An image recognition-based die bonder positioning method and system

Abstract

The present application relates to the technical field of image processing, and particularly relates to a positioning method and system of a die bonder based on image recognition, comprising: collecting images of each chip to be positioned and a standard chip at different angles; segmenting the images of each chip to be positioned and the standard chip respectively to extract target regions in the images of each chip to be positioned and the standard chip; identifying feature points of the target regions of each chip to be positioned and the standard chip, and obtaining a number of target feature points of the feature points of the target regions of each chip to be positioned falling within the target region of the standard chip; calculating matching degrees of each chip to be positioned and the standard chip, and responding to the chip to be positioned with the largest matching degree as the optimal chip to be positioned. The present application solves the problem of low positioning accuracy of a die bonder when positioning chips.

Description

Technical Field

[0001] This invention relates to the field of image processing technology. More specifically, this invention relates to a die bonder positioning method and system based on image recognition. Background Technology

[0002] With the rapid development of the semiconductor industry, the requirements for production efficiency and precision in the die bonding process of chip packaging are constantly increasing. As a crucial piece of equipment specifically designed for chip packaging, the die bonder undertakes the critical task of accurately attaching chips to the substrate. The accuracy of the die bonding process directly affects the performance and reliability of the chip; therefore, achieving high-precision chip positioning has become one of the core factors in improving the quality of the die bonding process.

[0003] With the rapid development of image processing and computer vision technologies, image recognition-based positioning methods are increasingly being applied in precision manufacturing. In die bonding processes, image recognition not only enables non-contact positioning but also allows for parameter adjustments via software, reducing reliance on mechanical systems and improving equipment flexibility and reliability. However, existing image recognition technologies often fail to accurately extract feature regions when dealing with chip images from different angles and under varying lighting conditions, leading to positioning errors. Furthermore, image segmentation methods based on fixed thresholds are easily affected by interference when handling complex backgrounds, impacting the accurate identification of target areas.

[0004] Traditional die-bonding positioning methods typically rely on mechanical guidance or optical inspection systems. While these methods can achieve certain positioning effects in some applications, their positioning accuracy and efficiency are often limited due to the small size and complex shape of chips and substrates. Specifically, mechanical guidance systems require high-precision equipment calibration, and in actual production, equipment wear and environmental changes can lead to positioning errors. Furthermore, optical inspection systems are also susceptible to the effects of lighting conditions and environmental interference under different illumination and angle conditions. These factors collectively contribute to a decrease in die-bonding positioning accuracy, making it difficult to further improve positioning accuracy during the die-bonding process. Summary of the Invention

[0005] To address the problem mentioned in the background art of difficulty in further improving positioning accuracy during the die bonding process, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a die bonder positioning method based on image recognition, comprising: acquiring images of each chip to be positioned and a standard chip image from different angles; segmenting each chip image to be positioned and the standard chip image respectively to extract target regions from the chip images to be positioned and the standard chip image; wherein the process of segmenting each chip image to be positioned and the standard chip image respectively comprises: setting an initial local threshold of an adaptive threshold binarization algorithm, segmenting each chip image to be positioned and the standard chip image respectively, updating the initial local threshold, performing multiple segmentations, obtaining the change in the area of ​​connected components before and after each segmentation, and selecting the largest change... The connected components corresponding to the values ​​are used as the target regions; feature points of the target regions of each chip image to be located and the standard chip image are identified, and the transformed positions of the feature points of the target regions of each chip image to be located in the standard chip image are obtained. Based on the transformed positions, the number of target feature points of each chip image to be located falling within the target region of the standard chip image is obtained; the matching degree between each chip image to be located and the standard chip image is calculated. The matching degree is inversely correlated with the number of chip images to be located and the number of feature points in the target regions of each chip image to be located, and positively correlated with the number of target feature points; the chip image to be located with the highest matching degree is the optimal chip image to be located.

[0007] The aforementioned technical solution acquires images of the chip to be located and a standard chip image from multiple angles, employs an adaptive threshold binarization algorithm for multiple segmentations and connected component area change analysis, accurately extracts the target region, and combines feature point matching and transformation position calculation to ensure high-precision matching of the chip image. Furthermore, it is adaptable to different types of chips and has significant technical advantages in chip manufacturing and packaging processes.

[0008] Furthermore, the first The matching degree between the chip image to be located and the standard chip image for:

[0009] ; In the formula, Indicates the number of collections The number of images of the chip to be positioned from different angles. Indicates the number of collections Zhang is waiting to be positioned chip The number of feature points in the target region of the image from each angle. Indicates the number of collections Zhang is waiting to be positioned chip The number of target feature points in the target area of ​​the image at each angle that fall within the target area of ​​the standard chip image.

[0010] The above technical solution enables accurate matching and evaluation of the chip to be located and standard chip images from different angles, comprehensively considering multi-angle information and improving the accuracy and robustness of the matching process. The matching degree quantifies the similarity between chip images by calculating the deviation ratio of feature points, thereby efficiently selecting the optimal chip image to be located and significantly improving the automation and recognition accuracy of chip image matching.

[0011] Furthermore, a CMOS camera is used to acquire images of each chip to be located and a standard chip from different angles.

[0012] Furthermore, it also includes performing image grayscale processing and median filtering processing on images of the chip to be located at different angles and standard chip images at standard angles.

[0013] Furthermore, the update of the initial local threshold specifically involves: The initial local threshold update range is set to [ , The update rule is: add the previous local threshold. As the local threshold after the current update Indicates the initial local threshold. This indicates the termination of the local threshold.

[0014] The above technical solution achieves precise control of adaptive threshold adjustment, effectively improving the flexibility and accuracy of the segmentation process, ensuring that the connected components corresponding to the maximum change are captured in multiple segmentations, thereby extracting the target region more accurately and improving the efficiency and robustness of image segmentation.

[0015] Furthermore, the change of position specifically refers to: Feature points are extracted from the target regions of each chip image to be located and the standard chip image using a feature point detection algorithm. Then, a nearest neighbor search matching algorithm is used to match the feature points of the target regions of each chip image to be located with the feature points of the target regions of the standard chip image. Based on the matching results, a geometric transformation matrix is ​​calculated between the feature points of the target regions of each chip image to be located and the feature points of the target regions of the standard chip image. Using the geometric transformation matrix, the feature points of each chip image to be located are transformed into the target regions of the standard chip image, thereby obtaining the transformed positions of the feature points of each chip image to be located in the target regions of the standard chip image.

[0016] The above technical solution accurately extracts and matches feature points of the chip to be located and the standard chip through feature point detection and nearest neighbor search matching algorithms, calculates their geometric transformation matrix, and realizes the spatial transformation of feature points. By transforming the feature points of the image of the chip to be located to the target area of ​​the standard chip image, the transformed position of the feature points can be accurately obtained, thereby effectively improving the accuracy and robustness of image matching and helping to achieve efficient and accurate chip localization and matching.

[0017] Furthermore, the feature point detection algorithm includes: SIFT, SURF, or ORB algorithm.

[0018] In a second aspect, the present invention provides a die bonder positioning system based on image recognition, comprising a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the image recognition-based die bonder positioning method described above is implemented.

[0019] The beneficial effects of this invention are as follows: This invention segments the image of the chip to be located and a standard chip image, dynamically adjusts the threshold using an adaptive threshold binarization algorithm, extracts the target region, and calculates the matching degree of the images based on feature point matching and geometric transformation. The matching degree is measured based on the number and distribution of feature points, ensuring accurate comparison between the chip to be located and the standard chip. This process effectively reduces errors, improves the accuracy of image segmentation and feature matching, and ultimately selects the optimal chip image to be located based on the matching degree, achieving efficient and accurate chip image localization. Attached Figure Description

[0020] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein: Figure 1 This is a schematic flowchart illustrating a die bonder positioning method based on image recognition according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the structure of a die bonder positioning system based on image recognition according to an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0023] An embodiment of a die bonder positioning method based on image recognition.

[0024] like Figure 1 As shown, a flowchart of a die bonder positioning method based on image recognition according to an embodiment of the present invention is included, comprising the following steps: S1: Acquire images of each chip to be located and a standard chip from different angles.

[0025] In one embodiment, a high-resolution CMOS camera can be used to acquire images of each chip to be located and a standard chip from different angles. Alternatively, a CCD or other camera can be used to acquire images of each chip to be located and a standard chip.

[0026] Specifically, the chip to be positioned is fixed on the positioning device, and the angle is controlled by a robotic arm or a rotating table to adjust the shooting angle of the camera one by one to obtain images of the chip to be positioned from multiple different angles, while simultaneously capturing images of a standard chip.

[0027] In this embodiment, after acquiring chip images, preprocessing is first performed on the images of each chip to be located and the standard chip image. The image preprocessing includes grayscale conversion and median filtering. For grayscale conversion: the color image is converted to grayscale to remove redundant color information, simplifying subsequent image processing. The grayscale conversion process can use a weighted method based on RGB color channels to calculate the grayscale value of each pixel. The grayscale image can more intuitively present the chip's boundaries and structural features, laying the foundation for subsequent feature point extraction and matching. For median filtering: salt-and-pepper noise is eliminated while preserving image edge details. The steps of median filtering are as follows: for each pixel in the image, a neighborhood window is set around it, the median value of the pixels in the window is calculated, and this median value replaces the original pixel value. This process can significantly reduce image noise, improve image quality, and ensure that chip features are clearly presented in the image.

[0028] S2: Segment each chip image to be located and the standard chip image respectively to extract the target region in each chip image to be located and the standard chip image.

[0029] In one embodiment, the process of segmenting each chip image to be located and the standard chip image is as follows: set the initial local threshold of the adaptive threshold binarization algorithm, segment each chip image to be located and the standard chip image, update the initial local threshold, perform multiple segmentations, obtain the change in the area of ​​the connected region before and after each segmentation, and take the connected region corresponding to the maximum value of the change as the target region. Specifically, updating the initial local threshold involves setting the update range of the initial local threshold to [ ]. , The update rule is: add the previous local threshold. As the local threshold after the current update Indicates the initial local threshold. This indicates the termination of the local threshold.

[0030] For example, suppose a high-resolution CMOS camera captures five images of the chip to be located from different angles and one standard chip image. An adaptive threshold binarization algorithm is used for initial segmentation. Based on the image brightness distribution, the initial local threshold range is set to [100, 150]. Let's say we have a threshold of 5. Assuming we segment the image with an initial local threshold of 100 to obtain the connected region area M, and then update the initial local threshold to 105 to obtain the connected region area N, the change in the updated connected region area is MN. We gradually update the initial local threshold to obtain the change in the connected region area each time, and take the connected region corresponding to the maximum change as the target region.

[0031] S3: Identify the feature points of the target regions of each chip image to be located and the standard chip image, and obtain the transformed positions of the feature points of the target regions of each chip image to be located in the standard chip image.

[0032] In one embodiment, obtaining the transformed positions of feature points in the target regions of each chip image to be located within the standard chip image specifically involves: extracting feature points from the target regions of both the chip image to be located and the standard chip image using a feature point detection algorithm; matching the feature points of the target regions of each chip image to be located with the feature points of the target regions of the standard chip image using a nearest neighbor search matching algorithm; calculating the geometric transformation matrix between the feature points of the target regions of each chip image to be located and the feature points of the target regions of the standard chip image based on the matching results; and using the geometric transformation matrix to transform the feature points of each chip image to be located into the target regions of the standard chip image, thereby obtaining the transformed positions of the feature points of each chip image to be located within the target regions of the standard chip image. In this embodiment, the feature point detection algorithm used is the SIFT algorithm.

[0033] In another embodiment, the feature point detection algorithm includes: SIFT, SURF, or ORB algorithms. The SIFT algorithm has strong scale and rotation invariance and is suitable for target matching in complex backgrounds. The SURF algorithm is faster than SIFT and is suitable for application scenarios with high real-time requirements. The ORB algorithm is an efficient feature detection and matching algorithm that can reduce computational complexity while ensuring accuracy and is more suitable for resource-constrained devices.

[0034] S4: Calculate the matching degree between each chip image to be located and the standard chip image based on the changed position, and the chip image to be located with the highest matching degree is the optimal chip image to be located.

[0035] In one embodiment, the first The matching degree between the chip image to be located and the standard chip image for: ; In the formula, Indicates the number of collections The number of images of the chip to be positioned from different angles. Indicates the number of collections Zhang is waiting to be positioned chip The number of feature points in the target region of the image from each angle. Indicates the number of collections Zhang is waiting to be positioned chip The number of target feature points in the target area of ​​the image at each angle that fall within the target area of ​​the standard chip image.

[0036] For example, suppose the system acquires three images of the chip to be located from different angles and matches them with a standard chip image. For the first... The image of the chip to be located has 3 images from different angles. We assume the following data: for the image from the first angle, the number of feature points in the target region... The number of feature points falling within the target area of ​​the standard chip image is 50. The value is 5; for the image from the second angle, the number of feature points in the target region is... The number of feature points falling within the target area of ​​the standard chip image is 60. The value is 10; for the image at the third angle, the number of feature points in the target region is... The value is 55, and the number of feature points falling within the target area of ​​the standard chip image is 55. The value is 8; substituting it into the matching degree formula, we can obtain the th... The matching degree between the image of the chip to be located and the standard chip image.

[0037] The matching degree of all images of the chip to be located is obtained, and the image of the chip to be located with the highest matching degree is selected as the optimal image of the chip to be located.

[0038] The present invention acquires images of the chip to be located and a standard chip image from different angles, and uses an adaptive threshold binarization algorithm for segmentation to effectively extract feature points in the target region, thereby improving the accuracy of image processing. By calculating the transformed positions of feature points in the target region and their matching degree in the standard chip image, the matching effect of each chip image to be located is evaluated, and the optimal chip image to be located is identified, thus optimizing the efficiency of chip localization and recognition. This process not only enhances the reliability of feature point detection but also improves the accuracy of the matching algorithm, providing a more effective solution for practical applications.

[0039] An embodiment of a die bonder positioning system based on image recognition: like Figure 2 As shown, a block diagram of a die bonder positioning system based on image recognition according to an embodiment of the present invention is provided, including a processor and a memory.

[0040] This invention also provides a die bonder positioning system based on image recognition. For example... Figure 2 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the image recognition-based die bonder positioning method according to the present invention.

[0041] The image recognition-based die bonder positioning system also includes other components well known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0042] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions that can be stored or otherwise maintained by such a computer-readable medium.

[0043] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise explicitly specified.

[0044] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A die bonder positioning method based on image recognition, characterized in that, include: Acquire images of the chip to be located from multiple different angles, and acquire images of a standard chip; Each chip image to be located and the standard chip image are segmented to extract the target region from each chip image to be located and the standard chip image; The process of segmenting each chip image to be located and the standard chip image is as follows: An initial local threshold is set for the adaptive threshold binarization algorithm; each chip image to be located and the standard chip image are segmented, and the initial local threshold is updated. Multiple segmentations are performed to obtain the change in the area of ​​the connected components before and after each segmentation. The connected component corresponding to the maximum change is taken as the target region. Feature points of the target regions of each chip image to be located and the standard chip image are identified, and the transformed positions of the feature points of the target regions of each chip image to be located in the standard chip image are obtained. Based on the transformed positions, the number of target feature points of each chip image to be located falling within the target region of the standard chip image is obtained. The matching degree between each chip image to be located and the standard chip image is calculated. The matching degree is inversely correlated with the number of chip images to be located and the number of feature points in the target regions of each chip image to be located, and positively correlated with the number of target feature points. The transformation position is specifically as follows: Feature points are extracted from the target regions of each chip image to be located and the standard chip image using a feature point detection algorithm; the feature points of the target regions of each chip image to be located are matched with the feature points of the target regions of the standard chip image using a nearest neighbor search matching algorithm; based on the matching results, a geometric transformation matrix is ​​calculated between the feature points of the target regions of each chip image to be located and the feature points of the target regions of the standard chip image; using the geometric transformation matrix, the feature points of each chip image to be located are transformed into the target regions of the standard chip image; the matching degree is measured based on the number and distribution of feature points to ensure accurate comparison between the chip to be located and the standard chip, thereby obtaining the transformed position of the feature points of each chip image target region in the target regions of the standard chip image; The image of the chip to be located with the highest matching degree is the optimal image of the chip to be located.

2. The die bonder positioning method based on image recognition according to claim 1, characterized in that, No. Matching degree between the chip image to be located and the standard chip image for: ； In the formula, Indicates the number of collections The number of images of the positioning chip from different angles. Indicates the number of collections Zhang is waiting to be positioned chip The number of feature points in the target region of the image from each angle. Indicates the number of collections Zhang is waiting to be positioned chip The number of target feature points in the target area of ​​the image at each angle that fall within the target area of ​​the standard chip image.

3. The die bonder positioning method based on image recognition according to claim 1, characterized in that, Images of each chip to be located and a standard chip are acquired from different angles using a CMOS camera.

4. The die bonder positioning method based on image recognition according to claim 1, characterized in that, It also includes performing image grayscale processing and median filtering processing on images of the chip to be located at different angles and standard chip images at standard angles.

5. The die bonder positioning method based on image recognition according to claim 1, characterized in that, The specific method for updating the initial local threshold is as follows: The initial local threshold update range is set to [ , The update rule is: add the previous local threshold. As the local threshold after the current update Indicates the initial local threshold. This indicates the termination of the local threshold.

6. The die bonder positioning method based on image recognition according to claim 1, characterized in that, The feature point detection algorithm includes: SIFT, SURF, or ORB algorithm.

7. A die bonder positioning system based on image recognition, characterized in that, It includes a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the die bonding machine positioning method based on image recognition as described in any one of claims 1 to 6 is implemented.

Images