert transform of each IMF, and obtain its instantaneous amplitude and instantaneous phase. ; is the Hilbert transform of the kth IMF, is the circumference of a circle, s is the adjacent moment of time t; calculate the instantaneous amplitude of the kth IMF and instantaneous phase , the specific calculation expression is: ; In the formula, is the Hilbert transform of the kth IMF, is the instantaneous amplitude of the kth IMF; ; In the formula, is the instantaneous phase of the kth IMF;

[0092] Calculate the peak amplitude of each IMF , ; Calculate the standard deviation of the peak amplitude of each IMF, and calculate the peak amplitude fluctuation index after summing up the standard deviations of the peak amplitudes of all IMFs.

[0093] The obtained peak amplitude fluctuation index is compared with the preset peak amplitude fluctuation index reference threshold. If the peak amplitude fluctuation index is greater than or equal to the peak amplitude fluctuation index reference threshold, it means that the amplitude fluctuation of the time domain signal is larger and the distortion of the time domain signal waveform is more obvious, and an amplitude fluctuation abnormal signal is generated at this time; if the peak amplitude fluctuation index is less than the peak amplitude fluctuation index reference threshold, it means that the amplitude fluctuation of the time domain signal is smaller and the distortion of the time domain signal waveform is less obvious, and a amplitude fluctuation normal signal is generated at this time.

[0094] The higher the peak amplitude fluctuation index, the more serious the distortion of the time domain signal waveform. A high fluctuation index indicates that the amplitude of the signal changes frequently and drastically in a short period of time, reflecting the poor stability of the signal. This phenomenon is usually caused by significant inhomogeneity or defects in the material, such as cracks, holes or impurities.

[0095] Severe signal distortion means that the signal shape deviates greatly from the ideal or expected shape and contains more high-frequency components. This can cause the detection system to misjudge material properties, affecting the accuracy and reliability of detection. By analyzing the peak amplitude fluctuation index, the degree of signal distortion can be identified and quantified, thereby evaluating structural changes and potential defects within the material.

[0096] Therefore, monitoring and analyzing the peak amplitude fluctuation index is crucial for evaluating internal defects in materials. A high fluctuation index indicates that there may be complex internal structural changes or defects, which helps to timely detect potential problems in materials, improve detection efficiency, and ensure product quality and safety.

[0097] S4: After comprehensively analyzing the stability of the scattering point density distribution and the degree of distortion of the time domain signal waveform, the quality of the appearance image of the casting product captured by the imaging system is evaluated.

[0098] The scattering point density distribution anomaly index and the peak amplitude fluctuation index are normalized, and the quality assessment coefficient of the casting product appearance image captured by the imaging system is calculated based on the normalized scattering point density distribution anomaly index and the peak amplitude fluctuation index.

[0099] For example, the present invention can use the following formula to calculate the quality assessment coefficient of the casting product appearance image captured by the imaging system, and the calculation expression is: ; In the formula, AQ is the quality assessment coefficient of the casting product appearance image captured by the imaging system, AS is the scattering point density distribution anomaly index, EK is the peak amplitude fluctuation index, are the proportional coefficients of the scattering point density distribution anomaly index and the peak amplitude fluctuation index, respectively, and .

[0100] S5: The quality of the casting product appearance images captured by the imaging system is divided into high-quality images, medium-quality images and low-quality images. For high-quality images, defect detection and analysis are directly performed, and for low-quality images, defect detection and analysis are performed.

[0101] Compare the quality assessment coefficient of the casting product appearance image captured by the imaging system with the gradient standard threshold, the gradient standard threshold includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold, and compare the quality assessment coefficient of the casting product appearance image captured by the imaging system with the first standard threshold and the second standard threshold respectively;

[0102] If the quality assessment coefficient of the casting product appearance image captured by the imaging system is greater than the second standard threshold, the corresponding casting product appearance image is classified as a high-quality image, and defect detection analysis is directly performed on the high-quality image;

[0103] If the quality assessment coefficient of the casting product appearance image captured by the imaging system is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, the corresponding casting product appearance image is classified as a medium quality image, and the medium quality image is further analyzed;

[0104] If the quality assessment coefficient of the casting product appearance image captured by the imaging system is less than the first standard threshold, the corresponding casting product appearance image is classified as a low-quality image, and the low-quality image is excluded.

[0105] S6: For medium-quality images, analyze the accuracy of the imaging system in capturing the appearance image of the casting product within a fixed time period. Based on the analysis results, dynamically adjust the emission power and scanning speed of the terahertz imaging device to improve the imaging quality.

[0106] When the appearance image of the casting product captured by the imaging system is classified as a medium quality image, that is, the quality assessment coefficient generated within a fixed time period is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, the quality assessment coefficients generated within a subsequent fixed time period that are greater than or equal to the first standard threshold and less than or equal to the second standard threshold are collected, and a corresponding data set is established, and the mean and standard deviation of the data set are calculated, and the mean and standard deviation of the quality assessment coefficients in the data set are further analyzed to determine the accuracy of the appearance image of the casting product captured by the imaging system within the subsequent fixed time period, divide the accuracy of the appearance image of the casting product into different categories, divide the accuracy of the appearance image of the casting product into accurate images and inaccurate images, and dynamically adjust the emission power and scanning speed of the terahertz imaging device for the inaccurate images to improve the imaging quality.

[0107] If the mean value of the quality assessment coefficient in the data set is greater than or equal to the reference threshold value of the mean value of the quality assessment coefficient, and the standard deviation of the quality assessment coefficient is less than the reference threshold value of the standard deviation of the quality assessment coefficient, it means that the image quality is stable and good, and it is marked as an accuracy image;

[0108] If the mean value of the quality assessment coefficient is greater than or equal to the reference threshold value of the mean value of the quality assessment coefficient, and the standard deviation of the quality assessment coefficient is greater than or equal to the reference threshold value of the standard deviation of the quality assessment coefficient, it means that the image quality is good but unstable, and it is marked as an inaccurate image;

[0109] If the mean value of the quality assessment coefficient is less than the reference threshold value of the mean value of the quality assessment coefficient, and the standard deviation of the quality assessment coefficient is greater than or equal to the reference threshold value of the standard deviation of the quality assessment coefficient, it means that the image quality is poor and unstable, and it is marked as an inaccurate image;

[0110] If the mean of the quality assessment coefficient is less than the reference threshold of the mean of the quality assessment coefficient, and the standard deviation of the quality assessment coefficient is less than the reference threshold of the standard deviation of the quality assessment coefficient, it means that the image quality is poor but stable, and it is marked as an inaccurate image.

[0111] For inaccurate images, the imaging quality can be improved by adjusting the emission power and scanning speed of the terahertz imaging device.

[0112] If the signal noise is large, the transmit power can be appropriately reduced to reduce the impact of high-energy noise. If the image contrast is low, the transmit power can be appropriately increased to enhance the signal strength and improve the image contrast.

[0113] If the image resolution is low, the scanning speed can be appropriately reduced to improve the spatial resolution and capture more details. If there is motion blur, the scanning speed can be appropriately increased to reduce the scanning time and reduce image blur.

[0114] Monitor the adjusted image quality in real time. If the adjustment effect is not obvious, further fine-tune the transmission power and scanning speed until the image quality reaches satisfactory standards.

[0115] In this embodiment, the casting materials are classified according to the characteristic information and subjected to real-time terahertz wave detection, which can accurately identify materials with abnormal absorption rates, and deeply analyze their scattering images and time domain signals through terahertz waves and time domain spectrum scanning, evaluate the internal defects and quality stability of the materials, and then divide and process the images captured by the imaging system, directly perform defect detection on high-quality images, exclude low-quality images, and dynamically adjust the transmission power and scanning speed of the terahertz imaging device to optimize medium-quality images, thereby improving the accuracy, efficiency and reliability of the overall detection and ensuring the high quality and safety of the casting products. Example 2

[0116] See also Figure 2 As shown, the casting product appearance defect detection system based on machine vision described in this embodiment includes a casting material analysis module, a scattering point density analysis module, a time domain signal waveform evaluation module, a comprehensive analysis module, an image quality classification module and an imaging quality adjustment module;

[0117] Casting material analysis module: According to the characteristic information of each casting material on the production line, it is divided into different material categories, and terahertz wave detection is performed on different categories of casting materials to obtain their absorption rate in real time. The real-time absorption rate of different casting materials is analyzed to identify casting materials with abnormal absorption rate;

[0118] Scattering point density analysis module: performs terahertz wave scanning on casting materials with abnormal absorption rate, obtains scattering images, determines abnormal changes in the scattering point density distribution area in the scattering image, and evaluates the stability of the scattering point density distribution;

[0119] Time domain signal waveform evaluation module: Scan the casting material with abnormal absorption rate through the terahertz time domain spectrometer, obtain the time domain signal when the terahertz wave passes through the casting material, analyze the amplitude fluctuation state of the time domain signal, and evaluate the distortion degree of the time domain signal waveform;

[0120] Comprehensive analysis module: After comprehensively analyzing the stability of the scattering point density distribution and the degree of distortion of the time domain signal waveform, the quality of the appearance image of the casting product captured by the imaging system is evaluated;

[0121] Image quality classification module: the quality of the casting product appearance images captured by the imaging system is classified into high-quality images, medium-quality images and low-quality images. For high-quality images, defect detection and analysis are directly performed, and for low-quality images, defects are excluded.

[0122] Imaging quality adjustment module: For medium-quality images, the accuracy of the imaging system in capturing the appearance image of the casting product within a fixed time period is analyzed. Based on the analysis results, the emission power and scanning speed of the terahertz imaging device are dynamically adjusted to improve the imaging quality.

[0123] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0124] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0125] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0126] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0127] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0128] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for detecting appearance defects of casting products based on machine vision, characterized in that: The following steps are involved: S1: According to the characteristic information of each casting material on the production line, it is divided into different material categories, and terahertz wave detection is performed on different categories of casting materials to obtain their absorption rate in real time. The real-time absorption rate of different casting materials is analyzed to identify casting materials with abnormal absorption rate; S2: Perform terahertz wave scanning on casting materials with abnormal absorption rate, obtain scattering images, determine the abnormal changes in the scattering point density distribution area in the scattering image, and evaluate the stability of the scattering point density distribution; The step S2 specifically includes: generating a scattering point density distribution anomaly index according to the abnormal changes in the scattering point density distribution area in the scattering map, and evaluating the stability of the scattering point density distribution. The method for obtaining the scattering point density distribution anomaly index is: Get the scatter point density distribution data set, the data set contains N samples, each sample has d features, ;in, Represents the eigenvector of the i-th sample, standardizes the data, and calculates the covariance matrix Σ of the standardized data set. ; In the formula, EX is the mean vector of the standardized data, exi is the i-th sample after standardization, and T is the matrix transpose; perform eigenvalue decomposition on the covariance matrix Σ to obtain the eigenvalues ​​λ1,λ2,…,λd and the corresponding eigenvectors v1,v2,…,vd; , d is a positive integer greater than 0; select the matrix V composed of the first k eigenvectors as the principal component matrix V=[v1,v2,…,vk] according to the size of the eigenvalue; project the standardized data X′ into the principal component space to obtain a low-dimensional representation Z, ; Where Z is an N×k matrix, representing the low-dimensional data projected into the principal component space, and the reconstruction error of each sample is calculated, that is, the difference between the original data and the reconstructed data. ;in, is the reconstructed data, and the scattering point density distribution anomaly index is calculated. The square sum of the reconstruction error is used as the scattering point density distribution anomaly index. The specific calculation expression is: ; Where AS is the anomaly index of scattering point density distribution; S3: Scan the casting material with abnormal absorption rate through the terahertz time-domain spectrometer to obtain the time domain signal when the terahertz wave passes through the casting material, analyze the amplitude fluctuation state of the time domain signal, and evaluate the distortion degree of the time domain signal waveform; The step S3 specifically includes: extracting characteristic information from the processed time domain signal, analyzing the amplitude fluctuation state of the time domain signal, generating a peak amplitude fluctuation index according to the amplitude change of the terahertz wave, and evaluating the degree of distortion of the time domain signal waveform. The method for obtaining the peak amplitude fluctuation index is: The time domain signal m(t) is decomposed into several intrinsic mode functions IMF and a residual term. The decomposition expression is: ; In the formula, is the kth eigenmode function, r(t) is the residual term; Perform Hilbert transform on each IMF, calculate the Hilbert transform of each IMF, and obtain its instantaneous amplitude and instantaneous phase. is the Hilbert transform of the kth IMF, is the circumference of a circle, s is the adjacent moment of time t; calculate the instantaneous amplitude of the kth IMF and instantaneous phase , the specific calculation expression is: ; In the formula, is the Hilbert transform of the kth IMF, is the instantaneous amplitude of the kth IMF; ; In the formula, is the instantaneous phase of the kth IMF; Calculate the peak amplitude of each IMF ; Calculate the standard deviation of the peak amplitude of each IMF, and calculate the peak amplitude fluctuation index after summing up the standard deviations of the peak amplitudes of all IMFs; S4: After comprehensively analyzing the stability of the scattering point density distribution and the distortion degree of the time domain signal waveform, the quality of the casting product appearance image captured by the imaging system is evaluated; S5: Divide the quality of the casting product appearance images captured by the imaging system into high-quality images, medium-quality images and low-quality images. For high-quality images, directly perform defect detection and analysis, and for low-quality images, exclude them; S6: For medium-quality images, analyze the accuracy of the imaging system in capturing the appearance image of the casting product within a fixed time period. Based on the analysis results, dynamically adjust the emission power and scanning speed of the terahertz imaging device to improve the imaging quality.

2. The method for detecting appearance defects of casting products based on machine vision according to claim 1, characterized in that: The random forest algorithm is used to analyze the real-time terahertz wave absorption rate data to identify casting materials with abnormal absorption rates, specifically: Obtain the absorption rate data of casting materials in real time from the terahertz wave detection equipment; Using historical data, normal samples are marked as 0 and abnormal samples are marked as 1 to form a labeled data set; Divide the dataset into training and testing sets; Use the training set to train the random forest model and use the test set to evaluate the model performance; Input the terahertz wave absorption rate data acquired in real time into the trained random forest model; Extract features from real-time data and input the extracted feature data into the random forest model for classification and judgment; If the model output is 1, the absorption rate of the casting material is abnormal, and it is marked as a casting material with abnormal absorption rate; if the model output is 0, the absorption rate of the casting material is normal, and it is marked as a casting material with normal absorption rate.

3. The method for detecting appearance defects of casting products based on machine vision according to claim 1, characterized in that: The obtained scattering point density distribution anomaly index is compared with the preset scattering point density distribution anomaly index reference threshold. If the scattering point density distribution anomaly index is greater than or equal to the scattering point density distribution anomaly index reference threshold, the more obvious the abnormal change of the scattering point density distribution area in the scattering map, the lower the stability of the scattering point density distribution, and a scattering point density distribution anomaly signal is generated at this time; If the scattering point density distribution anomaly index is less than the scattering point density distribution anomaly index reference threshold, the less obvious the abnormal change of the scattering point density distribution area in the scattering map, the higher the stability of the scattering point density distribution, and a normal scattering point density distribution signal is generated.

4. The method for detecting appearance defects of casting products based on machine vision according to claim 1, characterized in that: The obtained peak amplitude fluctuation index is compared with a preset peak amplitude fluctuation index reference threshold. If the peak amplitude fluctuation index is greater than or equal to the peak amplitude fluctuation index reference threshold, the more obvious the distortion of the time domain signal waveform is, the more an amplitude fluctuation abnormal signal is generated. If the peak amplitude fluctuation index is less than the peak amplitude fluctuation index reference threshold, the distortion degree of the time domain signal waveform is less obvious, and a normal amplitude fluctuation signal is generated at this time.

5. The method for detecting appearance defects of casting products based on machine vision according to claim 1, characterized in that: The scattering point density distribution anomaly index and the peak amplitude fluctuation index are normalized, and the quality assessment coefficient of the casting product appearance image captured by the imaging system is calculated by the normalized scattering point density distribution anomaly index and the peak amplitude fluctuation index; Compare the quality assessment coefficient of the casting product appearance image captured by the imaging system with the gradient standard threshold, the gradient standard threshold includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold, and compare the quality assessment coefficient of the casting product appearance image captured by the imaging system with the first standard threshold and the second standard threshold respectively; If the quality assessment coefficient of the casting product appearance image captured by the imaging system is greater than the second standard threshold, the corresponding casting product appearance image is classified as a high-quality image, and defect detection analysis is directly performed on the high-quality image; if the quality assessment coefficient of the casting product appearance image captured by the imaging system is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, the corresponding casting product appearance image is classified as a medium-quality image, and further analysis is performed on the medium-quality image; if the quality assessment coefficient of the casting product appearance image captured by the imaging system is less than the first standard threshold, the corresponding casting product appearance image is classified as a low-quality image, and the low-quality image is excluded.

6. The method for detecting appearance defects of casting products based on machine vision according to claim 5, characterized in that: When the appearance image of the casting product captured by the imaging system is classified as a medium quality image, that is, the quality assessment coefficient generated within a fixed time period is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, the quality assessment coefficients generated within a subsequent fixed time period that are greater than or equal to the first standard threshold and less than or equal to the second standard threshold are collected, and a corresponding data set is established, and the mean and standard deviation of the data set are calculated, and the mean and standard deviation of the quality assessment coefficients in the data set are further analyzed to determine the accuracy of the appearance image of the casting product captured by the imaging system within the subsequent fixed time period, divide the accuracy of the appearance image of the casting product into different categories, divide the accuracy of the appearance image of the casting product into accurate images and inaccurate images, and dynamically adjust the emission power and scanning speed of the terahertz imaging device for the inaccurate images to improve the imaging quality.

7. The method for detecting appearance defects of casting products based on machine vision according to claim 6, characterized in that: If the mean of the quality assessment coefficients in the data set is greater than or equal to the reference threshold of the mean of the quality assessment coefficients, and the standard deviation of the quality assessment coefficients is less than the reference threshold of the standard deviation of the quality assessment coefficients, it is marked as an accurate image; if the mean of the quality assessment coefficients is greater than or equal to the reference threshold of the mean of the quality assessment coefficients, and the standard deviation of the quality assessment coefficients is greater than or equal to the reference threshold of the standard deviation of the quality assessment coefficients, it is marked as an inaccurate image; if the mean of the quality assessment coefficients is less than the reference threshold of the mean of the quality assessment coefficients, and the standard deviation of the quality assessment coefficients is greater than or equal to the reference threshold of the standard deviation of the quality assessment coefficients, it is marked as an inaccurate image; if the mean of the quality assessment coefficients is less than the reference threshold of the mean of the quality assessment coefficients, and the standard deviation of the quality assessment coefficients is less than the reference threshold of the standard deviation of the quality assessment coefficients, it is marked as an inaccurate image.

8. A casting product appearance defect detection system based on machine vision, used to implement the casting product appearance defect detection method based on machine vision according to any one of claims 1 to 7, characterized in that: It includes casting material analysis module, scattering point density analysis module, time domain signal waveform evaluation module, comprehensive analysis module, image quality division module and imaging quality adjustment module; Casting material analysis module: According to the characteristic information of each casting material on the production line, it is divided into different material categories, and terahertz wave detection is performed on different categories of casting materials to obtain their absorption rate in real time. The real-time absorption rate of different casting materials is analyzed to identify casting materials with abnormal absorption rate; Scattering point density analysis module: performs terahertz wave scanning on casting materials with abnormal absorption rate, obtains scattering images, determines abnormal changes in the scattering point density distribution area in the scattering image, and evaluates the stability of the scattering point density distribution; Time domain signal waveform evaluation module: Scan the casting material with abnormal absorption rate through the terahertz time domain spectrometer, obtain the time domain signal when the terahertz wave passes through the casting material, analyze the amplitude fluctuation state of the time domain signal, and evaluate the distortion degree of the time domain signal waveform; Comprehensive analysis module: After comprehensively analyzing the stability of the scattering point density distribution and the degree of distortion of the time domain signal waveform, the quality of the appearance image of the casting product captured by the imaging system is evaluated; Image quality classification module: the quality of the casting product appearance images captured by the imaging system is classified into high-quality images, medium-quality images and low-quality images. For high-quality images, defect detection and analysis are directly performed, and for low-quality images, defects are excluded. Imaging quality adjustment module: For medium-quality images, the accuracy of the imaging system in capturing the appearance image of the casting product within a fixed time period is analyzed. Based on the analysis results, the emission power and scanning speed of the terahertz imaging device are dynamically adjusted to improve the imaging quality.

Images