HDI circuit board and method for intelligent mobile communication terminal

Abstract

The present application provides an HDI circuit board and method for intelligent mobile communication terminals, which relates to the field of circuit board detection technology. It uses image detection and analysis technology based on computer vision to extract image distortion features and enhance spatial features of blind hole stacking hole perspective reference images extracted from a database and blind hole stacking hole perspective images obtained by X-ray detector perspective, so as to intelligently obtain distortion correction coefficients based on the semantic metric value between the enhanced blind hole stacking hole perspective reference features and blind hole stacking hole perspective detection features. In this way, the accuracy of offset measurement can be improved. Moreover, by adopting image processing technology based on computer vision, not only the detection accuracy is improved, but also the intelligent offset measurement is realized, making blind hole offset detection in complex environments more reliable and efficient.

Description

Technical Field

[0001] The present application relates to the technical field of circuit board detection, and more specifically, to an HDI circuit board and method for intelligent mobile communication terminals. Background Art

[0002] HDI (High Density Interconnect) circuit boards play an important role in smart mobile communication terminals due to their high-density wiring and miniaturization, especially in the context of the increasing requirements for miniaturization and high performance of today's smartphones and other portable electronic devices. HDI circuit boards achieve higher integration and more compact designs by adding structures such as blind vias and buried vias. However, the alignment accuracy of blind vias has a crucial impact on the function and performance of the circuit board.

[0003] Patent CN113280737A proposes a method for detecting the offset of blind holes in high-order HDI printed circuit boards. The method sets a laser blind hole array in the non-working area of ​​the circuit board, and after the layers are pressed together, uses an X-ray detector to see through the overlapping blind holes to obtain a perspective image of the overlapping blind holes. On this basis, the offset between the blind holes in the image is measured, i.e., the second offset, and the offset of the blind holes on the actual circuit board is obtained by proportional conversion, i.e., the third offset.

[0004] However, in the above patent, when using blind hole overlapped hole perspective images to perform blind hole distance testing, in the process of converting the second offset to obtain the third offset, only the proportional relationship between the perspective image and the real object is considered (for example, calibration through a standard reference object of known size), while the influence of factors such as image distortion on the proportional relationship is not considered. That is, image distortion such as lens distortion, image deformation, etc. will change the actual distance relationship in the image, especially the distortion in the edge area is more serious. If the distortion is not considered, the measurement of the third offset will have a large error, affecting the accuracy of the measurement.

[0005] Therefore, an optimized solution for HDI circuit boards for smart mobile communication terminals is desired. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide an HDI circuit board and method for an intelligent mobile communication terminal.

[0007] According to one aspect of the present application, a method for an HDI circuit board for an intelligent mobile communication terminal is provided, which includes:

[0008] A plurality of laser blind hole arrays are arranged outside the working area of ​​each layer of the circuit board, and the position and size of the laser blind hole arrays on each layer are ensured to be consistent; each layer of the circuit board is pressed together to form a blind hole stacking hole; the blind hole stacking hole is detected by an X-ray detector to obtain a blind hole stacking hole perspective image; based on the blind hole stacking hole perspective image, an offset of the blind hole of the circuit board is obtained, and the blind hole of the circuit board is formally drilled based on the offset; wherein, based on the blind hole stacking hole perspective image, the offset of the blind hole of the circuit board is obtained, including:

[0009] Acquire the blind hole stacking hole perspective image;

[0010] Determine the distance between the centers of any two blind hole objects in the blind hole stacked hole perspective image based on machine vision to obtain a second offset;

[0011] Based on the blind hole-on-hole perspective image, determining the distortion correction coefficient includes: extracting a blind hole-on-hole perspective reference image annotated with a first distortion correction coefficient from a database; performing image twin distortion detection on the blind hole-on-hole perspective reference image and the blind hole-on-hole perspective image to obtain a blind hole-on-hole perspective reference feature map and a blind hole-on-hole perspective detection feature map; performing feature enhancement based on spatial perception on the blind hole-on-hole perspective reference feature map and the blind hole-on-hole perspective detection feature map to obtain a blind hole-on-hole perspective reference enhancement feature and a blind hole-on-hole perspective detection enhancement feature; and obtaining the distortion correction coefficient based on the image semantic similarity between the blind hole-on-hole perspective reference enhancement feature and the blind hole-on-hole perspective detection enhancement feature.

[0012] Based on the distortion correction coefficient and the proportional relationship between the perspective image and the real object, the second offset is converted to obtain a third offset.

[0013] According to another aspect of the present application, a HDI circuit board for a smart mobile communication terminal is provided. The HDI circuit board for a smart mobile communication terminal is manufactured by the method for manufacturing a HDI circuit board for a smart mobile communication terminal as described above.

[0014] Compared with the prior art, the HDI circuit board and method for intelligent mobile communication terminals provided by the present application use computer vision-based image detection and analysis technology to extract image distortion features and enhance spatial features of blind hole stacking hole perspective reference images extracted from a database and blind hole stacking hole perspective images obtained by X-ray detector perspective, so as to intelligently obtain the distortion correction coefficient based on the semantic metric value between the enhanced blind hole stacking hole perspective reference features and the blind hole stacking hole perspective detection features. In this way, the measurement error caused by image distortion can be effectively reduced and the accuracy of offset measurement can be improved. Moreover, by adopting image processing technology based on computer vision, not only the detection accuracy is improved, but also the intelligent offset measurement is realized, making the blind hole offset detection in complex environments more reliable and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 The present invention is a flow chart of obtaining the offset of the blind hole of the circuit board based on the blind hole stacking hole perspective image in the method of the HDI circuit board for the intelligent mobile communication terminal according to the embodiment of the present application.

[0017] Figure 2 The present invention is a flow chart of determining a distortion correction coefficient based on the blind hole stacking hole perspective image in a method for an HDI circuit board for an intelligent mobile communication terminal according to an embodiment of the present application.

[0018] Figure 3 The present invention is a schematic diagram of data flow for determining a distortion correction coefficient based on the blind hole stacking hole perspective image in a method for an HDI circuit board for an intelligent mobile communication terminal according to an embodiment of the present application.

[0019] Figure 4 In the method for an HDI circuit board for an intelligent mobile communication terminal according to an embodiment of the present application, the blind hole stacking hole perspective reference feature map and the blind hole stacking hole perspective detection feature map are subjected to feature enhancement based on spatial perception to obtain a blind hole stacking hole perspective reference enhancement feature and a blind hole stacking hole perspective detection enhancement feature.

[0020] Figure 5In the method for an HDI circuit board for an intelligent mobile communication terminal according to an embodiment of the present application, a set of blind hole stacking hole perspective reference pixel granularity feature vectors is subjected to semantic association and attenuation modulation based on soft space constraints to obtain a blind hole stacking hole perspective reference semantic association space soft constraint score topology matrix.

[0021] Figure 6 In the method for an HDI circuit board for an intelligent mobile communication terminal according to an embodiment of the present application, based on the blind hole stacking hole perspective reference semantic association space soft constraint score topology matrix, a set of the blind hole stacking hole perspective reference pixel granularity feature vectors is convolutionally enhanced to obtain the blind hole stacking hole perspective reference enhancement feature. DETAILED DESCRIPTION

[0022] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0023] In the field of intelligent mobile communication terminals, high-density interconnect (HDI) circuit boards play a key role because they can achieve high-density wiring and miniaturization. With the growing demand for miniaturization and high performance in smartphones and other portable electronic devices, HDI circuit boards have achieved higher integration and more compact layouts by adding designs such as blind and buried vias. However, the alignment accuracy of the blind vias is critical to the function and performance of the circuit board.

[0024] Patent CN113280737A provides a method for detecting the offset of blind holes in high-order HDI printed circuit boards. The method involves setting up a laser blind hole array in the non-working area of ​​the circuit board, and after the layers are pressed together, using an X-ray detector to perform perspective on the overlapping blind holes to obtain a perspective image of the blind hole overlap. By measuring the offset between the blind holes in the image, i.e., the second offset, and performing proportional conversion, the offset of the blind holes on the actual circuit board, i.e., the third offset, is obtained.

[0025] However, this patent has certain limitations when using blind hole overlap perspective images to test blind hole distance. In the process of converting the second offset into the third offset, only the proportional relationship between the perspective image and the real object is considered (for example, calibration through a standard reference object of known size), and it does not fully consider image distortion, such as lens distortion and image deformation, which may change the actual distance relationship in the image, especially in the edge area of ​​the image. If these distortions are ignored, it may cause large errors in the measurement results of the third offset, thereby affecting the accuracy of the measurement results.

[0026] Based on this, the present application proposes a method for an HDI circuit board for an intelligent mobile communication terminal, which includes: setting a plurality of laser blind hole arrays outside the working area of ​​each layer of the circuit board, and ensuring that the position and size of the laser blind hole arrays on each layer are consistent; performing a pressing process on each layer of the circuit board to obtain a blind hole stacking hole; performing a perspective detection of the blind hole stacking hole through an X-ray detector to obtain a blind hole stacking hole perspective image; based on the blind hole stacking hole perspective image, obtaining the offset of the blind hole of the circuit board, and formally drilling the blind hole of the circuit board based on the offset.

[0027] It should be understood that first, multiple laser blind hole arrays need to be arranged outside the working area of ​​each layer of the circuit board, and the position and size of the laser blind hole arrays on each layer are ensured to be consistent. Here, the working area of ​​the circuit board refers to the internal graphic distribution area of ​​the circuit board, which is generally used for circuit component layout. In this application, a blind hole refers to a conductive hole that connects the surface layer and the inner layer without penetrating the entire board, and a laser blind hole array is formed by arranging multiple blind holes. Generally speaking, in order not to interfere with the layout of circuit components, the laser blind hole array is usually arranged at the edge of the board outside the graphic distribution area, and each board edge is provided with a laser blind hole array, and the laser blind hole array corresponds to the position of each layer of the board. Then, it is necessary to perform a lamination process on each layer of the circuit board to obtain a blind hole stacking hole. Since the laser blind hole array corresponds to the position of each layer of the board, after the circuit board is pressed, the blind holes of each layer overlap correspondingly, forming a blind hole stacking hole, that is, a plurality of blind hole stacking hole arrays. Then, the blind hole stacking hole is perspectively inspected by an X-ray detector to obtain a blind hole stacking hole perspective image. Specifically, an X-ray detector is used to see through the blind hole stacking holes, and perspective imaging is performed on a single blind hole stacking hole in each blind hole stacking hole array. This can avoid interference between multiple blind hole stacking hole images when imaging the blind hole stacking hole array as a whole. Therefore, the obtained blind hole stacking hole perspective image is also a perspective image of a single blind hole stacking hole. Finally, based on the blind hole stacking hole perspective image, the offset of the blind hole of the circuit board is obtained, and the blind hole of the circuit board is formally drilled based on the offset. By performing offset detection on the blind hole stacking hole perspective image and calculating the offset, if the offset is within the threshold range, it is proved that the blind hole offset caused by the expansion and contraction of each layer of the board is within an acceptable range, and formal drilling can be performed. This is conducive to ensuring the alignment accuracy of the formal drilling, thereby making the circuit connection of the circuit board more stable.

[0028] Specifically, Figure 1 The present invention is a flowchart of obtaining the offset of the blind hole of the circuit board based on the blind hole stacking perspective image and performing formal drilling of the blind hole of the circuit board based on the offset in the method of the HDI circuit board for the intelligent mobile communication terminal according to the embodiment of the present application. Figure 1As shown, based on the blind hole stacking perspective image, the offset of the blind hole of the circuit board is obtained, and the blind hole of the circuit board is formally drilled based on the offset, including: S110, acquiring the blind hole stacking perspective image; S120, determining the distance between the centers of any two blind hole objects in the blind hole stacking perspective image based on machine vision to obtain a second offset; S130, determining the distortion correction coefficient based on the blind hole stacking perspective image; S140, converting the second offset to obtain a third offset based on the distortion correction coefficient and the proportional relationship between the perspective image and the real object.

[0029] In step S110, the blind via stacking perspective image is obtained. It should be understood that the blind via stacking perspective image contains the relative position information of the blind vias in the circuit board, which is the data source for analyzing whether there is an offset between the blind vias and calculating the offset between the blind vias.

[0030] In step S120, the distance between the centers of any two blind hole objects in the blind hole stacked hole perspective image is determined based on machine vision to obtain a second offset. It should be understood that since the size of each blind hole is equal, if any two blind holes on the circuit board are offset, the two blind hole images will appear to be intersecting in the blind hole stacked hole perspective image, that is, the centers of the two blind hole images do not overlap, and the offset of the blind hole can be obtained by measuring the distance between the two centers. In the present application, the distance between the centers of the blind hole objects is determined by machine vision, which can provide a more accurate second offset measurement, which helps to improve the efficiency and accuracy of the measurement. In particular, in a specific implementation of the embodiment of the present application, a possible implementation method for determining the distance between the centers of any two blind hole objects in the blind hole stacked hole perspective image based on machine vision may be: a. applying a Canny edge detector to identify the edge of the blind hole object, and using the Hough circle transform to detect the circle of the blind hole from the edge image; b. for each detected blind hole, calculating the coordinates of its center; c. selecting any two blind holes, using the Euclidean distance formula to calculate the distance between their centers, and converting this distance into an actual physical size, i.e., a second offset.

[0031] In step S130, a distortion correction coefficient is determined based on the blind hole stacked hole perspective image. Accordingly, in determining the distortion correction coefficient based on the blind hole stacked hole perspective image, the technical concept of the present application is to extract a blind hole stacked hole perspective reference image annotated with a first distortion correction coefficient from a database, and use image detection and analysis technology based on computer vision to extract image distortion features and enhance spatial features of the blind hole stacked hole perspective reference image and the blind hole stacked hole perspective image, so as to intelligently obtain the distortion correction coefficient based on the semantic metric value between the enhanced blind hole stacked hole perspective reference feature and the blind hole stacked hole perspective detection feature. In this way, the measurement error caused by image distortion can be effectively reduced and the accuracy of offset measurement can be improved. By using image processing technology based on computer vision, not only the detection accuracy is improved, but also the intelligent offset measurement is realized, making the blind hole offset detection in complex environments more reliable and efficient.

[0032] Specifically, Figure 2 The present invention is a flow chart of determining a distortion correction coefficient based on the blind hole stacking hole perspective image in a method for an HDI circuit board for an intelligent mobile communication terminal according to an embodiment of the present application. Figure 3 The present invention is a data flow diagram of determining the distortion correction coefficient based on the blind hole stacking hole perspective image in the method for the HDI circuit board for intelligent mobile communication terminals according to the embodiment of the present application. Figure 2 and Figure 3 As shown, based on the blind hole stacked hole perspective image, a distortion correction coefficient is determined, including: S131, extracting a blind hole stacked hole perspective reference image annotated with a first distortion correction coefficient from a database; S132, performing image twin distortion detection on the blind hole stacked hole perspective reference image and the blind hole stacked hole perspective image to obtain a blind hole stacked hole perspective reference feature map and a blind hole stacked hole perspective detection feature map; S133, performing feature enhancement based on spatial perception on the blind hole stacked hole perspective reference feature map and the blind hole stacked hole perspective detection feature map to obtain a blind hole stacked hole perspective reference enhancement feature and a blind hole stacked hole perspective detection enhancement feature; S134, obtaining the distortion correction coefficient based on the image semantic similarity between the blind hole stacked hole perspective reference enhancement feature and the blind hole stacked hole perspective detection enhancement feature.

[0033] In step S131, a blind hole-on-hole perspective reference image annotated with a first distortion correction coefficient is extracted from a database. It should be understood that the blind hole-on-hole perspective reference image annotated with a first distortion correction coefficient provides a known distortion benchmark. The blind hole-on-hole perspective reference image and the acquired blind hole-on-hole perspective image usually contain similar features. By establishing a corresponding relationship between the two images, the distortion features can be more accurately extracted and compared, thereby helping to accurately determine the distortion correction coefficient, thereby improving the accuracy of the final measurement result.

[0034] In step S132, image twin distortion detection is performed on the blind hole stacking hole perspective reference image and the blind hole stacking hole perspective image to obtain a blind hole stacking hole perspective reference feature map and a blind hole stacking hole perspective detection feature map. Specifically, in an embodiment of the present application, image twin distortion detection is performed on the blind hole stacking hole perspective reference image and the blind hole stacking hole perspective image to obtain a blind hole stacking hole perspective reference feature map and a blind hole stacking hole perspective detection feature map, including: inputting the blind hole stacking hole perspective reference image and the blind hole stacking hole perspective image into an image twin distortion detector including a first image distortion feature extractor and a second image distortion feature extractor to obtain the blind hole stacking hole perspective reference feature map and the blind hole stacking hole perspective detection feature map. It should be understood that both the blind hole stacking hole perspective reference image and the blind hole stacking hole perspective image contain information about the internal structure of the circuit board, especially data about the position, shape and alignment of the blind hole, and these data information have an important influence on offset detection. In order to better capture and extract the feature information of blind hole overlapped holes in the reference image and the detection image, so as to better improve the accuracy of subsequent distortion detection, in the technical solution of the present application, the blind hole overlapped hole perspective reference image and the blind hole overlapped hole perspective image are input into an image twin distortion detector including a first image distortion feature extractor and a second image distortion feature extractor to obtain a blind hole overlapped hole perspective reference feature map and a blind hole overlapped hole perspective detection feature map. It is worth noting that the first image distortion feature extractor and the second image distortion feature extractor specifically refer to convolutional neural network models with the same structure, which can ensure that the same feature extraction algorithm and parameters are applied to the blind hole overlapped hole perspective reference image and the blind hole overlapped hole perspective image, thereby ensuring the consistency of the feature extraction process, and this consistency helps to accurately compare the distortion features between the two images.

[0035] In step S133, the blind hole overlapped hole perspective reference feature map and the blind hole overlapped hole perspective detection feature map are enhanced based on spatial perception to obtain blind hole overlapped hole perspective reference enhancement features and blind hole overlapped hole perspective detection enhancement features. Specifically, Figure 4 A flowchart of performing spatial perception-based feature enhancement on the blind hole stacking hole perspective reference feature map and the blind hole stacking hole perspective detection feature map to obtain blind hole stacking hole perspective reference enhancement feature and blind hole stacking hole perspective detection enhancement feature in a method for an HDI circuit board for an intelligent mobile communication terminal according to an embodiment of the present application. Figure 4As shown, the blind hole overlapped hole perspective reference feature map and the blind hole overlapped hole perspective detection feature map are subjected to feature enhancement based on spatial perception to obtain blind hole overlapped hole perspective reference enhancement features and blind hole overlapped hole perspective detection enhancement features, including: S1331, feature dispersion of the blind hole overlapped hole perspective reference feature map along the channel dimension to obtain a set of blind hole overlapped hole perspective reference pixel granularity feature vectors; S1332, semantic association and attenuation modulation based on soft spatial constraints on the set of blind hole overlapped hole perspective reference pixel granularity feature vectors to obtain a blind hole overlapped hole perspective reference semantic association space soft constraint score topology matrix; S1333, based on the blind hole overlapped hole perspective reference semantic association space soft constraint score topology matrix, convolution coding enhancement is performed on the set of blind hole overlapped hole perspective reference pixel granularity feature vectors to obtain the blind hole overlapped hole perspective reference enhancement features.

[0036] Accordingly, considering that both the blind hole stacking perspective reference feature map and the blind hole stacking perspective detection feature map contain important and significant feature information about the blind hole stacking, and in order to make them clearer and more prominent so as to better identify the position and shape of the blind hole, thereby improving the accuracy of subsequent distortion correction. In the technical solution of the present application, the blind hole stacking perspective reference feature map and the blind hole stacking perspective detection feature map are subjected to feature enhancement based on spatial perception to obtain a blind hole stacking perspective reference enhancement feature map as a blind hole stacking perspective reference enhancement feature and a blind hole stacking perspective detection enhancement feature map as a blind hole stacking perspective detection enhancement feature.

[0037] In particular, the blind hole stacked hole perspective reference feature map is taken as an example for detailed description. Specifically, first, the blind hole stacked hole perspective reference feature map is feature scattered along the channel dimension to obtain a set of blind hole stacked hole perspective reference pixel granularity feature vectors. That is, by feature scattered the entire feature map, the multi-channel information of each pixel in the feature map can be converted into a feature vector, thereby retaining more detail information to better capture subtle changes in the image.

[0038] Then, the set of blind hole stacked hole perspective reference pixel granularity feature vectors is subjected to semantic association and attenuation modulation based on soft space constraints to obtain a blind hole stacked hole perspective reference semantic association space soft constraint score topological matrix. In this way, it is possible to not only focus on the information of a single pixel, but also consider the connection between adjacent pixels, so as to better understand the structural information in the image, and make important features more prominent by adjusting the weight or strength of the feature vector. Specifically, first, in order to more accurately quantify the semantic similarity between any two vectors, the set of blind hole stacked hole perspective reference pixel granularity feature vectors is mapped to the Poincare space. It can be understood that the Poincare space is a non-Euclidean geometric space, in which the distance measurement method is different from the traditional Euclidean space. That is, mapping to the Poincare space can better maintain distance invariance, that is, feature vectors that maintain similar distances in high-dimensional space still maintain similar distances in low-dimensional space. Afterwards, the Poincare distance between any two mapped vectors is calculated as the semantic association score between the two to obtain the blind hole stacked hole perspective reference semantic association score topological matrix. In particular, the distance measurement method in the Poincare space (such as the Poincare distance) can better reflect the semantic similarity between feature vectors, which means that even if two feature vectors have large differences in some dimensions, as long as they are semantically related, the Poincare distance can also reflect this relationship. Then, based on the spatial distance between any two pixel feature vectors, the spatial soft constraint score topology matrix is ​​spatially attenuated modulated to obtain the blind hole overlap perspective reference semantic association spatial soft constraint score topology matrix. It should be understood that the closer the spatial distance of the pixel point, the higher its similarity score should be. Therefore, the importance of pixel features that are closer and more semantically related on adjacent pixels can be accurately determined through spatial attenuation modulation, and the importance of each position can be dynamically adjusted.

[0039] Next, based on the blind hole stacking perspective reference semantic association space soft constraint score topology matrix, the set of the blind hole stacking perspective reference pixel granularity feature vectors is convolutionally enhanced to obtain the blind hole stacking perspective reference enhancement feature. Specifically, first, the blind hole stacking perspective reference semantic association space soft constraint score topology matrix is ​​subjected to hole convolution encoding to more deeply capture and refine the semantic association between each pixel. Afterwards, the set of the score topology feature matrix and the pixel granularity heat vector obtained by convolution encoding is subjected to global semantic context association feature expression through the graph convolution coding module to obtain a set of blind hole stacking perspective reference context pixel granularity feature vectors. Finally, the set of context pixel granularity feature vectors is reshaped to obtain the final blind hole stacking perspective reference enhancement feature map.

[0040] More specifically, Figure 5The present invention is a flowchart of performing semantic association and attenuation modulation based on soft space constraints on the set of blind hole stacking hole perspective reference pixel granularity feature vectors in the method for HDI circuit board for intelligent mobile communication terminal according to the embodiment of the present application to obtain the blind hole stacking hole perspective reference semantic association space soft constraint score topology matrix. Figure 5 As shown, the set of blind hole overlap perspective reference pixel granularity feature vectors is subjected to semantic association and attenuation modulation based on soft space constraints to obtain a blind hole overlap perspective reference semantic association space soft constraint score topology matrix, including: S13321, calculating the semantic association score between any two blind hole overlap perspective reference pixel granularity feature vectors in the set of blind hole overlap perspective reference pixel granularity feature vectors to obtain a blind hole overlap perspective reference semantic association score topology matrix; S13322, based on the spatial distance between any two blind hole overlap perspective reference pixel granularity feature vectors in the set of blind hole overlap perspective reference pixel granularity feature vectors, performing spatial attention attenuation modulation on the blind hole overlap perspective reference semantic association score topology matrix to obtain the blind hole overlap perspective reference semantic association space soft constraint score topology matrix.

[0041] More specifically, in an embodiment of the present application, the semantic association score between any two blind hole overlay hole perspective reference pixel granularity feature vectors in the set of blind hole overlay hole perspective reference pixel granularity feature vectors is calculated to obtain a blind hole overlay hole perspective reference semantic association score topological matrix, including: mapping each blind hole overlay hole perspective reference pixel granularity feature vector in the set of blind hole overlay hole perspective reference pixel granularity feature vectors to the Poincare space to obtain a set of mapped blind hole overlay hole perspective reference pixel granularity feature vectors; calculating the semantic association score between any two mapped blind hole overlay hole perspective reference pixel granularity feature vectors in the set of mapped blind hole overlay hole perspective reference pixel granularity feature vectors to obtain the blind hole overlay hole perspective reference semantic association score topological matrix composed of multiple blind hole overlay hole perspective reference semantic association score values.

[0042] More specifically, in an embodiment of the present application, each blind hole overlay hole perspective reference pixel granularity feature vector in the set of blind hole overlay hole perspective reference pixel granularity feature vectors is mapped to the Poincare space to obtain a set of mapped blind hole overlay hole perspective reference pixel granularity feature vectors, including: matrix-vector multiplication of a first weight matrix and the blind hole overlay hole perspective reference pixel granularity feature vector, and then vector-matrix multiplication of the obtained blind hole overlay hole perspective reference pixel granularity weight vector and a second weight matrix to obtain the mapped blind hole overlay hole perspective reference pixel granularity feature vector.

[0043] More specifically, in the embodiment of the present application, the semantic association score between any two post-mapping blind hole overlay hole perspective reference pixel granularity feature vectors in the set of post-mapping blind hole overlay hole perspective reference pixel granularity feature vectors is calculated to obtain the blind hole overlay hole perspective reference semantic association score topological matrix composed of multiple blind hole overlay hole perspective reference semantic association score values, including: extracting the first blind hole overlay hole perspective reference pixel granularity feature vector from the set of post-mapping blind hole overlay hole perspective reference pixel granularity feature vectors; The reference pixel granularity feature vector of blind hole overlap perspective after mapping and the a blind hole overlap perspective reference pixel granularity feature vector after mapping; and calculate the first the mapped blind hole overlap perspective reference pixel granularity feature vector and the first The mapped blind hole overlap hole perspective reference pixel granularity difference vector is obtained by calculating the position difference between the mapped blind hole overlap hole perspective reference pixel granularity feature vectors, and the square of the one-norm of the mapped blind hole overlap hole perspective reference pixel granularity difference vector is calculated to obtain the mapped blind hole overlap hole perspective reference feature difference norm value; the constant one is calculated and the first The difference between the squares of the norms of the reference pixel granularity feature vectors of the blind hole overlap perspective after mapping is used to obtain the first a blind hole overlapped hole perspective reference feature difference after mapping; calculate constant one and the first The difference between the squares of the norms of the reference pixel granularity feature vectors of the blind hole overlap perspective after mapping is used to obtain the first a blind hole overlapped hole perspective reference feature difference after mapping; and calculating the first The difference between the blind hole overlap perspective reference feature after mapping and the first The product of the difference values ​​of the perspective reference features of the mapped blind holes and holes is obtained by multiplying the difference norm value of the perspective reference features of the mapped blind holes and holes by the product value of the perspective reference features of the mapped blind holes and holes. The obtained value is first multiplied by a constant two and then added to a constant one to obtain a semantic metric value of the perspective reference features of the mapped blind holes and holes. The semantic metric value of the perspective reference features of the mapped blind holes and holes is input into the arccosh function to obtain the first the mapped blind hole overlap perspective reference pixel granularity feature vector and the first The blind hole overlay perspective reference semantic association score value between the mapped blind hole overlay perspective reference pixel granularity feature vectors.

[0044] More specifically, in an embodiment of the present application, based on the spatial distance between any two blind hole overlapped hole perspective reference pixel granularity feature vectors in the set of blind hole overlapped hole perspective reference pixel granularity feature vectors, the blind hole overlapped hole perspective reference semantic association score topology matrix is ​​subjected to spatial attention attenuation modulation to obtain the blind hole overlapped hole perspective reference semantic association spatial soft constraint score topology matrix, including: calculating the spatial distance between any two blind hole overlapped hole perspective reference pixel granularity feature vectors to obtain a set of blind hole overlapped hole perspective reference spatial distance values; calculating the spatial distance between each blind hole overlapped hole perspective reference pixel granularity feature vector in the set of blind hole overlapped hole perspective reference spatial distance values ​​with the natural constant e as the base; The blind hole perspective reference space distance value is an exponential function of the exponent to obtain a set of blind hole stacking hole perspective reference space distance index values; the set of blind hole stacking hole perspective reference space distance index values ​​is multiplied by the attenuation modulation value by position to obtain a set of blind hole stacking hole perspective reference space distance weighted values; each blind hole stacking hole perspective reference semantic association score value in the blind hole stacking hole perspective reference semantic association score topological matrix is ​​divided by the corresponding blind hole stacking hole perspective reference space distance weighted value to obtain the blind hole stacking hole perspective reference semantic association space soft constraint score topological matrix composed of multiple blind hole stacking hole perspective reference semantic association space soft constraint score values.

[0045] More specifically, Figure 6 In the method for HDI circuit board for intelligent mobile communication terminal according to the embodiment of the present application, based on the blind hole stacking hole perspective reference semantic association space soft constraint score topology matrix, the set of the blind hole stacking hole perspective reference pixel granularity feature vectors is convolutionally enhanced to obtain the blind hole stacking hole perspective reference enhancement feature. Figure 6 As shown, based on the blind hole stacked hole perspective reference semantic association space soft constraint score topological matrix, the set of the blind hole stacked hole perspective reference pixel granularity feature vectors is convolutionally enhanced to obtain the blind hole stacked hole perspective reference enhancement feature, including: S13331, the blind hole stacked hole perspective reference semantic association space soft constraint score topological matrix is ​​hole convolutionally encoded to obtain the blind hole stacked hole perspective reference semantic association space soft constraint score topological feature matrix; S13332, the set of the blind hole stacked hole perspective reference pixel granularity feature vectors and the blind hole stacked hole perspective reference semantic association space soft constraint score topological feature matrix are input into a graph convolutional coding module to obtain a set of blind hole stacked hole perspective reference context pixel granularity feature vectors; S13333, the set of the blind hole stacked hole perspective reference context pixel granularity feature vectors is feature reshaped to obtain a blind hole stacked hole perspective reference enhancement feature map as the blind hole stacked hole perspective reference enhancement feature.

[0046] In the embodiment of the present application, specifically, the blind hole stacking hole perspective reference feature map is enhanced based on spatial perception, which can be expressed by the formula:

[0047]

[0048]

[0049]

[0050]

[0051]

[0052]

[0053]

[0054] in, is the blind hole stacking perspective reference characteristic diagram, and , is the set of real numbers, and Respectively represent the width and height of each feature matrix of the blind hole stacking perspective reference feature map along the channel dimension, represents the number of channels of the blind hole stacking hole perspective reference feature map, To reshape the feature map, is the first pixel in the set of blind hole stacking perspective reference pixel granularity feature vectors blind hole overlap hole perspective reference pixel granularity feature vector, and , is the set of the blind hole stacking perspective reference pixel granularity feature vectors, and are the first weight matrix and the second weight matrix respectively, is the set of reference pixel granularity feature vectors of blind hole overlap perspective after mapping, are respectively the first, second, ..., first, second, third, fourth, fifth, fifth, sixth, sixth, fifth ... ,..., ,..., The reference pixel granularity feature vector of blind hole overlap perspective after mapping, is the square of the one-norm of the vector, is the inverse hyperbolic cosine function, For calculation and The blind hole overlap hole perspective reference semantic association score between them, is the eigenvalue of each position in the blind hole stacking hole perspective reference semantic association score topology matrix, for and The blind hole overlap perspective reference space distance value between them, is the attenuation modulation value, is the eigenvalue of each position in the topological matrix of the soft constraint score of the blind hole stacking hole perspective reference semantic association space, is the soft constraint score topological matrix of the blind hole stacking perspective reference semantic association space, is the dilated convolutional coding operation, is the topological feature matrix of the blind hole stacking perspective reference semantic association space soft constraint score, is the graph convolutional encoding operation, is the set of pixel granularity feature vectors of blind hole overlap perspective reference context, is the blind hole overlap hole perspective reference enhancement feature map. Similarly, the processing method of the blind hole overlap hole perspective detection feature map is also as shown in the above formula.

[0055] In step S134, the distortion correction coefficient is obtained based on the image semantic similarity between the blind hole overlapped hole perspective reference enhancement feature and the blind hole overlapped hole perspective detection enhancement feature. Specifically, in the embodiment of the present application, the distortion correction coefficient is obtained based on the image semantic similarity between the blind hole overlapped hole perspective reference enhancement feature and the blind hole overlapped hole perspective detection enhancement feature, including: calculating the image semantic similarity between the blind hole overlapped hole perspective reference enhancement feature map and the blind hole overlapped hole perspective detection enhancement feature map; multiplying the image semantic similarity by the first distortion correction coefficient to obtain the distortion correction coefficient.

[0056] It should be understood that in order to quantify the difference between the blind hole stacked hole perspective detection and the reference image and help determine the distortion correction coefficient more accurately, in the technical solution of the present application, the image semantic similarity between the blind hole stacked hole perspective reference enhanced feature map and the blind hole stacked hole perspective detection enhanced feature map is calculated. In particular, in a specific embodiment of the present application, the semantic similarity between the two images can be represented by calculating the cosine distance between the blind hole stacked hole perspective reference enhanced feature map and the blind hole stacked hole perspective detection enhanced feature map. Then, the image semantic similarity is multiplied by the first distortion correction coefficient to obtain the distortion correction coefficient. It should be understood that the image semantic similarity reflects the similarity between the feature vectors in the blind hole stacked hole perspective detection and the reference image, and the first distortion correction coefficient is obtained based on previous experience or calibration. Based on this, in the technical solution of the present application, the image semantic similarity is multiplied by the first distortion correction coefficient to obtain the distortion correction coefficient. In this way, by combining the image semantic similarity with the first distortion correction coefficient, the semantic information of the image (such as the similarity between the feature vectors) and the known distortion correction information can be considered at the same time to more comprehensively reflect the real situation of the image, thereby improving the accuracy of the distortion correction and ensuring that the final correction result is closer to the actual situation. In this way, the measurement error caused by image distortion can be effectively reduced and the accuracy of offset measurement can be improved. By adopting image processing technology based on computer vision, not only the detection accuracy is improved, but also the intelligent offset measurement is realized, making the blind hole offset detection in complex environments more reliable and efficient.

[0057] In step S140, based on the distortion correction coefficient and the proportional relationship between the perspective image and the real object, the second offset is converted to obtain a third offset. Specifically, in an embodiment of the present application, based on the distortion correction coefficient and the proportional relationship between the perspective image and the real object, the second offset is converted to obtain a third offset, including: multiplying the second offset by the proportional relationship between the perspective image and the real object and then multiplying it by the distortion correction coefficient to obtain the third offset. It should be understood that during the perspective imaging process, there is a scaling ratio between the actual size of the object and its displayed size in the image, and this proportional relationship allows the measured value in the image to be converted into the actual physical size. If this proportional relationship is not considered, directly using the offset in the image may cause the calculated offset to be inconsistent with the actual. By multiplying the second offset by the proportional relationship between the perspective image and the real object, the offset in the image can be converted into the offset in the real object. This ensures that the calculated offset is closer to the actual situation in the real object. The introduction of the distortion correction coefficient is to eliminate factors such as nonlinear deformation in the image, so that the calculation of the offset is more accurate. In this way, scale conversion and distortion correction can be comprehensively considered to obtain a more accurate third offset.

[0058] In particular, when the blind hole-on-hole perspective reference feature map and the blind hole-on-hole perspective detection feature map represent the image distortion semantic features of the blind hole-on-hole perspective reference image and the blind hole-on-hole perspective image, respectively, when performing feature enhancement based on spatial soft constraints and mapping them to the common image semantic similarity space, the blind hole-on-hole perspective enhancement feature map and the blind hole-on-hole perspective detection enhancement feature map may also cause soft constraint enhancement diversity due to the difference in image semantic space distribution, affect the iterative consistency of the common space mapping, and reduce the calculation accuracy of image semantic similarity.

[0059] Preferably, calculating the image semantic similarity between the blind hole overlapped hole perspective reference enhancement feature map and the blind hole overlapped hole perspective detection enhancement feature map comprises: respectively expanding the blind hole overlapped hole perspective reference enhancement feature map and the blind hole overlapped hole perspective detection enhancement feature map into a blind hole overlapped hole perspective reference enhancement feature vector and a blind hole overlapped hole perspective detection enhancement feature vector, and further cascading them into a blind hole overlapped hole perspective reference-detection enhancement cascade feature vector; calculating the product of a first blind hole overlapped hole perspective reference-detection enhancement cascade feature value of the blind hole overlapped hole perspective reference-detection enhancement cascade feature vector and the image semantic similarity to obtain a first blind hole overlapped hole perspective reference-detection enhancement cascade feature probability product. ; Calculate the product of the second blind hole overlap hole perspective reference-detection enhancement cascade feature value of the blind hole overlap hole perspective reference-detection enhancement cascade feature vector and the difference between one and the image semantic similarity to obtain a second blind hole overlap hole perspective reference-detection enhancement cascade feature inverse probability product; Calculate the quotient of the first blind hole overlap hole perspective reference-detection enhancement cascade feature value divided by the difference between one and the image semantic similarity to obtain a first blind hole overlap hole perspective reference-detection enhancement cascade feature inverse probability quotient; Calculate the quotient of the second blind hole overlap hole perspective reference-detection enhancement cascade feature value and the image semantic similarity to obtain a second blind hole overlap hole perspective reference-detection enhancement cascade feature probability quotient; After adding the first blind hole overlap hole perspective reference-detection enhancement cascade feature probability product and the second blind hole overlap hole perspective reference-detection enhancement cascade feature inverse probability product, the difference between the first blind hole overlap hole perspective reference-detection enhancement cascade feature inverse probability quotient and the second blind hole overlap hole perspective reference-detection enhancement cascade feature probability quotient is subtracted to obtain a blind hole overlap hole perspective reference-detection enhancement cascade correction feature value; for each first blind hole overlap hole perspective reference-detection enhancement cascade feature value and second blind hole overlap hole perspective reference-detection enhancement cascade feature value of the blind hole overlap hole perspective reference-detection enhancement cascade feature vector, the blind hole overlap hole perspective reference-detection enhancement cascade correction feature value is composed of The blind hole overlap hole perspective reference-detection enhancement cascade correction feature matrix is ​​matrix-multiplied with the blind hole overlap hole perspective reference-detection enhancement cascade feature vector to obtain a corrected blind hole overlap hole perspective reference-detection enhancement cascade feature vector; the corrected blind hole overlap hole perspective reference-detection enhancement cascade feature vector is split into a corrected blind hole overlap hole perspective reference enhancement feature vector and a corrected blind hole overlap hole perspective detection enhancement feature vector, and restored to a corrected blind hole overlap hole perspective reference enhancement feature map and a corrected blind hole overlap hole perspective detection enhancement feature map; the image semantic similarity between the corrected blind hole overlap hole perspective reference enhancement feature map and the corrected blind hole overlap hole perspective detection enhancement feature map is calculated.

[0060] The correction process of the blind hole overlapped hole perspective reference-detection enhanced cascade feature vector is expressed as:

[0061]

[0062]

[0063] in, is the blind hole overlap perspective reference-detection enhanced cascade feature vector, is the image semantic similarity, and They represent the first The first blind hole overlap perspective reference-detection enhanced cascade feature value and the first The second blind hole overlapped hole perspective reference-detection enhanced cascade feature value at the position, represents the first in the blind hole overlap perspective reference-detection enhanced cascade correction feature matrix Blind hole overlap perspective reference-detection enhanced cascade correction feature value at each position, is the blind hole overlap perspective reference-detection enhanced cascade correction feature matrix, represents matrix multiplication, A cascade feature vector is provided for the corrected blind hole overlap perspective reference-detection enhancement.

[0064] That is, the present application is directed to the region-boundary integral relationship of the joint high-dimensional feature manifold of the blind hole stacked hole perspective reference enhancement feature map and the blind hole stacked hole perspective detection enhancement feature map, and by defining the common image semantic similarity space mapping probability distribution of the feature set of the blind hole stacked hole perspective reference enhancement feature map and the blind hole stacked hole perspective detection enhancement feature map as a manifold constraint boundary, to approximate the high-dimensional manifold of the feature set of the blind hole stacked hole perspective reference enhancement feature map and the blind hole stacked hole perspective detection enhancement feature map composed of eigenvalue pairs. The single connected region representation, so as to avoid the mapping ambiguity of the diversified feature representation of the blind hole stacked hole perspective reference enhancement feature map and the blind hole stacked hole perspective detection enhancement feature map to the local manifold representation in the high-dimensional feature convergence space, improve the execution iteration consistency of each local feature distribution in the mapping task, so as to improve the feature mapping effect and improve the calculation accuracy of the image semantic similarity. In this way, the measurement error caused by image distortion can be effectively reduced, and the accuracy of offset measurement can be improved. By adopting image processing technology based on computer vision, not only the detection accuracy is improved, but also the intelligent offset measurement is realized, making blind hole offset detection in complex environments more reliable and efficient.

[0065] In summary, a method for an HDI circuit board for an intelligent mobile communication terminal based on an embodiment of the present application is explained, which uses image detection and analysis technology based on computer vision to extract image distortion features and enhance spatial features of a blind hole stacking hole perspective reference image extracted from a database and a blind hole stacking hole perspective image obtained by an X-ray detector, thereby intelligently obtaining a distortion correction coefficient based on the semantic metric between the enhanced blind hole stacking hole perspective reference feature and the blind hole stacking hole perspective detection feature. In this way, the measurement error caused by image distortion can be effectively reduced and the accuracy of offset measurement can be improved. By adopting image processing technology based on computer vision, not only the detection accuracy is improved, but also the intelligent offset measurement is realized, making blind hole offset detection in complex environments more reliable and efficient.

[0066] In the embodiment of the present application, there is also provided an HDI circuit board for a smart mobile communication terminal, wherein the HDI circuit board for a smart mobile communication terminal is manufactured by the method for manufacturing an HDI circuit board for a smart mobile communication terminal as described above. Figures 1 to 6 It has been described in detail in the description of , so its repeated description will be omitted.

[0067] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details of the above application are only for the purpose of illustration and ease of understanding, rather than limitation, and the above details do not limit the present application to the use of the above specific details to be implemented.

Claims

1. A method for an HDI circuit board for an intelligent mobile communication terminal, comprising: Arrange a plurality of laser blind hole arrays outside the working area of ​​each layer of the circuit board, and ensure that the position and size of the laser blind hole arrays on each layer are consistent; The layers of the circuit board are pressed together to form blind holes; the blind holes are detected by an X-ray detector to obtain a blind hole perspective image; based on the blind hole perspective image, the offset of the blind hole of the circuit board is obtained, and the blind hole of the circuit board is formally drilled based on the offset; characterized in that the offset of the blind hole of the circuit board is obtained based on the blind hole perspective image, including: Acquire the blind hole stacking hole perspective image; Determine the distance between the centers of any two blind hole objects in the blind hole stacked hole perspective image based on machine vision to obtain a second offset; Based on the blind hole-on-hole perspective image, determining the distortion correction coefficient includes: extracting a blind hole-on-hole perspective reference image annotated with a first distortion correction coefficient from a database; performing image twin distortion detection on the blind hole-on-hole perspective reference image and the blind hole-on-hole perspective image to obtain a blind hole-on-hole perspective reference feature map and a blind hole-on-hole perspective detection feature map; performing feature enhancement based on spatial perception on the blind hole-on-hole perspective reference feature map and the blind hole-on-hole perspective detection feature map to obtain a blind hole-on-hole perspective reference enhancement feature and a blind hole-on-hole perspective detection enhancement feature; and obtaining the distortion correction coefficient based on the image semantic similarity between the blind hole-on-hole perspective reference enhancement feature and the blind hole-on-hole perspective detection enhancement feature. Based on the distortion correction coefficient and the proportional relationship between the perspective image and the real object, converting the second offset to obtain a third offset; The blind hole overlap hole perspective reference feature map and the blind hole overlap hole perspective detection feature map are subjected to feature enhancement based on space perception to obtain a blind hole overlap hole perspective reference enhancement feature and a blind hole overlap hole perspective detection enhancement feature, including: Performing feature dispersion on the blind hole stacking perspective reference feature map along the channel dimension to obtain a set of blind hole stacking perspective reference pixel granularity feature vectors; Performing semantic association and attenuation modulation based on soft space constraints on the set of blind hole overlap perspective reference pixel granularity feature vectors to obtain a blind hole overlap perspective reference semantic association space soft constraint score topology matrix; Based on the blind hole overlap perspective reference semantic association space soft constraint score topology matrix, a set of the blind hole overlap perspective reference pixel granularity feature vectors is convolutionally enhanced to obtain the blind hole overlap perspective reference enhancement feature.

2. The method for HDI circuit board for intelligent mobile communication terminal according to claim 1, characterized in that: Performing image twin distortion detection on the blind hole-on-hole perspective reference image and the blind hole-on-hole perspective image to obtain a blind hole-on-hole perspective reference feature map and a blind hole-on-hole perspective detection feature map, comprising: inputting the blind hole-on-hole perspective reference image and the blind hole-on-hole perspective image into an image twin distortion detector including a first image distortion feature extractor and a second image distortion feature extractor to obtain the blind hole-on-hole perspective reference feature map and the blind hole-on-hole perspective detection feature map.

3. The method for HDI circuit board for intelligent mobile communication terminal according to claim 2, characterized in that: The set of blind hole overlapped hole perspective reference pixel granularity feature vectors is subjected to semantic association and attenuation modulation based on soft space constraints to obtain a blind hole overlapped hole perspective reference semantic association space soft constraint score topology matrix, including: Calculating the semantic association score between any two blind hole overlap perspective reference pixel granularity feature vectors in the set of blind hole overlap perspective reference pixel granularity feature vectors to obtain a blind hole overlap perspective reference semantic association score topological matrix; Based on the spatial distance between any two blind hole overlay perspective reference pixel granularity feature vectors in the set of blind hole overlay perspective reference pixel granularity feature vectors, the blind hole overlay perspective reference semantic association score topology matrix is ​​spatially attended attenuated and modulated to obtain the blind hole overlay perspective reference semantic association spatial soft constraint score topology matrix.

4. The method for HDI circuit board for intelligent mobile communication terminal according to claim 3, characterized in that: Calculating the semantic association score between any two blind hole overlap perspective reference pixel granularity feature vectors in the set of blind hole overlap perspective reference pixel granularity feature vectors to obtain a blind hole overlap perspective reference semantic association score topological matrix, including: Mapping each blind hole overlay perspective reference pixel granularity feature vector in the set of blind hole overlay perspective reference pixel granularity feature vectors to the Poincare space to obtain a set of mapped blind hole overlay perspective reference pixel granularity feature vectors; The semantic association score between any two mapped blind hole overlay perspective reference pixel granularity feature vectors in the set of mapped blind hole overlay perspective reference pixel granularity feature vectors is calculated to obtain the blind hole overlay perspective reference semantic association score topological matrix composed of multiple blind hole overlay perspective reference semantic association score values.

5. The method for HDI circuit board for intelligent mobile communication terminal according to claim 4, characterized in that: The method of performing spatial attention attenuation modulation on the blind hole overlap perspective reference semantic association score topology matrix based on the spatial distance between any two blind hole overlap perspective reference pixel granularity feature vectors in the set of blind hole overlap perspective reference pixel granularity feature vectors to obtain the blind hole overlap perspective reference semantic association spatial soft constraint score topology matrix comprises: Calculating the spatial distance between any two blind hole overlap perspective reference pixel granularity feature vectors to obtain a set of blind hole overlap perspective reference spatial distance values; Calculate an exponential function with the natural constant e as the base, with each blind hole stacking hole perspective reference space distance value in the set of blind hole stacking hole perspective reference space distance values ​​as an exponent, to obtain a set of blind hole stacking hole perspective reference space distance index values; Multiplying the set of blind hole stacking hole perspective reference space distance index values ​​by the attenuation modulation value according to the position to obtain a set of blind hole stacking hole perspective reference space distance weighted values; Each blind hole stacking hole perspective reference semantic association score value in the blind hole stacking hole perspective reference semantic association score topological matrix is ​​divided by the corresponding blind hole stacking hole perspective reference space distance weighted value to obtain the blind hole stacking hole perspective reference semantic association space soft constraint score topological matrix composed of multiple blind hole stacking hole perspective reference semantic association space soft constraint score values.

6. The method for HDI circuit board for intelligent mobile communication terminal according to claim 5, characterized in that: Based on the blind hole overlapped hole perspective reference semantic association space soft constraint score topology matrix, a set of the blind hole overlapped hole perspective reference pixel granularity feature vectors is convolutionally enhanced to obtain the blind hole overlapped hole perspective reference enhancement feature, including: Performing dilated convolution coding on the blind hole stacking hole perspective reference semantic association space soft constraint score topology matrix to obtain a blind hole stacking hole perspective reference semantic association space soft constraint score topology feature matrix; Input the set of blind hole overlapped hole perspective reference pixel granularity feature vectors and the blind hole overlapped hole perspective reference semantic association space soft constraint score topological feature matrix into a graph convolutional coding module to obtain a set of blind hole overlapped hole perspective reference context pixel granularity feature vectors; The set of the blind hole-over-hole perspective reference context pixel granularity feature vectors is reshaped to obtain a blind hole-over-hole perspective reference enhancement feature map as the blind hole-over-hole perspective reference enhancement feature.

7. The method for HDI circuit board for intelligent mobile communication terminal according to claim 6, characterized in that: The distortion correction coefficient is obtained based on the image semantic similarity between the blind hole overlapped hole perspective reference enhancement feature and the blind hole overlapped hole perspective detection enhancement feature, including: Calculating the image semantic similarity between the blind hole overlapped hole perspective reference enhancement feature map and the blind hole overlapped hole perspective detection enhancement feature map; The image semantic similarity is multiplied by the first distortion correction coefficient to obtain the distortion correction coefficient.

8. The method for HDI circuit board for intelligent mobile communication terminal according to claim 7, characterized in that: Based on the distortion correction coefficient and the proportional relationship between the perspective image and the real object, the second offset is converted to obtain a third offset, including: multiplying the second offset by the proportional relationship between the perspective image and the real object and then multiplying by the distortion correction coefficient to obtain the third offset.

9. An HDI circuit board for intelligent mobile communication terminals, characterized in that: The HDI circuit board for smart mobile communication terminals is manufactured by the method for manufacturing an HDI circuit board for smart mobile communication terminals as claimed in any one of claims 1 to 8.

Images