Detection method of chip gold thread

Abstract

The invention discloses a detection method of a chip gold thread. The detection method comprises the following steps: acquiring and preprocessing an image; gold thread candidate region extraction; identifying a starting point and an ending point; performing gold thread track geometric analysis; and defect judgment and output marking: outputting final abnormal information according to a trajectory analysis result. According to the method, the geometric continuity and the starting and ending point connection constraint of the gold thread are fully considered, the gold thread candidate region is extracted by adopting image preprocessing, the actual trajectory of the gold thread is accurately reconstructed in combination with a specially optimized A * path algorithm, and then the anomaly recognition is realized through trajectory analysis.

Description

Technical Field

[0001] This invention relates to the field of semiconductor testing technology, and in particular to a method for testing gold wires on a chip. Background Technology

[0002] Wire bonding connects chip pads to package pins using gold, aluminum, or copper wires, creating a conductive path. During this process, factors such as the wire trace path, connection integrity, and spacing consistency directly impact the chip's electrical performance and reliability. After manufacturing, the gold wires require inspection. Currently, most companies still rely on manual inspection, where operators visually inspect each wire under a microscope. However, this method is inefficient, cannot meet high-volume production demands, has subjective and inconsistent inspection standards, leads to operator fatigue, high error rates, and is highly dependent on training and experience, resulting in high personnel replacement costs. Summary of the Invention

[0003] According to an embodiment of the present invention, a method for detecting gold wires in a chip is provided, comprising the following steps: Image acquisition and preprocessing: A high-precision industrial camera is used to acquire images of the chip packaging area. The resolution of the acquired images should meet the detection requirements of the minimum line width and spacing of the gold wires. The color image is converted to grayscale, high-frequency noise in the image is removed, and edge enhancement and local contrast enhancement are performed to highlight the gold wire structure. Gold line candidate region extraction: Use absolute threshold or adaptive threshold to extract the highlight areas that may be gold lines, convert the thick gold line areas into a single-pixel-width skeleton, remove the interference areas that are not gold lines, and combine the skeleton with edge information to form a gold line candidate region that can be used for pathfinding. Start and end point identification: Based on the encapsulation structure template, the welding start and end points of the gold wire are automatically identified, and all identifiable gold wire pairs constitute the input coordinates for path search; The A* path search algorithm: Based on the traditional heuristic function, it introduces the perpendicular distance from the current point to the line connecting the start and end points, so that the path approaches a straight line connection; Geometric analysis of the gold line trajectory: A local weighted regression line fitting algorithm is introduced to smooth the original trajectory; Defect identification and output marking: Based on the trajectory analysis results, the final abnormal information is output.

[0004] Furthermore, the distance from a point to a line in the path search algorithm is calculated using the following formula: ; Pre-calculate the linear coefficients A, B, C, and... It can be quickly reused each time a node is expanded.

[0005] Furthermore, all pixels outside the candidate region of the gold line are set as obstacles. During the search process, it can be dynamically determined whether they are obstacle points or out-of-bounds nodes. A custom lightweight priority queue is used to achieve priority expansion of nodes with the lowest cost.

[0006] Furthermore, the path search process includes: S1: Initialize the map, start and end points; S2: Open list is added to the starting point, closed list is empty; S3: Find the node with the minimum total cost F; visit all its adjacent nodes; if it is not visited, is not an obstacle, or is valid, calculate G and H; determine whether the path and cost need to be updated; add it to the closed list; S4: Repeat the operation in S3 until the endpoint is found. S5: After the search is completed, backtrack from the endpoint to the parent node to obtain the complete trajectory.

[0007] Furthermore, in the geometric analysis of the gold wire trajectory, for each trajectory point... Take n neighboring pixels to the left and right of the given point, for a total of 2n+1 points. Form a point set with these 2n+1 points, and perform a least-squares line fitting on this point set. The fitted line can be represented as: ; After obtaining the fitted line, the current point... Projecting onto this line yields a new smooth trajectory point. Then point The coordinates of the foot of the perpendicular to the line for: ; ; The final smooth trajectory is composed of all The composition provides better continuity and geometric interpretability.

[0008] Furthermore, the larger the window size n is, the smoother the fit will be, but it may lead to over-averaging. n should be between 3 and 7. During boundary processing, when the neighborhood is insufficient near the start and end points of the path, symmetrical expansion or a smaller window can be used. Multiple smoothing iterations can be performed in one or more rounds of smoothing.

[0009] Furthermore, the geometric features of the path trajectory are extracted and analyzed, including: curvature of the bend, connectivity analysis, spacing statistics, detection of repeated lines, detection of stray lines and redundant paths, and whether it deviates from the target range of the solder joint.

[0010] Furthermore, based on the trajectory analysis results, it is necessary to determine issues such as broken paths, disconnected paths, illegal trajectories, and abnormal distances.

[0011] Furthermore, the anomaly information is output in the form of image tags; The anomaly information is output to the structured data in the form of coordinates or defect type; The abnormal information is used to form a defect map or statistical report, which is used for automatic screening, alarm and rework by the system.

[0012] According to an embodiment of the present invention, a method for detecting gold wires in a chip fully considers the geometric continuity and start-end point connection constraints of the gold wires. The method uses image preprocessing to extract candidate regions for the gold wires, combines a specially optimized A* path algorithm to accurately reconstruct the actual trajectory of the gold wires, and then achieves anomaly identification through trajectory analysis.

[0013] It should be understood that both the foregoing general description and the following detailed description are exemplary and intended to provide further illustration of the claimed technology. Attached Figure Description

[0014] Figure 1 This is a schematic diagram illustrating the detection process of a chip gold wire detection method according to an embodiment of the present invention. Figure 1 ; Figure 2 This is a schematic diagram illustrating the detection process of a chip gold wire detection method according to an embodiment of the present invention. Figure 2 ; Figure 3 This is a schematic diagram illustrating the detection process of a chip gold wire detection method according to an embodiment of the present invention. Figure 3 . Detailed Implementation

[0015] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, further illustrating the present invention.

[0016] like Figures 1-3 As shown, an embodiment of the present invention provides a method for detecting gold wires in a chip, comprising the following steps: Image acquisition and preprocessing: A high-precision industrial camera is used to acquire images of the chip packaging area. The resolution of the acquired images should meet the detection requirements of the minimum line width and spacing of the gold wires. The color image is converted to grayscale, high-frequency noise in the image is removed, and enhancement operations such as edge enhancement and local contrast enhancement are performed to highlight the gold wire structure.

[0017] Gold line candidate region extraction: Use absolute threshold or adaptive threshold to extract the highlight areas that may be gold lines, convert the thick gold line areas into a single-pixel-width skeleton, remove the interference areas that are not gold lines, and combine the skeleton with edge information to form a gold line candidate region that can be used for pathfinding.

[0018] Start and end point identification: Based on the package structure template, the starting and ending points of gold wire welding are automatically identified. Pad positioning, template matching, deep learning and other methods can be used; all identifiable gold wire pairs constitute the input coordinates for path search.

[0019] The A* path search algorithm: Based on the traditional heuristic function, it introduces the vertical distance from the current point to the starting and ending points, so that the path is close to a straight line and avoids unnatural turns.

[0020] The distance from a point to a line in a pathfinding algorithm is calculated using the following formula: ; Pre-calculate the linear coefficients A, B, C, and... It can be quickly reused each time a node is expanded, thus improving pathfinding efficiency.

[0021] Pixels outside the candidate area of ​​the gold line are set as obstacles. During the search process, it can be dynamically determined whether they are obstacle points or out-of-bounds nodes. A custom lightweight priority queue is used to achieve priority expansion of nodes with the lowest cost.

[0022] The path search process includes: S1: Initialize the map, start and end points; S2: Open list is added to the starting point, closed list is empty; S3: Find the node with the minimum total cost F; visit all its adjacent nodes; if it is not visited, is not an obstacle, or is valid, calculate G and H; determine whether the path and cost need to be updated; add it to the closed list; S4: Repeat the operation in S3 until the endpoint is found. S5: After the search is completed, backtrack from the endpoint to the parent node to obtain the complete trajectory.

[0023] Geometric analysis of the gold line trajectory: A local weighted regression line fitting algorithm is introduced to smooth the original trajectory; In the geometric analysis of the gold line trajectory, for each trajectory point Take n neighboring pixels to the left and right of the given point, for a total of 2n+1 points. Form a point set with these 2n+1 points, and perform a least-squares line fitting on this point set. The fitted line can be represented as: ; After obtaining the fitted line, the current point... Projecting onto this line yields a new smooth trajectory point. Then point The coordinates of the foot of the perpendicular to the line for: ; ; The final smooth trajectory is composed of all The composition provides better continuity and geometric interpretability.

[0024] The function of a smooth trajectory is as follows: Improve the accuracy of gold line reconstruction: remove discrete jagged edges during the search process to reflect the true trace trajectory; Facilitates curvature analysis: The first and second derivatives are calculated using smooth curves, which makes it easier to determine whether the gold wire is abnormally curved; For easier comparison with standard templates: the smoothed trajectory is closer to the design path, and error quantification is more accurate; Suppressing image noise: The impact of individual pixel errors on path points is reduced, resulting in greater overall robustness.

[0025] The adjustable parameters are as follows: Window size n: The larger the window size, the smoother the fit, but it may lead to over-averaging. n should be between 3 and 7. Boundary handling: When the neighborhood is insufficient near the start and end points of the path, symmetrical expansion or a smaller window can be used; Multiple smoothing iterations: One or more rounds of smoothing can be performed to improve trajectory quality.

[0026] Extract and analyze the geometric features of the path trajectory, including: curvature of the bend, connectivity analysis, spacing statistics, detection of duplicate lines, detection of stray lines and redundant paths, and whether it deviates from the target range of the solder joint.

[0027] Defect identification and output marking: Based on the trajectory analysis results, the final abnormal information is output.

[0028] Based on the trajectory analysis results, it is necessary to determine issues such as broken paths, disconnected paths, illegal trajectories, and abnormal distances.

[0029] The final exception information will be output in the following format: Image tagging; Output coordinates or defect types to structured data; Defect maps or statistical reports are used for automatic system screening, alarms, and rework.

[0030] Unlike traditional manual detection or simple image comparison algorithms, this application fully considers the geometric continuity of the gold line and the connection constraints of the start and end points. It uses image preprocessing to extract candidate regions of the gold line, combines a specially optimized A* path algorithm to accurately reconstruct the actual trajectory of the gold line, and then achieves anomaly identification through trajectory analysis.

[0031] By eliminating human error through algorithmic modeling, it improves detection accuracy and consistency, supports rapid, batch, and full-image detection, enhances detection efficiency and coverage, adapts to different packaging structures, different gold wire materials and complex backgrounds, and has strong robustness and scalability. The output results can be used for rework, statistics, quality traceability and intelligent analysis systems, facilitating industrial integration applications.

[0032] Above, refer to Figures 1-3 This paper describes a method for detecting gold wires in a chip according to an embodiment of the present invention. Unlike traditional manual detection or simple image comparison algorithms, this application fully considers the geometric continuity and start-end point connection constraints of the gold wires. It uses image preprocessing to extract candidate regions for the gold wires, combines a specially optimized A* path algorithm to accurately reconstruct the actual trajectory of the gold wires, and then achieves anomaly identification through trajectory analysis.

[0033] It should be noted that, in this specification, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0034] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for detecting gold wires in a chip, characterized in that, Includes the following steps: Image acquisition and preprocessing: A high-precision industrial camera is used to acquire images of the chip packaging area. The resolution of the acquired images should meet the detection requirements of the minimum line width and spacing of the gold wires. The color image is converted to grayscale, high-frequency noise in the image is removed, and edge enhancement and local contrast enhancement are performed to highlight the gold wire structure. Gold line candidate region extraction: Use absolute threshold or adaptive threshold to extract the highlight areas that may be gold lines, convert the thick gold line areas into a single-pixel-width skeleton, remove the interference areas that are not gold lines, and combine the skeleton with edge information to form a gold line candidate region that can be used for pathfinding. Start and end point identification: Based on the encapsulation structure template, the welding start and end points of the gold wire are automatically identified, and all identifiable gold wire pairs constitute the input coordinates for path search; The A* path search algorithm: Based on the traditional heuristic function, it introduces the perpendicular distance from the current point to the line connecting the start and end points, so that the path approaches a straight line connection; Geometric analysis of the gold line trajectory: A local weighted regression line fitting algorithm is introduced to smooth the original trajectory; Defect identification and output marking: Based on the trajectory analysis results, the final abnormal information is output.

2. The method for detecting gold wires in a chip as described in claim 1, characterized in that, The distance from a point to a line in the path search algorithm is calculated using the following formula: ； Pre-calculate the linear coefficients A, B, C, and... It can be quickly reused each time a node is expanded.

3. The method for detecting gold wires in a chip as described in claim 1, characterized in that, Pixels outside the candidate region of the gold line are set as obstacles. During the search process, it can be dynamically determined whether they are obstacle points or out-of-bounds nodes. A custom lightweight priority queue is used to achieve priority expansion of nodes with the lowest cost.

4. The method for detecting gold wires in a chip as described in claim 1, characterized in that, The path search process includes: S1: Initialize the map, start and end points; S2: Open list is added to the starting point, closed list is empty; S3: Find the node with the minimum total cost F; visit all its adjacent nodes; if it is not visited, is not an obstacle, or is valid, calculate G and H; determine whether the path and cost need to be updated; add it to the closed list; S4: Repeat the operation in S3 until the endpoint is found. S5: After the search is completed, backtrack from the endpoint to the parent node to obtain the complete trajectory.

5. The method for detecting gold wires in a chip as described in claim 1, characterized in that, In the geometric analysis of the gold thread trajectory, for each trajectory point Take n neighboring pixels to the left and right of the given point, for a total of 2n+1 points. Form a point set from these 2n+1 points, and perform a least-squares line fitting on this point set. The fitted line can be represented as: ； After obtaining the fitted line, the current point... Projecting onto this line yields a new smooth trajectory point. , Then point The coordinates of the foot of the perpendicular to the line for: ； ； The final smooth trajectory is composed of all The composition provides better continuity and geometric interpretability.

6. The method for detecting gold wires in a chip as described in claim 5, characterized in that, A larger window size n results in a smoother fit, but may lead to over-averaging; n should be between 3 and 7. During boundary processing, when the neighborhood is insufficient near the start and end points of the path, symmetrical expansion or a smaller window can be used. Multiple smoothing iterations can be performed in one or more rounds of smoothing.

7. The method for detecting gold wires in a chip as described in claim 6, characterized in that, Extract and analyze the geometric features of the path trajectory, including: curvature of the bend, connectivity analysis, spacing statistics, detection of duplicate lines, detection of stray lines and redundant paths, and whether it deviates from the target range of the solder joint.

8. The method for detecting gold wires in a chip as described in claim 1, characterized in that, Based on the trajectory analysis results, it is necessary to determine issues such as broken paths, disconnected paths, illegal trajectories, and abnormal distances.

9. The method for detecting gold wires in a chip as described in claim 1, characterized in that, The anomaly information is output in the form of image markers; The anomaly information is output to the structured data in the form of coordinates or defect type; The abnormal information is used to form a defect map or statistical report, which is used for automatic screening, alarm and rework by the system.

Images