A method for automatic matching of components and pads

Abstract

This invention relates to the field of component pad matching technology, and in particular to a method for automatic matching of components and pads. The technical solution includes: dynamically calculating the coverage area directly based on component size; real-time detection of pad layers without the need for pre-stored mapping relationships; adding electrical feasibility verification that the number of pads is greater than or equal to the number of pins to address the risk of cold solder joints or missing solder joints; coordinate correction using the average value of the matching pad center point to correct coordinates; angle correction using AI to identify silkscreen polarity markings to achieve a two-stage correction technology; and adding polarity marking consistency verification to ensure the correct installation of high-reliability components and guarantee polarity direction. This not only reduces the workload of customers pre-storing templates but also improves matching accuracy through dynamic detection and closed-loop correction.

Description

Technical Field

[0001] This invention relates to the field of component pad matching technology, and in particular to a method for automatic matching of components and pads. Background Technology

[0002] The role of component pad matching is mainly reflected in three aspects: ensuring soldering reliability, improving assembly efficiency, and maintaining circuit performance. Mismatch between pad and component pin size can lead to insufficient overlap or overload, resulting in displacement or short circuit, and may also cause cold solder joints.

[0003] Current methods for matching pads and components involve pre-storing the mapping relationship between pads and components, then finding the pre-stored pad information based on the component, and finally comparing it with the actual current pad information. For customers, this requires pre-storing the mapping relationship between pads and components, which increases their workload in actual use. Furthermore, if the pre-stored information is incorrect, it will reduce the matching accuracy, increase the risk of cold solder joints or missing solder joints, and also reduce the correctness of installation.

[0004] Therefore, we propose a method for automatic matching of components and pads to solve the existing problems. Summary of the Invention

[0005] The purpose of this invention is to provide a method for automatic matching of components and pads, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for automatic matching of components and pads, comprising:

[0007] S1: Obtain the size information of the components and calculate the coverage area of ​​the components on the circuit board based on the size information;

[0008] S2: Detect all pads within the coverage area in the pad layer to form a candidate pad set;

[0009] S3: Determine whether the number of pads in the candidate pad set is greater than or equal to the number of pins of the component; if yes, the candidate pad set is successfully matched with the component; if no, the match fails.

[0010] S4: Perform at least one of coordinate correction and angle correction on the successfully matched components.

[0011] Furthermore, in coordinate correction, the average center coordinates are calculated based on the center point coordinates of each pad in the candidate pad set, and the mounting coordinates of the components are corrected accordingly.

[0012] Furthermore, in angle correction, the actual polarity direction of the component is determined by identifying the polarity mark in the silkscreen layer using AI, and the mounting angle of the component is rotated to align with the polarity mark.

[0013] Furthermore, the steps for determining whether the number of pads matches include: if the number of pads equals the number of pins, then directly bind the candidate pad set to the component; if the number of pads is greater than the number of pins, then select the pad set with the highest overlap with the pin position for binding.

[0014] Furthermore, the coordinate correction steps specifically include: calculating the average coordinate value of the center points of all pads in the candidate pad set; and shifting the component's mounting coordinates to the position of the average coordinate value.

[0015] Furthermore, the angle correction steps specifically include: identifying the polarity marking direction associated with the component in the silkscreen layer; calculating the deviation value between the polarity marking direction and the current angle of the component; and rotating the mounting angle of the component according to the deviation value.

[0016] Furthermore, for polarized components, after angle correction, the process also includes: verifying whether the polarity direction of the component is consistent with the polarity marking direction of the pad; if they are inconsistent, a polarity error alarm is output.

[0017] Furthermore, the coverage area is a polygonal region centered on the component coordinates and covering its physical boundary.

[0018] Furthermore, after determining that the matching has failed, the coverage area is expanded by a preset step size, and the pads are re-detected until the matching is successful or the maximum expansion threshold is reached.

[0019] Furthermore, the acquisition of component size information includes: identifying the physical boundaries of the component body through computer vision, and extracting the length, width and contour features; the calculation of the coverage area includes: constructing an initial rectangular coverage area with the actual installation coordinates of the component as the center, combined with the length and width dimensions, and expanding it into a polygonal coverage area when the contour features are irregular shapes.

[0020] Compared with the prior art, the beneficial effects of the present invention are:

[0021] This invention dynamically calculates the coverage area based on component dimensions and detects the pad layer in real time, eliminating the need for pre-stored mapping relationships. It adds electrical feasibility verification for pad numbers being greater than or equal to pin numbers, mitigating the risk of cold solder joints or missing solder joints. Coordinate correction uses the average value of the pad center point to correct coordinates, and angle correction uses AI to recognize silkscreen polarity markings, achieving a two-stage correction technology. A new polarity marking consistency verification ensures the correct installation of high-reliability devices, guaranteeing polarity direction. This not only reduces the workload of customers pre-storing templates but also improves matching accuracy through dynamic detection and closed-loop correction. Attached Figure Description

[0022] Figure 1 This is a schematic diagram illustrating the workflow of a method for automatic matching of components and pads according to the present invention. Detailed Implementation

[0023] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0024] The following is a description of the method in Example 1:

[0025] like Figure 1 As shown, a method for automatically matching components with pads includes:

[0026] S1: Obtain the size information of the components and calculate the coverage area of ​​the components on the circuit board based on the size information; the coverage area is a polygonal area centered on the component coordinates and covering its physical boundary;

[0027] S2: Detect all pads within the coverage area in the pad layer to form a candidate pad set;

[0028] S3: Determine whether the number of pads in the candidate pad set is greater than or equal to the number of pins of the component; if yes, determine that the candidate pad set and the component are successfully matched; if no, determine that the match is unsuccessful; after determining that the match is unsuccessful, expand the coverage area by a preset step size and re-detect the pads until the match is successful or the maximum expansion threshold is reached.

[0029] S4: Perform at least one of coordinate correction and angle correction on the successfully matched components; in coordinate correction, calculate the average center coordinate based on the center point coordinates of each pad in the candidate pad set, and use this to correct the component's mounting coordinates; in angle correction, use AI to identify the polarity mark in the silkscreen layer, determine the actual polarity direction of the component, and rotate the component's mounting angle to align with the polarity mark.

[0030] The steps to determine whether the number of pads matches the number of pins include: if the number of pads equals the number of pins, then directly bind the candidate pad set to the component; if the number of pads is greater than the number of pins, then select the pad set with the highest overlap with the pin position for binding.

[0031] The coordinate correction steps specifically include: calculating the average coordinate value of the center points of all pads in the candidate pad set; and shifting the component mounting coordinates to the position of the average coordinate value.

[0032] The specific steps of angle correction include: identifying the polarity marking direction associated with the component in the silkscreen layer; calculating the deviation between the polarity marking direction and the current angle of the component; and rotating the component's mounting angle according to the deviation value.

[0033] For polarized components, after angle correction, the following steps are also included: verifying whether the polarity direction of the component is consistent with the polarity marking direction of the pad; if they are inconsistent, a polarity error alarm is output.

[0034] The working principle of the automatic component and pad matching method based on Embodiment 1 is as follows:

[0035] The acquisition of component size information includes: identifying the physical boundaries of the component body through computer vision, and extracting the length, width and contour features; the calculation of the coverage area includes: constructing an initial rectangular coverage area with the actual installation coordinates of the component as the center, combined with the length and width dimensions, and expanding it into a polygonal coverage area when the contour features are irregular shapes.

[0036] The steps for obtaining dimensions specifically include: performing edge detection on the component body image to generate the minimum bounding rectangle; when the body has protrusion or concavity features, calculating the convex hull polygon based on the contour point set and recording the vertex coordinates.

[0037] The specific steps for calculating the coverage area include: if the contour feature is a rectangle, generating a coverage area that coincides with the smallest bounding rectangle; if the contour feature is a polygon other than a rectangle, generating a minimum area convex polygon coverage area that includes all convex hull vertices.

[0038] The formation of the candidate pad set includes pad coarse screening, topology relationship construction, and hierarchical clustering. Pad coarse screening involves identifying all conductive patterns in the pad layer within the component coverage area and extracting patterns that meet area thresholds and shape rules as the initial pad set. Topology relationship construction involves calculating the spatial topological relationships between pads in the initial pad set and generating a topology graph containing distance and azimuth matrices. Hierarchical clustering involves using the DBSCAN algorithm to cluster the pads based on the topology graph, outputting pad clusters that satisfy the component pin distribution rules as the candidate pad set.

[0039] The specific steps for constructing the topology relationship include: establishing a polar coordinate system with the center of the coverage area as the origin; calculating the polar radius ρ and polar angle θ of each pad center point to generate the azimuth matrix: M angle [i][j]=|θ i -θ j |, i and j are pad indices, and ρ i <ρ j Generate distance matrix: M dist [i][j]= x i With y i x represents the x and y coordinates corresponding to pad index i. j With y j This represents the horizontal and vertical coordinate values ​​corresponding to pad index j.

[0040] The hierarchical clustering process specifically includes a first-level clustering and a second-level filtering. The first-level clustering takes the distance matrix as input and sets the neighborhood radius to 1.5 times the pin spacing, grouping pads with a spacing smaller than the neighborhood radius into the same cluster. The second-level filtering calculates the standard deviation of the azimuth matrix in each cluster. If the standard deviation is greater than the angle tolerance threshold, the cluster is split until each sub-cluster satisfies the standard deviation being less than or equal to the threshold.

[0041] The hierarchical clustering steps also include pad type verification: extracting the contour features of each pad in the candidate pad set; if the component is a polarized component, identifying the pads containing polarity identifiers and marking them as reference pads; verifying whether the relative positions of the reference pads and the other pads conform to the pin distribution template of the component.

[0042] The dynamic generation method of the pin distribution template includes: counting the center coordinates of all pads in the candidate pad set; fitting the minimum bounding polygon of the pad center point; comparing the number of vertices of the polygon with the number of component pins, and if they match, generating a distribution template based on the vertices.

[0043] Before the pad coarse screening step, grayscale difference enhancement processing is also included. The operations on the pad layer image include: extracting the channel with the maximum contrast between the pad and the substrate in the RGB channels; performing histogram equalization on the channel to enhance the contrast of the pad edge; and using adaptive threshold segmentation to obtain a binarized pad mask.

[0044] Coordinate correction includes pad weight allocation, weighted centroid calculation, outlier robustness handling, and coordinate offset execution. Pad weight allocation calculates weight factors based on the geometric attributes of each pad in the candidate pad set, including pad area, shape regularity, and position confidence. Weighted centroid calculation calculates the correction reference point coordinates using a weighted average algorithm based on the center coordinates of each pad and its weight factors. Outlier robustness handling uses an iterative reweighted least squares method to remove pads whose offset from the reference point exceeds a threshold, and recalculates the weighted centroid until convergence. Coordinate offset execution corrects the component mounting coordinates to the final reference point position.

[0045] The specific steps for pad weight allocation include: calculating the area weighting factor W. area =(A i / A max ) α , where A i Given the current pad area, A max Given the largest pad area in the candidate set, α = 0.5-1.0; calculate the shape regularity factor W. shape =1-|1-[Pi 2 / (4πAi)]|, where P i W represents the circumference of a circular pad. shape =1; Overall weight W i=β·W area +γ·W shape , where β+γ=1, and β>γ.

[0046] The robust outlier handling steps include: initializing the weights of all pads to W. i ; Calculate the weighted centroid coordinates (μ) x ,μ y ): μ x =Σ(W i ·x i ) / ΣW i μ y =Σ(W i ·y i ) / ΣW i Calculate the Euclidean distance di from each pad to the centroid; if di > k·σ d (σ) d If the distance is the standard deviation (k=2.0-3.0), then reset the weight of the pad to zero; repeat the above process until no pads are removed or the maximum number of iterations is reached.

[0047] The location confidence level is obtained by performing subpixel edge detection on the pad image and fitting the standard deviation of the edge points as the location fluctuation σ. edge Location confidence level: C pos =1 / (1+σ edge / δ), where δ is the maximum allowable edge fluctuation threshold; add a confidence term W to the weighting factors. i '=W i ·C pos .

[0048] Angle correction includes multimodal feature extraction, feature fusion, and progressive angle optimization. The multimodal feature extraction step includes: simultaneously acquiring image data of the silkscreen layer, solder mask layer, and pad layer; extracting the contour direction and azimuth of the polarity identifier on the silkscreen layer, the asymmetric geometric features of the windowed area on the solder mask layer, and the size difference ratio between polar and non-polar pads on the pad layer. The feature fusion step includes: fusing the three types of features through an attention-weighted network to output the theoretical polarity direction vector of the component. The progressive angle optimization step includes: starting from the current mounting angle, iteratively rotating and searching within a ±180° range according to the golden ratio; calculating the cosine similarity between the actual polarity direction and the theoretical vector at each angle; and using the angle with the highest similarity as the correction output.

[0049] The feature fusion steps include constructing a dual-channel attention mechanism and outputting a theoretical polarity direction vector. In constructing the dual-channel attention mechanism, spatial attention generates a heatmap of the polarity symbol contours of the silkscreen layer, focusing on high-gradient regions; channel attention assigns weights to the three types of features, with the silkscreen layer's weight W... silkscreen=0.5, the weight W of the solder mask layer soldermask =0.3, pad layer weight Wpad=0.2. Output theoretical polarity direction vector V. theory =Σ(W i ·V i ) / ||Σ(W i ·V i )||, where V i This is the normalized direction vector for the features of each layer.

[0050] The incremental angle optimization steps include: initialization, golden section iteration, and outputting the globally optimal angle θ. opt In the initialization, the angle search interval is set to [θ]. min ,θ max [[Current angle - 180°, Current angle + 180°]] = [Current angle - 180°, Current angle + 180°]. The steps of the golden section iteration include: calculating the division point θ. a =θ max -(θ max -θ min ) / φ,θ b =θ min +(θ max -θ min ) / φ, where φ is the golden ratio of 1.618; calculate θ a θ b The cosine similarity at point S(θ) = cos(V) actual (θ),V theory ); retain the sub-intervals corresponding to higher values ​​of S(θ); repeat the iteration until the interval length is <0.1° or 10 iterations are reached.

[0051] Progressive angle optimization also includes a pad position verification step: at θ opt At an angle, calculate the average distance D between the component pin projection point and the center of the candidate pad. avg If D avg If the pin diameter is greater than 50%, conflict handling is activated. Activating conflict handling includes: in [θ] opt -15°, θ opt The interval [+15°] is scanned in 1° increments; the selected angles that simultaneously meet the condition are: S(θ) > 0.95 and D avg (θ)≤30% of the pin diameter.

[0052] The extraction of asymmetric geometric features includes: identifying the centroid of the solder mask window region; calculating the major axis direction of the second-order moment ellipse of the window region; and determining the existence of polarity features when the major axis eccentricity is >0.7 and the centroid deviates from the geometric center.

[0053] Angle correction also includes a deep learning enhancement step: a pre-trained polarity recognition model is used to process the blurred silkscreen image, inputting a multi-scale pyramid image of the ROI region of the silkscreen layer, and outputting the polarity symbol category (triangle / horizontal bar / dot) and orientation confidence; when the orientation confidence > 0.9, the model output orientation is directly used as V. silkscreen .

[0054] The polarity consistency verification method includes spatiotemporal synchronous acquisition, dual-stream feature extraction, and adaptive decision-making. Spatiotemporal synchronous acquisition involves continuously capturing multiple frames of relative position images between components and pads using a high-speed camera during the placement head's movement. In dual-stream feature extraction, the spatial stream extracts the component polarity identification orientation angle θ in the current frame. component and pad polarity marking direction angle θ pad Temporal flow analysis of θ in N consecutive frames (N≥3) component With θ pad The trend of angular difference change. Adaptive decision-making is based on the spatial deviation Δθ = |θ component -θ pad |and the time-flow stability coefficient K t Execute a tiered response: if Δθ ≤ threshold θ th And K t If ≥0.8, the polarity is considered consistent; if Δθ>θ th But K t If Δθ > 0.9, it is determined to be transient occlusion and re-verification is initiated; if Δθ > θ th And K t If the value is less than 0.6, a real-time emergency stop alarm will be triggered.

[0055] Time-flow stability coefficient K t The calculation includes: calculating the mean μ of the angle difference between consecutive frames. Δθ and standard deviation σ Δθ Stability coefficient K t =exp(-σ Δθ / μ Δθ ).

[0056] threshold θ th The dynamic calculation includes: the base threshold θ base =arctan(pin diameter / pad center distance); speed compensation term θ vel =k·v·Δt, where v is the mounting head speed, Δt is the image sampling interval, and k is a coefficient; the final threshold θ th =θ base +θ vel .

[0057] The re-verification operation includes: controlling the placement head to rise to a safe height; starting multi-angle scanning, rotating the component around the Z-axis in 10° steps, and acquiring images in 12 directions; constructing a polarity direction probability distribution map, and taking the direction with the highest probability as the final polarity direction.

[0058] The re-verification operation also includes a deep learning arbitration step: when the graded response result is questionable, multiple frames of images are input into a pre-trained polar arbitration network; the network outputs a three-dimensional arbitration vector [consistency probability, inconsistency probability, uncertainty]; if the consistency probability is greater than 0.7 and the uncertainty is less than 0.2, the original judgment result is overwritten.

[0059] The polarity arbitration network adopts a spatiotemporal fusion architecture: in the spatial branch, 3D-ResNet extracts single-frame spatial features; in the temporal branch, LSTM encodes the temporal relationship of consecutive frames; and in the fusion layer, spatial-temporal features are weighted through an attention mechanism.

[0060] The above specific embodiments are merely several preferred embodiments of the present invention. Based on the technical solutions of the present invention and the relevant teachings of the above embodiments, those skilled in the art can make various alternative improvements and combinations to the above specific embodiments.

Claims

1. A method for automatically matching components with pads, characterized in that, include: S1: Obtain the size information of the components and calculate the coverage area of ​​the components on the circuit board based on the size information; S2: Detect all pads within the coverage area in the pad layer to form a candidate pad set; S3: Determine whether the number of pads in the candidate pad set is greater than or equal to the number of pins of the component; if yes, determine that the candidate pad set and the component are successfully matched; if no, determine that the match is unsuccessful; after determining that the match is unsuccessful, expand the coverage area by a preset step size and re-detect the pads until the match is successful or the maximum expansion threshold is reached. S4: Perform at least one of coordinate correction and angle correction on the successfully matched components; in coordinate correction, calculate the average center coordinate based on the center point coordinates of each pad in the candidate pad set, and use this to correct the component's mounting coordinates; in angle correction, use AI to identify the polarity mark in the silkscreen layer, determine the actual polarity direction of the component, and rotate the component's mounting angle to align with the polarity mark.

2. The method for automatic matching of components and pads according to claim 1, characterized in that, The steps to determine whether the number of pads matches the number of pins include: if the number of pads equals the number of pins, then directly bind the candidate pad set to the component; if the number of pads is greater than the number of pins, then select the pad set with the highest overlap with the pin position for binding.

3. The method for automatic matching of components and pads according to claim 1, characterized in that, The coordinate correction steps specifically include: calculating the average coordinate value of the center points of all pads in the candidate pad set; and shifting the component mounting coordinates to the position of the average coordinate value.

4. The method for automatic matching of components and pads according to claim 1, characterized in that, The specific steps of angle correction include: identifying the polarity marking direction associated with the component in the silkscreen layer; calculating the deviation between the polarity marking direction and the current angle of the component; and rotating the component's mounting angle according to the deviation value.

5. The method for automatic matching of components and pads according to claim 4, characterized in that, For polarized components, after angle correction, the following steps are also included: verifying whether the polarity direction of the component is consistent with the polarity marking direction of the pad; if they are inconsistent, a polarity error alarm is output.

6. The method for automatic matching of components and pads according to claim 1, characterized in that: The coverage area is a polygonal region centered on the component's coordinates and covering its physical boundaries.

7. The method for automatic matching of components and pads according to claim 1, characterized in that, The acquisition of component size information includes: identifying the physical boundaries of the component body through computer vision, and extracting the length, width and contour features; the calculation of the coverage area includes: constructing an initial rectangular coverage area with the actual installation coordinates of the component as the center, combined with the length and width dimensions, and expanding it into a polygonal coverage area when the contour features are irregular shapes.

Images