Fusion imaging PCB via hole blockage detection method and system

Abstract

The invention relates to the technical field of PCB (Printed Circuit Board) detection methods, in particular to a PCB via hole blockage detection method based on fusion imaging, which comprises the following steps: generating a hole position mask matrix according to PCB design information, and synchronously acquiring a transient thermal characteristic sequence and a local dielectric response diagram of a via hole; extracting an independent physical state signature for each via hole from the transient thermal characteristic sequence and the local dielectric response diagram according to the hole position mask matrix; constructing a local context decoding area for each via hole, and calculating a local consistency anomaly score of each via hole; performing quality evaluation on the thermal signature and the electromagnetic signature of each via hole to generate a comprehensive abnormal confidence coefficient; according to the method, the positions of the via holes are accurately locked, an independent physical state signature is established for each via hole, weak anomalies of the single blocked via holes are converted into significant differences in a local range, and the detection accuracy of the single blocked via holes in a dense area is improved.

Description

Technical Field

[0001] This invention relates to the field of PCB inspection methods, and in particular to a PCB via blockage detection method and system using fusion imaging. Background Technology

[0002] Via blockage is a common defect in PCB manufacturing, severely impacting the electrical connectivity and reliability of the circuit board. Existing automated optical inspection methods primarily rely on visible light or single-modal imaging techniques to identify defects by analyzing the texture, contrast, or edge features of the image.

[0003] When faced with high-density, regularly arranged via arrays, the characteristic signals exhibited by a single blocked via in the optical image are often very weak, easily overwhelmed by the strong background features formed by the numerous surrounding normal vias, leading to serious missed detection problems. Furthermore, traditional image fusion techniques, such as weighted averaging and Laplacian pyramid fusion, while improving the overall image quality to some extent when dealing with such dense, regular structures, are essentially linear, global smoothing operations. These operations are prone to feature smoothing effects, systematically suppressing sparsely distributed anomalous signals and exacerbating the overwhelming effect. Therefore, a detection method capable of more accurately identifying individual blocked vias in dense regions is needed. Summary of the Invention

[0004] The main objective of this invention is to provide a PCB board via blockage detection method and system using fusion imaging, aiming to solve the technical problems mentioned in the background art.

[0005] This invention proposes a PCB board via blockage detection method using fusion imaging, comprising: The theoretical center coordinates and theoretical radius of all vias are obtained based on the PCB design information, and a via mask matrix is ​​generated based on the theoretical center coordinates and theoretical radius. The transient thermal characteristic sequence and local dielectric response map of the via are synchronously acquired based on the area defined by the via mask matrix; Based on the aperture mask matrix, an independent physical state signature is extracted from each via from the transient thermal feature sequence and the local dielectric response map, wherein the physical state signature includes at least a thermal signature and an electromagnetic signature; A local context decoding region is constructed for each via, and a local consistency anomaly score for each via is calculated based on the difference between the physical state signature of each via and the physical state signature of its neighboring vias within the local context decoding region. The thermal and electromagnetic signatures of each via are evaluated for quality, and a comprehensive anomaly confidence level is dynamically generated based on the results of the quality evaluation. The via is determined to be blocked based on the local consistency anomaly score and the overall anomaly confidence score.

[0006] Preferably, the step of decoupling each via from the transient thermal feature sequence and the local dielectric response map based on the via mask matrix and extracting an independent physical state signature includes: Based on the aperture mask matrix, extract the temperature decay curve of the average temperature of each via over time from the transient thermal feature sequence; For each via, the temperature decay curve is fitted with an exponential decay rate starting from the temperature peak point to obtain the decay time constant. Calculate the thermal signature of each via based on the decay time constant; The mean amplitude and mean phase of each via are extracted from the dielectric response map based on the via mask matrix. The reference amplitude and reference phase of the via-free background region are obtained based on the dielectric response diagram. The composite scattering intensity of each via is calculated based on the mean amplitude, mean phase, reference amplitude, and reference phase. The composite scattering intensity of all vias is normalized to obtain the electromagnetic signature of each via. The physical state signatures of all vias are obtained based on the electromagnetic signature and the thermal signature.

[0007] Preferably, the step of obtaining the physical state signatures of all vias based on the electromagnetic signature and the thermal signature includes: Isolated vias that are all more than a preset safety distance from all adjacent vias are selected from the via mask matrix, and a reference sample set is established based on the electromagnetic signature and thermal signature of the isolated vias, wherein the reference sample set includes a reference thermal signature set and a reference electromagnetic signature set. The distribution centers of the reference thermal signature set and the reference electromagnetic signature set are calculated respectively, and the ratio between the two distribution centers is calculated as the cross-field alignment coefficient. The thermal signature and electromagnetic signature of all vias are scaled by the cross-field alignment coefficient, and the numerical ranges of the thermal signature and the electromagnetic signature are aligned with the two distribution centers respectively. The aligned thermal signature and the electromagnetic signature are combined to form a two-dimensional vector, which serves as the physical state signature for all vias.

[0008] Preferably, the step of constructing a local context decoding region for each via and calculating the local consistency anomaly score for each via based on the difference between the physical state signature of each via and the physical state signatures of its neighboring vias within the corresponding local context decoding region includes: The typical hole spacing between each via and all other vias is obtained based on the hole mask matrix. Centered on each via, a preset number of nearest-neighbor vias are selected according to the typical via spacing to form several local context decoding regions. The physical state signatures of all nearest-neighbor vias in the local context decoding region are combined into a decoding region physical state signature set. Calculate the mean vector and covariance matrix of the physical state signature set of the decoding area for each via; Calculate the Mahalanobis distance between the physical state signature of each via and the mean vector and covariance matrix of the corresponding physical state signature set of the decoding area; The Mahalanobis distance is mapped to a probability value through an exponential function, which is then used as the local consistency anomaly score.

[0009] Preferably, the step of performing a quality assessment on the thermal and electromagnetic signatures of each via, and dynamically fusing them to generate a comprehensive anomaly confidence level based on the results of the quality assessment, includes: Calculate the determination coefficient of the exponential decay fit, and obtain the thermal characteristic quality score of each via based on the determination coefficient; Obtain the amplitude standard deviation and amplitude mean of each via in the local dielectric response diagram, obtain the coefficient of variation based on the amplitude standard deviation and amplitude mean, and obtain the electromagnetic characteristic quality score of each via based on the coefficient of variation. Calculate the dynamic weight of thermal characteristics and the dynamic weight of electromagnetic characteristics for each via based on the quality score; By performing global Z-score normalization on the thermal signatures and electromagnetic signatures of all vias respectively, the thermal anomaly amplitude and electromagnetic anomaly amplitude are obtained. The thermal anomaly amplitude and the electromagnetic anomaly amplitude are fused based on the dynamic weights of the thermal features and the electromagnetic features to obtain a comprehensive anomaly confidence level.

[0010] Preferably, the step of determining whether the via is blocked based on the local consistency anomaly score and the overall anomaly confidence level includes: A first initial global threshold is obtained by statistically analyzing the mean and standard deviation of the local consistency anomaly scores of all vias, and a second initial global threshold is obtained by statistically analyzing the mean and standard deviation of the comprehensive anomaly confidence scores of all vias. The process tolerance coefficient is obtained according to the type and process specifications of the PCB board. The first initial global threshold and the second initial global threshold are adjusted according to the process tolerance coefficient to obtain the first adaptive threshold and the second adaptive threshold. The local consistency anomaly scores and the overall anomaly confidence scores of all vias are compared with the first adaptive threshold and the second adaptive threshold, respectively. If the local consistency anomaly score is less than the first adaptive threshold, it is determined to be blocked; If only the overall anomaly confidence level is greater than the second adaptive threshold, the corresponding via is marked as a suspected blockage, and a secondary verification is performed based on the deviation direction of the thermal signature and the electromagnetic signature relative to their respective global mean. If the relative deviation direction of the thermal signature and the electromagnetic signature is consistent and both are lower than their respective global mean, then it is finally determined to be a blockage; otherwise, it is determined to be normal.

[0011] The present invention also provides a PCB board via blockage detection system based on fusion imaging, comprising: The hole modeling module is used to obtain the theoretical center coordinates and theoretical radius of all vias based on the PCB board design information, and generate a hole mask matrix based on the theoretical center coordinates and theoretical radius. The synchronous imaging module is used to synchronously acquire the transient thermal characteristic sequence and local dielectric response map of the via according to the area defined by the aperture mask matrix; The signature extraction module is used to decouple and extract an independent physical state signature for each via from the transient thermal feature sequence and the local dielectric response map based on the via mask matrix, wherein the physical state signature includes at least a thermal signature and an electromagnetic signature; The local scoring module is used to construct a local context decoding region for each via and calculate the local consistency anomaly score for each via based on the difference between the physical state signature of each via and the physical state signature of its neighboring vias in the corresponding local context decoding region. The dynamic fusion module is used to perform quality assessment on the thermal signature and electromagnetic signature of each via, and dynamically fuse them to generate a comprehensive anomaly confidence level based on the results of the quality assessment. The comprehensive judgment module is used to determine whether the via is blocked based on the local consistency anomaly score and the comprehensive anomaly confidence level.

[0012] The local scoring module includes: A hole spacing acquisition unit is used to acquire the typical hole spacing between each via and all other vias based on the hole position mask matrix. The decoding area construction unit is used to select a preset number of nearest-neighbor vias to form several local context decoding areas, with each via as the center and according to the typical via spacing, and to combine the physical state signatures of all nearest-neighbor vias in the local context decoding area into a decoding area physical state signature set. The matrix calculation unit is used to calculate the mean vector and covariance matrix of the physical state signature set of the decoding area for each via; The Mahalanobis distance calculation unit is used to calculate the Mahalanobis distance between the physical state signature of each via and the mean vector and covariance matrix of the physical state signature set of the corresponding decoding area. Anomaly mapping unit is used to map the Mahalanobis distance into a probability value through an exponential function, which serves as the local consistency anomaly score.

[0013] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a PCB board via blockage detection method using fusion imaging.

[0014] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a PCB board via blockage detection method using fusion imaging.

[0015] The beneficial effects of this invention are as follows: This invention generates a via mask matrix based on PCB design information to accurately locate vias, eliminates noise interference from non-via areas from the detection starting point, and avoids feature aliasing caused by spatial positioning ambiguity in dense vias. By synchronously acquiring the transient thermal feature sequence and local dielectric response map of the vias, data is obtained from both thermal conduction and electromagnetic scattering dimensions. Compared with single-mode data, this provides richer anomaly information, reduces anomaly omissions caused by weak single-dimensional signals, and further combats signal suppression caused by background masking. Furthermore, an independent physical state signature is established for each via, focusing the detection unit from the overall image to a single via, breaking the limitation of feature superposition of dense vias, allowing the weak features of a single blocked via to be presented independently, solving the problem of features being submerged by the background. In addition, a local context decoding area is constructed for each via, and the local consistency anomaly score is calculated by analyzing its signature difference with neighboring vias. By utilizing the local consistency of normal vias in a high-density array, the weak anomaly of a single blocked via is transformed into a significant difference within a local range, avoiding the dilution of the abnormal signal by the overall background in global analysis and improving the sensitivity of weak anomaly identification. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of a method flow according to an embodiment of this application.

[0017] Figure 2 This is a schematic diagram of the system structure according to an embodiment of this application.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] like Figure 1 As shown, this application provides a PCB board via blockage detection method using fusion imaging, comprising: S1. Obtain the theoretical center coordinates and theoretical radius of all vias based on the PCB design information, and generate a via mask matrix based on the theoretical center coordinates and theoretical radius. By parsing the design file information, extract the theoretical center coordinates and theoretical radius of all vias, and create a two-dimensional blank matrix (initial values ​​are all 0) with the same spatial resolution as the subsequently acquired images. For each theoretical via, with its theoretical center coordinates as the center and a fixed value slightly larger than its theoretical radius (e.g., the theoretical radius is increased by 5 pixels) as the radius, draw a circular area with a fill value of 1 on this blank matrix. Thus, a binary via mask matrix is ​​output. S2, synchronously acquire the transient thermal characteristic sequence and local dielectric response map of the via based on the area defined by the via mask matrix; For acquiring transient thermal feature sequences: A uniform planar heat source (such as an infrared flash lamp) is used to apply a brief (e.g., millisecond-level) pulsed thermal excitation to the target area of ​​the PCB board. Simultaneously, a mid-wave or long-wave infrared thermal imager is controlled to record the temperature changes of the target area at a high frame rate (e.g., 100Hz), recording the entire process from the start of the thermal excitation until the area temperature essentially returns to the ambient temperature, lasting several seconds. The output is a three-dimensional data cube I(x,y,t), where (x,y) are the spatial pixel coordinates and t is the time frame number, completely recording the temperature value of each pixel over time. For the acquisition of local dielectric response maps: A microwave near-field probe (e.g., a probe station based on a vector network analyzer) is used to scan over the same target area. The probe emits a fixed high-frequency signal (e.g., 10 GHz) and measures the amplitude and phase of the reflected or transmitted signal at each scan point to output two two-dimensional images: a local dielectric response amplitude map and a local dielectric response phase map. These two maps are spatially precisely registered with the thermal image in the transient thermal feature sequence and the aperture mask matrix. S3, based on the aperture mask matrix, decouple each via from the transient thermal feature sequence and the local dielectric response map and extract an independent physical state signature, wherein the physical state signature includes at least a thermal signature and an electromagnetic signature; S4, construct a local context decoding region for each via, and calculate the local consistency anomaly score for each via based on the difference between the physical state signature of each via and the physical state signature of its neighboring vias in the corresponding local context decoding region. S5, perform quality assessment on the thermal signature and electromagnetic signature of each via, and dynamically fuse the results of the quality assessment to generate a comprehensive anomaly confidence level; S6. Determine whether the via is blocked based on the local consistency anomaly score and the comprehensive anomaly confidence level.

[0021] As described in steps S1-S6 above, the main physical change caused by via blockage is the replacement of the metal conductor inside the via with an insulating medium. This change leads to two differences in physical properties: first, a decrease in thermal conductivity, with normally metallized vias conducting heat quickly, while vias blocked by the medium conduct heat slowly; second, a change in electromagnetic scattering characteristics, with normal vias, as perpendicular metal conductors, strongly disturbing the surrounding electromagnetic field, while blocked vias have almost no effect on the electromagnetic field. Existing technologies mainly identify defects by analyzing the texture, contrast, or edge features of images. When faced with high-density, regularly arranged via arrays, the characteristic signals exhibited by a single blocked via in the optical image are often very weak and easily submerged by the strong background features formed by a large number of normal vias around it, leading to missed detections. Furthermore, traditional image fusion techniques, such as weighted averaging and Laplacian pyramid fusion, while improving the overall image quality to some extent when processing such dense and regular structures, are essentially linear, global smoothing operations that easily produce feature smoothing effects. They systematically suppress sparsely distributed abnormal signals, thus exacerbating the submersion effect. Therefore, a novel detection method is needed that can overcome the limitations of existing imaging and fusion modes and accurately identify individual blocked vias in dense areas, so as to capture weak anomalous signals under strong background interference of dense arrays.

[0022] This invention employs a physical field collaborative sensing strategy to overcome the limitations of traditional optical imaging. First, it extracts the theoretical positions and geometric parameters of all vias from the PCB design information, generating a via mask matrix. This matrix is ​​a two-dimensional digital array where each element is 1 at the theoretical center of the via and 0 in other areas. This provides precise spatial positioning for subsequent processing, allowing the system to more accurately focus on individual vias rather than the entire image, thus preventing features from being obscured by the background. Based on this, transient thermal feature sequences and local dielectric response maps are acquired simultaneously. A high-precision timestamp is added to both the transient thermal feature sequences and the local dielectric response maps, and data alignment is performed based on the timestamps to achieve synchronous acquisition. The transient thermal feature sequence is an image sequence of the temperature field in the target area changing over time after applying a transient thermal pulse excitation to the PCB board, recording the heat conduction process in the via area due to differences in the thermophysical properties of the material. The local dielectric response map is an image reflecting the distribution of the electromagnetic properties of the medium, obtained by scanning with a microwave near-field probe, including amplitude and phase maps. These two data sources provide complementary information directly related to the internal state of vias from two different physical principles: thermal conduction and electromagnetic scattering. This synchronous acquisition of multi-physics data avoids the problem of single optical features being easily affected by surface conditions. Since via blockage causes changes in both thermal conduction and electromagnetic scattering, single-physics data cannot fully capture this anomaly. Therefore, this invention uses physical state signatures to replace traditional image features. Guided by the aperture mask matrix, the system decouples an independent physical state signature for each via from the original physical field data. For thermal data, the temperature decay curve of each via region is extracted and exponentially fitted to obtain the decay time constant, which is then used to calculate the thermal signature. This signature quantifies the thermal conductivity of the via unit. Metallized vias have high thermal conductivity, resulting in a large thermal signature value, while dielectric-blocked vias have poor thermal conductivity and a small thermal signature value. For electromagnetic data, the composite scattering intensity is calculated and normalized to obtain the electromagnetic signature, which quantifies the via's scattering ability for near-field microwaves. Metallic vias scatter strongly, resulting in a high electromagnetic signature value, while dielectric-blocked vias scatter weakly, resulting in a low electromagnetic signature value. To further ensure that thermal signatures and electromagnetic signatures can be compared and fused on the same scale, this invention also introduces a cross-field signature alignment operation to construct a unique physical state signature for each via, transforming the detection unit from the image region to the physical hole unit. Compared with the limitations of feature superposition in traditional whole-image processing, it can directly characterize the via state at the physical level, without being affected by surface optical artifacts, and provides a more reliable feature basis for accurately identifying anomalies in dense backgrounds.

[0023] To address the flooding effect problem, this invention designs a local context decoding region. In a dense rule array, a blocked via is essentially an outlier in its local neighborhood. Therefore, for each via, the difference between its physical state signature and the physical state signatures of its neighboring vias within the local context decoding region is calculated, centered on the via. After considering the correlation of features between vias, a local consistency anomaly score is obtained. This transforms anomaly detection from global threshold judgment to local consistency comparison using the rules of the dense via array. Since the flooding effect occurs because the intensity of weak anomalous signals is much lower than the overall intensity of normal background signals globally, the local context decoding region dynamically establishes a small range of normal parameters for each via, containing only its nearest neighbors. The model shows that the fluctuations within the local normal range are much smaller than those within the global range. This allows the weak anomalous signals of individual blocked vias to be effectively highlighted against a strong global normal background. Furthermore, because any via in the dense, regular array of PCB vias shares high similarity with its neighbors in terms of physical structure, material properties, and environment, their physical state signatures should also cluster tightly together in the feature space. The local context decoding region utilizes this inherent spatial autocorrelation. Under normal circumstances, the features of vias should be uniform within a small local area. When a blocked via appears, it disrupts this local uniformity, becoming an outlier in the local feature space. This further allows the features of individual blocked vias to be highlighted, thus solving the problem of the flooding effect. Considering that the reliability of different physical field measurement data will dynamically change in actual testing, this invention performs a quality score on each signature and dynamically fuses them to generate a comprehensive anomaly confidence score. This ensures that each decision automatically tends to adopt more reliable physical field evidence, thereby maintaining high accuracy even in the face of measurement noise and environmental fluctuations. Finally, the via is determined to be blocked based on the local consistency anomaly score and the comprehensive anomaly confidence score. By verifying the consistency of multi-physics evidence, the accuracy of via blockage detection can be improved.

[0024] In one embodiment, the step of decoupling each via from the transient thermal feature sequence and the local dielectric response map based on the via mask matrix and extracting an independent physical state signature includes: S31, extract the temperature decay curve of the average temperature of each via over time from the transient thermal feature sequence based on the via mask matrix; S32, perform exponential decay fitting on the temperature decay curve of each via starting from the temperature peak point to obtain the decay time constant; The exponential decay fitting function is: ; In the formula, T i(t) represents the exponentially decaying fitting function, A represents the fitting magnitude coefficient, and τ i Let C represent the decay time constant of the i-th via, and let C represent the baseline constant. S33, calculate the thermal signature of each via based on the decay time constant, using the following formula: ; In the formula, K eff, i τ represents the thermal signature of the i-th via. i This represents the decay time constant of the i-th via; S34, extract the average amplitude and average phase of each via from the dielectric response map according to the via mask matrix; S35, obtain the reference amplitude and reference phase of the via-free background region based on the dielectric response map, and calculate the composite scattering intensity of each via based on the mean amplitude, mean phase, reference amplitude, and reference phase, using the following formula: ; In the formula, S i ma represents the composite scattering intensity of the i-th via. bac Indicates the reference amplitude, ma i Ph represents the average amplitude of the i-th via. bac Indicates the reference phase, ph i This represents the average phase value of the i-th via; S36, normalize the composite scattering intensity of all vias to obtain the electromagnetic signature of each via; S37, Obtain the physical state signatures of all vias based on the electromagnetic signature and the thermal signature.

[0025] As described in steps S31-S37 above, since the essence of via blockage is the replacement of the metal conductor inside the via with an insulating medium, this change directly leads to two key physical property alterations: First, a decrease in thermal conductivity. The thermal conductivity of metal is much higher than that of insulating media. Normal vias can conduct heat quickly due to their metallized structure, while blocked vias experience a slower cooling rate because the insulating medium hinders heat transfer. Second, a reduction in electromagnetic scattering capability. Metal conductors have a strong reflection and scattering effect on high-frequency electromagnetic fields. Normal vias significantly alter the amplitude and phase distribution of the surrounding electromagnetic field, while blocked vias, lacking a metal conductor, have a significantly reduced disturbance effect on the electromagnetic field. Therefore, this invention captures the physical property changes caused by via blockage by converting transient thermal characteristic sequences and local dielectric response diagrams into thermal and electromagnetic signatures that accurately characterize the physical state of vias.

[0026] In terms of thermal signature extraction, the average temperature of all pixels in the local area of ​​each via is precisely extracted from the transient thermal feature sequence based on the via mask matrix, forming the temperature decay curve of the via. This curve characterizes the unique heat dissipation process of the via region due to its specific material composition and geometry after the pulsed thermal excitation stops. Then, the temperature decay curve of each via is fitted with exponential decay starting from the temperature peak point. Its mathematical expression includes three parameters: amplitude coefficient, decay time constant, and baseline constant. The decay time constant quantifies how quickly the temperature of the via region recovers to a steady state, and its value is directly determined by the effective thermal diffusivity of the region. Based on this, the thermal signature of each via is obtained by calculating the reciprocal of the decay time constant. Its specific form is a scalar value characterizing the thermal conductivity of the material, representing the effective thermal conductivity of the via unit in the axial direction. Metallized vias provide an efficient heat conduction path from one side of the plate to the other due to the high thermal conductivity of the metal. Therefore, they dissipate heat very quickly after transient thermal excitation, resulting in a large effective thermal conductivity coefficient. Conversely, vias blocked by non-metallic media have a drastic decrease in thermal conductivity, resulting in slow heat dissipation and a sharp decrease in the effective thermal conductivity coefficient. This quantitative characterization based on the physical principle of thermal conduction makes thermal signatures more sensitive to blockage defects and resistant to optical surface interference.

[0027] In terms of electromagnetic signature extraction, the mean amplitude and mean phase of each via are extracted from the amplitude and phase diagrams of the local dielectric response map based on the via mask matrix. The mean amplitude and mean phase reflect the changes in the absorption / scattering intensity of incident microwave energy and the electromagnetic wave propagation speed of that region, respectively. To establish an effective reference, reference amplitude and reference phase are obtained from the dielectric response map for background regions far from any via and representing pure substrate material. Based on these parameters, the composite scattering intensity of each via is obtained by calculating the composite scattering intensity formula, where the amplitude relative attenuation term... The microwave amplitude variation caused by the presence of vias was quantified, reflecting energy loss and phase relative shift term. This quantifies the microwave phase change caused by the presence of vias, reflecting the signal distortion caused by path difference or dielectric constant changes. Combining these two factors, a composite scattering intensity is formed that comprehensively reflects the via's ability to disturb near-field electromagnetic waves. Subsequently, the maximum value of the composite scattering intensity of all vias on the current PCB board is normalized and mapped to the zero-one interval to obtain the electromagnetic signature of each via. The specific form of this electromagnetic signature is a dimensionless scalar value, which represents the via's effective ability to scatter the incident microwave field as a local scatterer compared to the surrounding background. A well-metallized via is similar to a miniature metal cylinder, which will significantly scatter and reflect near-field microwaves, causing a drastic change in the field distribution in the area above it. Therefore, the value of the electromagnetic signature approaches one. On the other hand, a via blocked by a dielectric material has a smaller difference in electromagnetic properties from the surrounding substrate material, and its ability to disturb the field is greatly weakened. Therefore, the value of the electromagnetic signature approaches zero. Through quantitative characterization based on the principle of electromagnetic scattering, the electromagnetic signature can reveal whether the via is blocked from an electromagnetic perspective.

[0028] Compared to traditional methods that rely heavily on visual features such as image grayscale and edges when extracting via features, which are easily affected by environmental factors and have a weak correlation with the physical nature of via blockage, resulting in low feature discrimination, this invention extracts features from two dimensions: thermal conduction and electromagnetic scattering. This not only improves the correlation between features and abnormal states but also effectively overcomes interference from environmental factors. It provides high-quality input features for reliably detecting submerged individual blockage defects in dense via arrays. Furthermore, the features from the two dimensions are complementary; when the measurement of one physical field is disturbed under certain circumstances, the other physical field may still provide reliable evidence.

[0029] In one embodiment, the step of obtaining the physical state signatures of all vias based on the electromagnetic signature and the thermal signature includes: S371, Select isolated vias from the via mask matrix whose distance from all adjacent vias is greater than a preset safety distance, and establish a reference sample set based on the electromagnetic signature and thermal signature of the isolated vias, wherein the reference sample set includes a reference thermal signature set and a reference electromagnetic signature set. S372, calculate the distribution centers of the reference thermal signature set and the reference electromagnetic signature set respectively, and calculate the ratio between the two distribution centers as the cross-field alignment coefficient; S373, the scale transformation is performed on at least one of the thermal signatures and the electromagnetic signatures of all vias by the cross-field alignment coefficient, and the numerical ranges of the thermal signatures and the electromagnetic signatures are aligned with the two distribution centers respectively. S374, The aligned thermal signature and the electromagnetic signature are combined to form a two-dimensional vector, which serves as the physical state signature for all vias.

[0030] As described in steps S371-S374 above, since thermal signatures and electromagnetic signatures originate from different physical fields, their original data dimensions, numerical ranges, and physical meanings differ. Thermal signatures may have values ​​concentrated in the range of 10-50 W / (m•K) due to factors such as the thermal conductivity of the PCB board and the thickness of the via metal plating. Electromagnetic signatures, after normalization, are distributed in the 0-1 range. If existing conventional methods are used for fusion, thermal signatures will dominate in subsequent calculations, and the abnormal information of electromagnetic signatures will be masked, resulting in soft submersion. Therefore, this invention achieves the alignment of the two signatures in the numerical space through a benchmark calibration mechanism based on isolated vias. This allows subsequent local consistency analysis and dynamic fusion decisions to fairly and accurately utilize dual-physical field information, avoiding the dominance of detection results by a single feature due to its numerical magnitude advantage, and further improving the identification accuracy of via blockage in dense areas.

[0031] This invention selects isolated vias from the via mask matrix whose distance from all adjacent vias is greater than a preset safety distance. The basis for selecting these isolated vias as reference samples is that, because they are far from interference from other vias, their thermal and electromagnetic field distributions best reflect the inherent physical characteristics of a single normal via, and are least affected by thermal or electromagnetic coupling from adjacent vias. Therefore, they better represent the ideal normal state. The thermal and electromagnetic signature sets of these isolated vias are defined as the reference thermal signature set and the reference electromagnetic signature set, respectively. Next, the distribution center (e.g., its mean) of the reference thermal signature set and the distribution center of the reference electromagnetic signature set are calculated, and the ratio of the two distribution centers is calculated as the cross-field alignment coefficient. Specifically, this coefficient is a dimensionless scaling factor that quantifies the inherent proportional relationship between thermal and electromagnetic characteristics on a numerical scale under normal conditions. This proportional relationship is jointly determined by the material's thermophysical properties (e.g., thermal conductivity) and electromagnetic properties (e.g., dielectric constant, electrical conductivity). For the same type of PCB material and process, this relationship is relatively stable.

[0032] Then, using the cross-field alignment coefficient, the electromagnetic signatures of all vias on the entire board are globally scaled. Alternatively, under the premise of mathematical equivalence, the thermal signatures can be scaled inversely to map two features originating from different physical principles and having different dimensions and numerical ranges to a common, commensurable numerical reference frame. This allows subsequent local consistency analysis and weighted fusion decisions to compare feature values ​​with more balanced scales. If the numerical scales of two features differ greatly, the feature with the larger numerical range will dominate in multivariate distance calculations or weighted summations, thus overshadowing the role of the other feature. Through alignment, thermal and electromagnetic evidence have more accurate evaluation value. Finally, the aligned thermal and electromagnetic signatures are combined into a two-dimensional vector as the final physical state signature.

[0033] Compared to existing technologies that often directly use the original feature values ​​or perform simple standardization (such as Z-score) when performing multi-feature fusion or analysis, the cross-field signature alignment operation in this embodiment establishes an intrinsic correlation model between thermal and electromagnetic features by utilizing known normal samples (isolated vias) on the board, and performs scale unification accordingly. This avoids decision bias caused by differences in feature dimensions and sensitivity, enabling subsequent steps to make balanced and full use of dual-physics information, thereby improving the ability to identify weak abnormal signals in complex backgrounds.

[0034] In one embodiment, the step of constructing a local context decoding region for each via and calculating a local consistency anomaly score for each via based on the difference between the physical state signature of each via and the physical state signatures of its neighboring vias within the corresponding local context decoding region includes: S41, Obtain the typical hole spacing between each via and all other vias according to the hole mask matrix; S42, taking each via as the center, select a preset number of the nearest vias (this preset number is greater than the minimum effective number of neighboring vias; if the number of neighboring vias is insufficient or their features are too concentrated, the search range can be automatically expanded or additional spatially adjacent vias can be introduced to ensure that the sample size for estimating local statistical characteristics is sufficient) to form several local context decoding regions, and combine the physical state signatures of all the nearest vias in the local context decoding region into a decoding region physical state signature set; S43, calculate the mean vector and covariance matrix of the physical state signature set of the decoding area for each via; S44, calculate the Mahalanobis distance between the physical state signature of each via and the mean vector and covariance matrix of the corresponding physical state signature set of the decoding area, using the following formula: ; In the formula, Di FP represents the Mahalanobis distance of the i-th via. i Let μ represent the physical state signature of the i-th via, and μ represent the mean vector. T This indicates the transpose operation. The inverse matrix of the covariance matrix; S45, the Mahalanobis distance is mapped to a probability value using an exponential function, which is used as the local consistency anomaly score, as follows: ; In the formula, L i D represents the local consistency anomaly score of the i-th via. i 2 This represents the square of the Mahalanobis distance of the i-th via.

[0035] As described in steps S41-S45 above, in a high-density via array, normal vias in the same region have highly consistent design parameters (hole diameter, metal plating thickness) and manufacturing processes. Their physical state signatures (aligned thermal and electromagnetic signatures) should exhibit clustered distribution characteristics. However, blocked vias, due to changes in physical properties, will have signatures that deviate from this cluster, becoming outliers in a local area. Traditional global thresholding methods use the feature distribution of all vias on the entire board as a benchmark, which is easily affected by process fluctuations in different areas of the board (e.g., there may be systematic differences in the thermal signatures of vias in the edge and center regions). This leads to a decrease in the sensitivity of identifying local sparse anomalies. The features of some blocked vias may be in the fuzzy range of the global threshold, neither meeting the global anomaly criteria nor being easily detected due to the strong background coverage of surrounding normal vias. This invention, by constructing a dedicated local context decoding area for each via, transforms anomaly detection from global comparison to local comparison, utilizing the local clustering characteristics of normal vias to highlight sparse anomalies, thereby improving the anomaly recognition capability in dense scenes.

[0036] This invention first calculates the Euclidean distance between each via and all other vias on the board based on the precise spatial coordinate information provided by the via mask matrix. Then, based on the typical via spacing parameters defined in the PCB design file, it dynamically determines the neighborhood range of each via and constructs a local context decoding area. For example, for a via array with a center-to-center spacing of 0.5 mm, the decoding area radius can be set to 1.0 mm, or the six nearest vias can be directly selected as neighbors. This is because, on PCBs with consistent manufacturing processes, adjacent vias have similar processing backgrounds, material properties, and environmental conditions. Therefore, they can collectively form a normal reference group. By establishing such a local reference frame for each via to be detected, the influence of factors such as board inhomogeneity, global temperature fluctuations, and lighting changes can be effectively eliminated. These factors often mask the weak abnormal signals of individual vias in traditional global detection methods. After constructing the local context decoding region, the physical state signatures of all neighboring vias within the local context decoding region are combined into a decoding region physical state signature set. This physical state signature set establishes a dynamic and adaptive normal baseline for the target via. Compared with the fixed threshold method, this locally statistical reference baseline can better adapt to process fluctuations in different regions and batches. For example, in areas with slightly lower thermal conductivity of the board material, the thermal signatures of all vias may be generally lower, but they still maintain relative consistency in the local context. Only when a via shows a significant difference relative to its direct neighbors is it considered a true anomaly. Next, the mean vector and covariance matrix of the decoding region physical state signature set for each via are calculated. The mean vector represents the central tendency of the via features in the local region, while the covariance matrix characterizes the interrelationships and variability among these features. It not only considers the discreteness of each feature dimension but also captures the intrinsic correlation between thermal and electromagnetic features. For example, in the normal region, the thermal and electromagnetic signatures of vias often have a positive correlation, while blocking the via will disrupt this inherent relationship. Based on these statistics, the Mahalanobis distance between the physical state signature of each via and its corresponding decoded area signature set is calculated. This Mahalanobis distance takes into account the geometric structure of the feature space, transforming the original feature space into a new coordinate system. In this new coordinate system, the variance in each direction is normalized, and the correlation between features is eliminated. This means that when calculating the distance, feature dimensions with larger variances (i.e., normal vias exhibit greater natural fluctuations in these dimensions) are automatically assigned smaller weights, while feature dimensions with smaller variances (i.e., normal vias exhibit stable behavior in these dimensions) are assigned larger weights. This adaptive weighting mechanism makes the Mahalanobis distance more sensitive to real anomalies while being more robust to normal natural fluctuations.Finally, the Mahalanobis distance is mapped to a local consistency anomaly score using an exponential function. This score represents the probability density of the physical state signature of the target via appearing in a multivariate distribution constructed by its local context. This mapping transforms the distance value into a probabilistic score between 0 and 1, where values ​​close to 0 indicate high anomaly and values ​​close to 1 indicate high normality, thus forming a more explicit probabilistic interpretation.

[0037] To address the issue that individual anomalous signals in dense arrays are easily overwhelmed by the global background, this embodiment establishes a local context decoding region and employs covariance matrix and Mahalanobis distance to capture the intrinsic correlation between features within the local region. This transforms the multidimensional collaborative deviation of vias into significant statistical anomalies, enabling the features of individual blocked vias to be highlighted and processed. Consequently, it can effectively identify weak anomalous signals that are overwhelmed by traditional global thresholding methods.

[0038] In one embodiment, the step of performing a quality assessment on the thermal and electromagnetic signatures of each via, and dynamically fusing them to generate a comprehensive anomaly confidence level based on the results of the quality assessment, includes: S51, calculate the determination coefficient of the exponential decay fitting, and obtain the thermal characteristic quality score of each via based on the determination coefficient; S52, obtain the amplitude standard deviation and amplitude mean of each via in the local dielectric response diagram, obtain the coefficient of variation based on the amplitude standard deviation and amplitude mean, and obtain the electromagnetic characteristic quality score of each via based on the coefficient of variation, using the following formula: ; In the formula, Q rf, i This represents the electromagnetic characteristic quality score of the i-th via. Std represents the coefficient of variation. i Mea represents the standard deviation of the amplitude of the local dielectric response plot corresponding to the i-th via. i This represents the average amplitude of the local dielectric response map corresponding to the i-th via; S53, calculate the dynamic weight of the thermal characteristic and the dynamic weight of the electromagnetic characteristic for each via based on the quality score, using the following formula: , ; In the formula, W the, i Q represents the dynamic weight of the thermal characteristics of the i-th via. the, i W represents the thermal characteristic quality score of the i-th via. rf, i Q represents the dynamic weight of the electromagnetic characteristics of the i-th via. rf, i This represents the electromagnetic characteristic quality score of the i-th via; S54, by performing global Z-score normalization on the thermal signatures and electromagnetic signatures of all vias respectively, the thermal anomaly amplitude and electromagnetic anomaly amplitude are obtained; S55, the thermal anomaly amplitude and the electromagnetic anomaly amplitude are fused according to the dynamic weights of the thermal features and the electromagnetic features to obtain the comprehensive anomaly confidence level, as shown in the formula: ; In the formula, FS i W represents the overall anomaly confidence level of the i-th via. the,i A represents the dynamic weight of the thermal characteristics of the i-th via. the,i W represents the thermal anomaly amplitude of the i-th via. rf,i A represents the dynamic weight of the electromagnetic characteristics of the i-th via. rf,i This represents the electromagnetic anomaly amplitude of the i-th via.

[0039] As described in steps S51-S55 above, the acquisition process of thermal and electromagnetic signatures is susceptible to various interference factors: thermal signatures rely on the fitting accuracy of the temperature decay curve. If there are airflow fluctuations, local reflections, or contamination on the via surface in the detection environment, the noise of the temperature data will increase, the goodness of fit will decrease, and the reliability of the thermal signature will be reduced. Electromagnetic signatures rely on the stable coupling of the high-frequency electromagnetic field. If the distance between the microwave probe and the PCB surface is uneven, or if there are tiny air bubbles in the dielectric on the board surface, the signal-to-noise ratio of the dielectric response signal will decrease, and the reliability of the electromagnetic signature will be reduced. Traditional multi-feature fusion does not consider the dynamic changes in feature quality. When the quality of a feature is reduced due to interference, it will still participate in the fusion at a fixed ratio, causing the comprehensive result to deviate from the true abnormal state. Therefore, this invention ensures that the comprehensive anomaly confidence is preferentially generated based on more reliable features through a quality assessment mechanism and a dynamic weight fusion strategy, avoiding interference from low-quality features in the detection results. At the same time, it fully leverages the synergistic and complementary advantages of the two physical fields to provide a more robust basis for via blockage determination.

[0040] In assessing the reliability of thermal characteristics, this invention does not use the original temperature decay curve data, but rather evaluates the measurement quality through exponential goodness-of-fit analysis. Specifically, it calculates the coefficient of determination between the temperature decay curve and the ideal exponential decay model. This coefficient of determination ranges from [0,1], with a value closer to 1 indicating a higher degree of agreement between the actual measurement data and the ideal physical model. This is because, under ideal conditions without interference, the temperature decay of a via made of uniform material should strictly follow an exponential law. However, when factors such as measurement noise, uneven thermal excitation, or changes in surface emissivity exist, the goodness of fit decreases. Therefore, the coefficient of determination quantifies the degree of interference in this thermal measurement, providing a more objective basis for subsequent weight allocation.

[0041] In terms of electromagnetic characteristic reliability assessment, this invention employs a method based on local region signal stability analysis. Specifically, it first extracts all pixel values ​​from the local region of interest for each via from the dielectric response amplitude map, and then calculates the coefficient of variation (the ratio of standard deviation to mean) of the signal in that region. The coefficient of variation is a dimensionless quantity that effectively eliminates the influence of absolute signal strength, purely reflecting the relative fluctuation of the signal within the region. Ideally, the electromagnetic field distribution above a perfectly metallized via should be relatively uniform, exhibiting a low coefficient of variation. However, when measurement noise, electromagnetic interference, or surface irregularities exist, the fluctuation of the signal within the region increases. Therefore, by calculating the coefficient of variation, the reliability of this electromagnetic measurement can be quantified. This statistical stability-based assessment method can effectively identify low-quality measurement data caused by probe jitter and external electromagnetic interference. After completing the credibility assessment, dynamic weight allocation is performed using a normalized weight calculation strategy based on relative quality comparison. The advantage of this weight allocation method is that when the measurement quality of one physical field is significantly better than another, the higher-quality field automatically receives a larger decision weight. When the quality of the two fields is comparable, the evidence from both is adopted in a balanced manner. This mechanism can effectively handle various extreme situations. For example, when thermal measurements are severely interfered with while electromagnetic measurements are of good quality, the system automatically transfers the decision-making power to the electromagnetic field evidence, preventing defective thermal data from introducing erroneous results. Z-score standardization transforms the thermal and electromagnetic signatures into a standard normal distribution with a mean of 0 and a standard deviation of 1. This allows the amplitudes of thermal and electromagnetic anomalies to directly reflect the degree of deviation from normal levels. The larger the absolute value of the thermal anomaly amplitude, the further the thermal signature deviates from the normal mean of the entire plate, and the more significant the thermal anomaly. The same applies to the electromagnetic anomaly amplitude. Finally, a weighted fusion is used to obtain a comprehensive anomaly confidence level, which further highlights signals with high reliability.

[0042] Compared to traditional multi-sensor fusion methods that often assume comparable data quality from each sensor and ignore the different interferences that may occur in measurements of different physical fields in practical applications, this invention takes data quality assessment as the core element of fusion decision-making. This enables the system to recognize its own sensing reliability, ensuring that the detection of weak anomalous signals submerged by the background will not fail due to a temporary decline in the quality of a certain physical field. It also prevents false anomalies caused by low-quality data from interfering with the detection results. Especially in the detection of high-density via arrays, it can fully utilize the complementary advantages of different physical fields under different operating conditions.

[0043] In one embodiment, the step of determining whether a via is blocked based on the local consistency anomaly score and the overall anomaly confidence score includes: S61, a first initial global threshold is obtained by statistically analyzing the mean and standard deviation of the local consistency anomaly scores of all vias, and a second initial global threshold is obtained by statistically analyzing the mean and standard deviation of the comprehensive anomaly confidence scores of all vias. S62, obtain the process tolerance coefficient according to the type and process specifications of the PCB board, and adjust the first initial global threshold and the second initial global threshold according to the process tolerance coefficient to obtain the first adaptive threshold and the second adaptive threshold. S63, compare the local consistency anomaly score and the overall anomaly confidence score of all vias with the first adaptive threshold and the second adaptive threshold, respectively; S631, if the local consistency anomaly score is less than the first adaptive threshold, it is determined to be blocked; S632, if only the comprehensive anomaly confidence level is greater than the second adaptive threshold, the corresponding via is marked as suspected blockage, and a secondary verification is performed based on the deviation direction of the thermal signature and the electromagnetic signature relative to their respective global mean. If the relative deviation direction of the thermal signature and the electromagnetic signature is consistent and both are lower than their respective global mean, then it is finally determined to be a blockage; otherwise, it is determined to be normal.

[0044] As described in steps S61-S63 above, from the perspective of actual industrial inspection needs, PCB via blockage detection must simultaneously meet two core requirements: first, to avoid missed detections, especially in high-end scenarios such as high-density interconnect boards and IC carrier boards, where a single blocked via can lead to the failure of the entire board, and missed detections can cause serious subsequent quality risks; second, to minimize false detections, as normal vias misjudged as blocked require manual verification, and too many false detections will significantly increase production costs and reduce inspection efficiency. Therefore, this invention establishes a layered and adaptive judgment logic to ensure a high detection rate for via blockages in dense areas, while reducing the risk of false judgments through secondary verification, ultimately achieving the goal of accurate blockage identification.

[0045] The dual-stage adaptive threshold setting strategy of this invention, in the first stage, statistically analyzes the local consistency anomaly scores of all vias, calculates their mean and standard deviation, and multiplies these scores by a corresponding transformation coefficient to obtain a first initial global threshold, with a value between 2 and 3. This is because the local consistency anomaly scores of normal vias follow a certain distribution pattern, while anomalies exhibit significant deviations from this distribution; therefore, the first initial global threshold quantifies this deviation. Simultaneously, a similar statistical analysis is performed on the comprehensive anomaly confidence scores of all vias, calculating their mean and standard deviation, and multiplying these scores by a corresponding transformation coefficient to obtain a second initial global threshold. By using the first and second initial global thresholds, the decision criteria can adapt to the actual characteristic distribution of the current PCB board, providing a unified solution for the inspection of PCB boards from different batches and using different processes. In the second stage, the initial threshold is finely adjusted using a process tolerance coefficient. This coefficient is obtained by querying a pre-generated baseline tolerance mapping table based on PCB type, number of layers, and customer specifications (such as type, copper thickness, hole diameter, aspect ratio, and customer quality level). This coefficient serves as the adjustment factor for the decision threshold, reflecting the stringency required for the batch of boards being inspected. When extremely high inspection quality is required and some false alarms are permissible, the process tolerance coefficient can be less than 1 to lower the threshold and increase inspection sensitivity. Conversely, when strict false alarm control is required, the coefficient can be greater than 1 to raise the threshold and enhance inspection specificity, allowing the invention to flexibly adapt to the inspection needs of different application scenarios. In the decision execution stage, a tiered verification decision logic is adopted. First, the local consistency anomaly score of each via is compared with a first adaptive threshold. If the local consistency anomaly score is less than the first adaptive threshold, the via is directly determined to be blocked. This is because the local consistency anomaly score reflects the relative degree of anomaly of the via in its most immediate neighborhood. When it significantly deviates from the local normal background, it represents direct evidence of anomaly, and a judgment can be made without further verification. For vias that fail to meet the direct judgment condition but whose overall anomaly confidence is greater than the second adaptive threshold, they are marked as suspected blockages, and a secondary verification mechanism is initiated. The core of the secondary verification is to check whether the original thermal signature and electromagnetic signature of the via are simultaneously lower than their respective global mean. This requires that the two independent physical field evidences must consistently point to the blockage characteristics. That is, not only is a high overall score required, but thermal measurements also require that thermal conductivity is lower than normal, and electromagnetic measurements also require that scattering ability is lower than normal. Through this verification mechanism based on the consistency of multi-physics field evidence, false anomaly signals caused by accidental interference in a single field measurement can be effectively filtered out, significantly improving the reliability of the judgment and thus ensuring the efficiency of detection.

[0046] like Figure 2As shown, the present invention also provides a PCB board via blockage detection system based on fusion imaging, comprising: The hole modeling module is used to obtain the theoretical center coordinates and theoretical radius of all vias based on the PCB board design information, and generate a hole mask matrix based on the theoretical center coordinates and theoretical radius. The synchronous imaging module is used to synchronously acquire the transient thermal characteristic sequence and local dielectric response map of the via according to the area defined by the aperture mask matrix; The signature extraction module is used to decouple and extract an independent physical state signature for each via from the transient thermal feature sequence and the local dielectric response map based on the via mask matrix, wherein the physical state signature includes at least a thermal signature and an electromagnetic signature; The local scoring module is used to construct a local context decoding region for each via and calculate the local consistency anomaly score for each via based on the difference between the physical state signature of each via and the physical state signature of its neighboring vias in the corresponding local context decoding region. The dynamic fusion module is used to perform quality assessment on the thermal signature and electromagnetic signature of each via, and dynamically fuse them to generate a comprehensive anomaly confidence level based on the results of the quality assessment. The comprehensive judgment module is used to determine whether the via is blocked based on the local consistency anomaly score and the comprehensive anomaly confidence level.

[0047] The local scoring module includes: A hole spacing acquisition unit is used to acquire the typical hole spacing between each via and all other vias based on the hole position mask matrix. The decoding area construction unit is used to select a preset number of nearest-neighbor vias to form several local context decoding areas, with each via as the center and according to the typical via spacing, and to combine the physical state signatures of all nearest-neighbor vias in the local context decoding area into a decoding area physical state signature set. The matrix calculation unit is used to calculate the mean vector and covariance matrix of the physical state signature set of the decoding area for each via; The Mahalanobis distance calculation unit is used to calculate the Mahalanobis distance between the physical state signature of each via and the mean vector and covariance matrix of the physical state signature set of the corresponding decoding area. Anomaly mapping unit is used to map the Mahalanobis distance into a probability value through an exponential function, which serves as the local consistency anomaly score.

[0048] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of a PCB board via blockage detection method using fusion imaging.

[0049] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a PCB board via blockage detection method using fusion imaging.

[0050] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0051] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for detecting via blockage on a PCB board using fusion imaging, characterized in that, include: The theoretical center coordinates and theoretical radius of all vias are obtained based on the PCB design information, and a via mask matrix is ​​generated based on the theoretical center coordinates and theoretical radius. The transient thermal characteristic sequence and local dielectric response map of the via are synchronously acquired based on the area defined by the via mask matrix; Based on the aperture mask matrix, an independent physical state signature is extracted from each via from the transient thermal feature sequence and the local dielectric response map, wherein the physical state signature includes at least a thermal signature and an electromagnetic signature; A local context decoding region is constructed for each via, and a local consistency anomaly score is calculated for each via based on the difference between the physical state signature of each via and the physical state signature of its neighboring vias in the corresponding local context decoding region. The thermal and electromagnetic signatures of each via are evaluated for quality, and a comprehensive anomaly confidence level is dynamically generated based on the results of the quality evaluation. The via is determined to be blocked based on the local consistency anomaly score and the overall anomaly confidence score.

2. The PCB board via blockage detection method based on fusion imaging according to claim 1, characterized in that, The step of decoupling each via from the transient thermal feature sequence and the local dielectric response map based on the via mask matrix and extracting an independent physical state signature includes: Based on the aperture mask matrix, extract the temperature decay curve of the average temperature of each via over time from the transient thermal feature sequence; For each via, the temperature decay curve is fitted with an exponential decay rate starting from the temperature peak point to obtain the decay time constant. Calculate the thermal signature of each via based on the decay time constant; The mean amplitude and mean phase of each via are extracted from the dielectric response map based on the via mask matrix. The reference amplitude and reference phase of the via-free background region are obtained based on the dielectric response diagram. The composite scattering intensity of each via is calculated based on the mean amplitude, mean phase, reference amplitude, and reference phase. The composite scattering intensity of all vias is normalized to obtain the electromagnetic signature of each via. The physical state signatures of all vias are obtained based on the electromagnetic signature and the thermal signature.

3. The PCB board via blockage detection method based on fusion imaging according to claim 2, characterized in that, The step of obtaining the physical state signatures of all vias based on the electromagnetic signature and the thermal signature includes: Isolated vias that are all more than a preset safety distance from all adjacent vias are selected from the via mask matrix, and a reference sample set is established based on the electromagnetic signature and thermal signature of the isolated vias, wherein the reference sample set includes a reference thermal signature set and a reference electromagnetic signature set. The distribution centers of the reference thermal signature set and the reference electromagnetic signature set are calculated respectively, and the ratio between the two distribution centers is calculated as the cross-field alignment coefficient. The thermal signature and electromagnetic signature of all vias are scaled by the cross-field alignment coefficient, and the numerical ranges of the thermal signature and the electromagnetic signature are aligned with the two distribution centers respectively. The aligned thermal signature and the electromagnetic signature are combined to form a two-dimensional vector, which serves as the physical state signature for all vias.

4. The PCB board via blockage detection method based on fusion imaging according to claim 1, characterized in that, The step of constructing a local context decoding region for each via and calculating the local consistency anomaly score for each via based on the difference between the physical state signature of each via and the physical state signatures of its neighboring vias within the corresponding local context decoding region includes: The typical hole spacing between each via and all other vias is obtained based on the hole mask matrix. Centered on each via, a preset number of nearest-neighbor vias are selected according to the typical via spacing to form several local context decoding regions. The physical state signatures of all nearest-neighbor vias in the local context decoding region are combined into a decoding region physical state signature set. Calculate the mean vector and covariance matrix of the physical state signature set of the decoding area for each via; Calculate the Mahalanobis distance between the physical state signature of each via and the mean vector and covariance matrix of the physical state signature set of the corresponding decoding area; The Mahalanobis distance is mapped to a probability value through an exponential function, which is used as the local consistency anomaly score.

5. The PCB board via blockage detection method using fusion imaging according to claim 2, characterized in that, The steps of performing quality assessment on the thermal and electromagnetic signatures of each via, and dynamically fusing them to generate a comprehensive anomaly confidence level based on the results of the quality assessment, include: Calculate the determination coefficient of the exponential decay fit, and obtain the thermal characteristic quality score of each via based on the determination coefficient; Obtain the amplitude standard deviation and amplitude mean of each via in the local dielectric response diagram, obtain the coefficient of variation based on the amplitude standard deviation and amplitude mean, and obtain the electromagnetic characteristic quality score of each via based on the coefficient of variation. Calculate the dynamic weight of thermal characteristics and the dynamic weight of electromagnetic characteristics for each via based on the quality score; By performing global Z-score normalization on the thermal signatures and electromagnetic signatures of all vias respectively, the thermal anomaly amplitude and electromagnetic anomaly amplitude are obtained. The thermal anomaly amplitude and the electromagnetic anomaly amplitude are fused based on the dynamic weights of the thermal features and the electromagnetic features to obtain a comprehensive anomaly confidence level.

6. The PCB board via blockage detection method based on fusion imaging according to claim 1, characterized in that, The step of determining whether a via is blocked based on the local consistency anomaly score and the overall anomaly confidence level includes: A first initial global threshold is obtained by statistically analyzing the mean and standard deviation of the local consistency anomaly scores of all vias, and a second initial global threshold is obtained by statistically analyzing the mean and standard deviation of the comprehensive anomaly confidence scores of all vias. The process tolerance coefficient is obtained according to the type and process specifications of the PCB board. The first initial global threshold and the second initial global threshold are adjusted according to the process tolerance coefficient to obtain the first adaptive threshold and the second adaptive threshold. The local consistency anomaly scores and the overall anomaly confidence scores of all vias are compared with the first adaptive threshold and the second adaptive threshold, respectively. If the local consistency anomaly score is less than the first adaptive threshold, it is determined to be blocked; If only the overall anomaly confidence level is greater than the second adaptive threshold, the corresponding via is marked as a suspected blockage, and a secondary verification is performed based on the deviation direction of the thermal signature and the electromagnetic signature relative to their respective global mean. If the relative deviation direction of the thermal signature and the electromagnetic signature is consistent and both are lower than their respective global mean, then it is finally determined to be a blockage; otherwise, it is determined to be normal.

7. A PCB board via blockage detection system using fusion imaging, characterized in that, include: The hole modeling module is used to obtain the theoretical center coordinates and theoretical radius of all vias based on the PCB board design information, and generate a hole mask matrix based on the theoretical center coordinates and theoretical radius. The synchronous imaging module is used to synchronously acquire the transient thermal characteristic sequence and local dielectric response map of the via according to the area defined by the aperture mask matrix; The signature extraction module is used to decouple and extract an independent physical state signature for each via from the transient thermal feature sequence and the local dielectric response map based on the via mask matrix, wherein the physical state signature includes at least a thermal signature and an electromagnetic signature; The local scoring module is used to construct a local context decoding region for each via and calculate the local consistency anomaly score for each via based on the difference between the physical state signature of each via and the physical state signature of its neighboring vias in the corresponding local context decoding region. The dynamic fusion module is used to perform quality assessment on the thermal signature and electromagnetic signature of each via, and dynamically fuse them to generate a comprehensive anomaly confidence level based on the results of the quality assessment. The comprehensive judgment module is used to determine whether the via is blocked based on the local consistency anomaly score and the comprehensive anomaly confidence level.

8. The PCB board via blockage detection system based on fusion imaging according to claim 7, characterized in that, The local scoring module includes: A hole spacing acquisition unit is used to acquire the typical hole spacing between each via and all other vias based on the hole position mask matrix. The decoding area construction unit is used to select a preset number of nearest-neighbor vias to form several local context decoding areas, with each via as the center and according to the typical via spacing, and to combine the physical state signatures of all nearest-neighbor vias in the local context decoding area into a decoding area physical state signature set. The matrix calculation unit is used to calculate the mean vector and covariance matrix of the physical state signature set of the decoding area for each via; The Mahalanobis distance calculation unit is used to calculate the Mahalanobis distance between the physical state signature of each via and the mean vector and covariance matrix of the corresponding physical state signature set of the decoding area. Anomaly mapping unit is used to map the Mahalanobis distance into a probability value through an exponential function, which serves as the local consistency anomaly score.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images