Wafer laser hidden cutting positioning method

Abstract

The application discloses a wafer laser hidden cutting positioning method and relates to the technical field of wafer processing methods. Specifically, the method comprises the following steps: presetting a wafer standard template; generating a complete 3D topographic map of a wafer surface; obtaining the flat edge and center point coordinates of the wafer, determining the actual position of the wafer, and determining the first rotation angle at which the wafer needs to be rotated; rotating the glass stage to complete the coarse rotation positioning; obtaining the first electrode contour image of the electrode below the wafer, obtaining the center point coordinates and direction angle of the wafer according to the first electrode contour image, determining the current position of the wafer, determining the second rotation angle at which the wafer needs to be rotated according to the current position of the wafer and the position of the wafer standard template, and rotating the glass stage to complete the fine rotation positioning of the wafer. The transverse coordinates and longitudinal coordinates of the cutting channel in the wafer are determined and fitted with the 3D topographic map to obtain the accurate coordinates of the cutting channel in the wafer. The method aims to guarantee cutting precision while reducing the manufacturing cost of the laser cutting equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wafer processing methods, in particular to a wafer laser hidden cutting positioning method. BACKGROUND

[0002] In modern semiconductor device manufacturing, the precision of wafer cutting directly affects the performance and quality of the product. Traditional wafer cutting methods mainly use mechanical cutting, but as the requirements for processing precision and material quality continue to improve, traditional cutting methods have been difficult to meet the demand for high precision and high efficiency.

[0003] Laser cutting technology has been widely used in wafer cutting due to its non-contact, high precision, and strong controllability. Laser hidden cutting technology is an innovative laser cutting technology, which mainly focuses the laser beam to a certain depth below the wafer surface, forms a modified layer at the desired depth, and then applies a certain external force to separate the wafer. Compared with traditional mechanical cutting methods, this technology has higher processing precision, less mechanical damage, and less heat-affected zone, and is suitable for the processing of high-precision and high-reliability semiconductor and optoelectronic devices.

[0004] However, although laser hidden cutting technology has obvious advantages in improving processing precision, the platform on which the wafer is placed may have a certain flatness difference during processing, which causes the laser beam to not always remain within the ideal focal range, affecting the depth and cutting quality of the laser cutting. At the same time, the wafer may be affected by factors such as thermal stress, material inhomogeneity, and unstable process parameters during the process, causing the wafer to warp. The warping phenomenon not only affects the stability of laser focusing, but also may cause the laser beam to not always focus at the predetermined depth of the wafer, resulting in inconsistent processing depth and further affecting the cutting quality of the product.

[0005] In order to ensure the consistency of the depth of the modified layer during laser hidden cutting, continuous monitoring and adjustment of the processing process is required. In traditional laser hidden cutting systems, real-time focusing sensors, range finders, and multiple cameras are usually equipped to accurately control the focal point position, monitor the relative position of the laser beam and the wafer surface, and adjust the power and cutting speed of the laser in real time. These devices monitor the surface state of the wafer and its relative position with the laser beam to help maintain processing precision and reduce the impact of warping and deviation on cutting effect. However, this high-precision detection method, while improving processing precision, also brings higher cost investment and equipment complexity, significantly increasing the cost of laser cutting equipment.

[0006] Therefore, how to reduce the manufacturing cost of laser cutting equipment while ensuring cutting precision has become a technical problem to be solved. SUMMARY

[0007] The main purpose of the present application is to provide a wafer laser hidden cutting positioning method, aiming to reduce the manufacturing cost of laser cutting equipment while ensuring cutting accuracy.

[0008] In order to achieve the above purpose, the present application provides a wafer laser hidden cutting positioning method, comprising the following steps:

[0009] A wafer standard template with the same size as the wafer to be cut and distributed with cutting lanes is preset;

[0010] A 3D camera is used to scan and generate a complete 3D topographic map of the wafer surface;

[0011] According to the generated 3D topographic map, the flat edge and center point coordinates of the wafer are obtained, the actual position of the wafer is determined, and according to the actual position of the wafer and the position of the wafer standard template, the first rotation angle of the wafer that needs to be rotated is determined;

[0012] According to the first rotation angle, the glass stage is rotated to complete the coarse rotation positioning of the wafer;

[0013] A plurality of electrodes are arranged in an array below the wafer, a first electrode contour image of the electrodes below the wafer is obtained, the center point coordinates and the direction angle of the wafer are obtained according to the obtained first electrode contour image, the current position of the wafer is determined, the second rotation angle of the wafer that needs to be rotated is determined according to the current position of the wafer and the position of the wafer standard template, and the glass stage is rotated according to the second rotation angle to complete the fine rotation positioning of the wafer.

[0014] In an embodiment of the present application, after the fine rotation positioning is completed, the following steps are further included:

[0015] According to the first electrode contour image, the first center point coordinate of the wafer is obtained;

[0016] The glass stage is translated horizontally along the plane, and a second electrode contour image of the electrodes below the wafer is obtained; according to the second electrode contour image, the second center point coordinate of the wafer is obtained;

[0017] It is judged whether the plane longitudinal coordinates of the first center point coordinate and the second center point coordinate are the same, if they are the same, it indicates that the wafer has moved to the limited position of the wafer standard template.

[0018] In an embodiment of the present application, the flat edge and center point coordinates of the wafer are obtained according to the generated 3D topographic map; and the center point coordinates and the direction angle of the wafer are obtained according to the obtained electrode contour image; and the first center point coordinate of the wafer is obtained according to the electrode contour image; and the second center point coordinate of the wafer is obtained according to the second electrode contour image; all of which are completed by using a template matching algorithm.

[0019] In an embodiment of the present application, the first electrode profile image of the electrode under the wafer is obtained, including the following steps:

[0020] The focal length adjustment model is constructed;

[0021] The wafer is photographed at different focal lengths to obtain original images of the wafer at different focal lengths, each original image is labeled to generate training images, and the focal length adjustment model is trained through the training images, so that the focal length adjustment model can automatically adjust the position between the coaxial camera and the wafer to make the coaxial camera always focus on the upper surface of the wafer;

[0022] After the coaxial camera focuses on the upper surface of the wafer, the coaxial camera is controlled to move downward by a standard thickness, so that the coaxial camera can obtain the first electrode profile image of the electrode under the wafer; wherein the standard thickness is equal to the thickness of the wafer.

[0023] In an embodiment of the present application, it is judged whether the plane longitudinal coordinates of the coordinates of the first center point and the coordinates of the second center point are the same, if not, the angle value between the line connecting the first center point and the second center point and the cutting groove on the wafer is obtained, and the wafer is rotated based on the first center point according to the angle value.

[0024] In an embodiment of the present application, the focal length adjustment model is constructed using a convolutional neural network.

[0025] In an embodiment of the present application, after the fine positioning of the wafer is completed, it further includes:

[0026] The cutting groove in the wafer standard template is mapped to the actual cutting groove of the wafer to determine the horizontal coordinates and the longitudinal coordinates of the cutting groove in the wafer when the wafer is completely flat; according to the horizontal coordinates and the longitudinal coordinates of the cutting groove in the wafer, the 3D topography map is fitted to obtain the horizontal coordinates, the longitudinal coordinates, and the height coordinates of the cutting groove in the wafer, and the accurate planning of the cutting groove is realized.

[0027] By using the above technical scheme, through the coarse rotation and fine rotation positioning of the wafer, the accurate alignment of the cutting groove and the electrode of the wafer in the cutting process is ensured. Through the cooperation of the 3D camera and the coaxial camera, the position information of the wafer can be obtained in real time, and the corresponding rotation angle is calculated according to the information to realize the accurate positioning of the wafer. This method can greatly improve the precision of wafer cutting, reduce the material waste caused by positioning error in the cutting process, and is suitable for different types of wafers and electrode structures, and has good universality. BRIEF DESCRIPTION OF DRAWINGS

[0028] The present application will be described in detail below with specific embodiments and drawings, in which:

[0029] Figure 1A flowchart of a first embodiment of the present application. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be described in detail below in combination with the drawings and embodiments. It should be understood that the following specific embodiments are only used to explain the present application and do not constitute a limitation on the present application.

[0031] As Figure 1 shown, in order to achieve the above-mentioned purpose, the present application proposes a wafer laser hidden cutting positioning method, comprising the following steps:

[0032] A wafer standard template is preset, which is the same as the wafer to be cut and has cutting lanes distributed thereon;

[0033] A 3D camera is used to scan to generate a complete 3D topographic map of the wafer surface;

[0034] The flat edge and center point coordinates of the wafer are obtained according to the generated 3D topographic map, the actual position of the wafer is determined, and the first rotation angle that the wafer needs to rotate is determined according to the actual position of the wafer and the position of the wafer standard template;

[0035] The glass stage is rotated according to the first rotation angle, and the coarse rotation positioning of the wafer is completed;

[0036] A plurality of electrodes are arranged in an array below the wafer, a first electrode contour image of the electrodes below the wafer is obtained, the center point coordinates and the direction angle of the wafer are obtained according to the obtained first electrode contour image, the current position of the wafer is determined, the second rotation angle that the wafer needs to rotate is determined according to the current position of the wafer and the position of the wafer standard template, and the fine rotation positioning of the wafer is completed by rotating the glass stage according to the second rotation angle.

[0037] Specifically, in the present embodiment, the preset wafer standard template is a template with the same size as the wafer to be cut, and the cutting lanes are distributed on the template according to the actual cutting lanes of the wafer to be cut, so as to ensure that the positions, number and size of the cutting lanes on the template are consistent with those of the wafer to be cut. The template is modeled using computer software and adjusted according to different wafer sizes in actual application. The use of the wafer standard template ensures the positioning accuracy of the wafer and the accurate matching of the cutting lanes.

[0038] A 3D camera is used to scan the surface of the wafer to be cut to obtain a complete 3D topographic map. The 3D camera can be selected to use a laser scanning method or a structured light scanning method, and the scanning process ensures that the morphological characteristics of the wafer surface, including the flat edge and surface texture, are obtained. The resolution of the 3D camera is at least 1 μm, which ensures that the scanning data accuracy is high enough for subsequent positioning and rotation.

[0039] After obtaining the complete 3D topography of the wafer, the flat edge information of the wafer is extracted using an image processing algorithm, and the center point coordinates of the wafer are calculated. The specific method includes that the flat edge contour can be extracted by an edge detection algorithm, and then the center point position is determined by geometric calculation. Through this step, the position of the wafer can be accurately determined.

[0040] Based on the comparison of the center point coordinates of the wafer and the position of the standard template, the first rotation angle that the wafer needs to rotate is calculated. The specific calculation method is to obtain the angle difference of the flat edge of the wafer relative to the standard template through geometric transformation, and determine the initial rotation angle of the wafer on the glass stage. Then, the rotation axis of the glass stage is controlled to rotate, so that the wafer on the stage is roughly matched with the standard template, and the coarse rotation positioning is completed.

[0041] In the electrode area below the wafer surface, a coaxial camera or a high-resolution camera is configured to take a picture to obtain a first electrode contour image. When taking the picture, the camera focal length and focusing position are ensured to be accurate, so that the electrode contour can be clearly obtained. After image processing, the shape information of the electrode can be extracted, and then the center point coordinates and direction angle of the wafer are calculated.

[0042] According to the difference between the current position of the wafer and the position of the standard template, the second rotation angle that the wafer needs to rotate is calculated. The rotation angle is mainly obtained by comparing the coordinates and direction angle of the electrode contour image with the corresponding part of the standard template, to obtain the accurate position of the wafer. By adjusting the rotation angle of the glass stage, the fine rotation positioning of the wafer is completed, and the position of the wafer on the glass stage is completely matched with the standard template.

[0043] By using the above technical scheme, the coarse rotation and fine rotation positioning of the wafer are realized, and the accurate alignment of the wafer cutting path and the electrode in the cutting process is ensured. By using the cooperation of the 3D camera and the coaxial camera, the position information of the wafer can be obtained in real time, and the corresponding rotation angle can be calculated according to the information to realize the accurate positioning of the wafer. This method can greatly improve the precision of wafer cutting, reduce the material waste caused by positioning error in the cutting process, and is suitable for different types of wafers and electrode structures, and has good universality.

[0044] In an embodiment of the present application, after the fine rotation positioning is completed, the following steps are further included:

[0045] According to the first electrode contour image, the first center point coordinate of the wafer is obtained;

[0046] The glass stage is translated along the plane in the transverse direction, and a second electrode contour image of the electrode below the wafer is obtained; according to the second electrode contour image, the second center point coordinate of the wafer is obtained;

[0047] The longitudinal coordinates of the first center point and the second center point are compared to determine whether they are the same. If they are the same, it indicates that the wafer has moved to the limited position of the wafer standard template.

[0048] Specifically, embodiment 1:

[0049] After the fine rotation positioning is completed, the first center point coordinate of the wafer is extracted by analyzing the electrode contour through an image processing algorithm based on the first electrode contour image obtained in the foregoing. The calculation of the coordinate is realized through image analysis techniques such as edge detection and contour matching, and the accuracy is ensured to reach the micron level. The first center point coordinate is the initial point of the positioning of the wafer on the glass stage.

[0050] The glass stage is controlled to move horizontally along the plane (i.e., along the X-axis direction) to ensure that the electrode area below the wafer can be photographed again by the coaxial camera or the high-resolution camera. During the movement, the movement speed and accuracy of the stage need to be strictly controlled to avoid positioning deviation in the subsequent steps due to the translation error. The second electrode contour image is obtained by scanning the electrode area below the wafer.

[0051] The second center point coordinate of the wafer is calculated through image processing of the second electrode contour image using the same electrode recognition algorithm. The calculation method of the second center point is consistent with that of the first center point, and the high-precision connection of the two steps is ensured.

[0052] The longitudinal coordinates (i.e., Y-axis coordinates) of the first center point coordinate and the second center point coordinate are compared to determine whether they are consistent. This determination is used to confirm whether the wafer has been accurately moved to the limited position of the wafer standard template. If the longitudinal coordinates are consistent, it indicates that the wafer has reached the accurate position and does not need to be further adjusted. If the longitudinal coordinates are inconsistent, it indicates that the wafer has not reached the positioning requirement of the standard template, and needs to be further adjusted and optimized.

[0053] By using the above technical solution, after the fine rotation positioning of the wafer is completed, the longitudinal coordinates of the two center points are compared to realize the accurate positioning of the wafer. By moving the wafer horizontally and obtaining the electrode contour image, the position accuracy of the wafer is further verified, and the processing error caused by the positioning error is effectively avoided.

[0054] In an embodiment of the present application, the flat side and the center point coordinate of the wafer are obtained according to the generated 3D topographic map; the center point coordinate and the direction angle of the wafer are obtained according to the obtained electrode contour image; the first center point coordinate of the wafer is obtained according to the obtained electrode contour image; and the second center point coordinate of the wafer is obtained according to the second electrode contour image; all of which are completed by using a template matching algorithm.

[0055] By adopting the technical scheme, the template matching algorithm can accurately find the target position in the image by comparing the similarity between the actual image and the preset standard template. Meanwhile, the template matching algorithm is relatively simple and easy to implement, and through the standardized tool and algorithm library, the application can be developed and deployed relatively quickly, thereby reducing the development cost and time.

[0056] In an embodiment of the present application, the first electrode profile image of the electrode under the wafer is obtained, including the following steps:

[0057] A focal length adjustment model is constructed.

[0058] The wafer is photographed at different focal lengths to obtain original images of the wafer at different focal lengths. Each original image is labeled to generate training images. The focal length adjustment model is trained through the training images, so that the focal length adjustment model can automatically adjust the position between the coaxial camera and the wafer to make the coaxial camera always focus on the upper surface of the wafer.

[0059] After the coaxial camera focuses on the upper surface of the wafer, the coaxial camera is controlled to move downward by a standard thickness, so that the coaxial camera can obtain the first electrode profile image of the electrode under the wafer; wherein the standard thickness is equal to the thickness of the wafer.

[0060] Specifically, a focal length adjustment model is constructed, which is used to adjust the focal length position between the coaxial camera and the wafer to ensure that the camera can always focus on the upper surface of the wafer. The construction process of the focal length adjustment model is trained by a large amount of image data at different focal lengths, so as to optimize the adjustment mechanism and make the coaxial camera automatically adjust to the most appropriate focal length.

[0061] The training process of the model is as follows:

[0062] The wafer is photographed at different focal lengths to obtain original images of the wafer at different focal lengths. These image data can cover different surface states and focal length positions of the wafer, and provide a basis for subsequent labeling and training.

[0063] Each original image obtained is labeled. The labeling content includes the corresponding relationship between the focal length position and the different positions on the upper surface of the wafer. Through these labeling information, a training image set is generated. Each training image corresponds to a specific focal length value. These training images will be used as an input data set to train the focal length adjustment model, so that it can adaptively adjust the focal length of the coaxial camera according to the actual shooting situation.

[0064] Using the generated training image set, the focal length adjustment model is trained. This model automatically learns the relationship between focal length and image sharpness from the training images through deep learning techniques, combined with convolutional neural networks (CNN) or other suitable network architectures. The training process gradually improves the accuracy of the model by optimizing the loss function to ensure that the coaxial camera can always maintain a sharp focus on the upper surface of the wafer.

[0065] Once the focal length adjustment model is trained and verified, the system controls the coaxial camera to automatically adjust the focal length to ensure that the coaxial camera can accurately focus on the upper surface of the wafer. At this time, through the control of the focal length adjustment model, the coaxial camera will make a clear image of the wafer surface.

[0066] After the coaxial camera focuses on the upper surface of the wafer, the coaxial camera is further controlled to move downward along the Z-axis direction. The moving distance is set to the standard thickness of the wafer, which is equal to the actual thickness of the wafer. At this time, the coaxial camera will focus on the lower side of the wafer and can obtain the first electrode profile image of the lower electrode.

[0067] After the coaxial camera moves downward by the standard thickness, the coaxial camera will obtain the first electrode profile image of the lower electrode of the wafer, thereby ensuring the clarity of the first profile.

[0068] An example is as follows:

[0069] A 6-inch wafer is placed on a self-made image acquisition device, 1000 base point coordinates are set for self-focusing, the defocus offset is set to 0.005 mm, the upper and lower limits of the offset are set to [-2, 2] mm, the automatic picture acquisition is started, and the picture and offset data are output.

[0070] The obtained picture is subjected to ROI processing, and a center 56*56 size picture is cropped and subjected to corresponding image enhancement processing; the above data is divided into a training set, a verification set, and a test set in a corresponding 6:2:2 ratio;

[0071] The PyTorch deep learning architecture is used, the ResNet50 network is migrated using the pre-training method, the important features in the input image are extracted, and then the self-defined loss function and optimizer are matched to complete the classification work through the corresponding linear layer.

[0072] Based on the built model, the cycle number is set to 50, the initial learning rate is set to 0.0001, and the sample size is set to 128 for training.

[0073] The trained model is tested on the test set data, and when the accuracy reaches 99%, the corresponding focusing task is completed. If the accuracy is not enough, the existing model needs to be modified and retrained based on the corresponding parameters.

[0074] By training the focal length adjustment model, the coaxial camera can automatically adapt to different focal length requirements, ensuring that the camera can accurately focus on the upper surface of the wafer and improving the clarity and quality of the image. Through precise control of the coaxial camera, the camera can not only focus on the upper surface of the wafer, but also accurately control the camera position to obtain a clear profile image of the electrode below the wafer, facilitating subsequent processing and analysis. The application of the focal length adjustment model makes the entire process automated, eliminating the need for manual adjustment of the focal length position, greatly improving work efficiency and reducing human error.

[0075] In an embodiment of the present application, it is determined whether the planar longitudinal coordinates of the first center point and the second center point are the same. If not, the angle between the line connecting the first center point and the second center point and the cutting path on the wafer is obtained, and the wafer is rotated based on the first center point.

[0076] Specifically, the first electrode profile image and the second electrode profile image of the electrode below the wafer are obtained. The first center point coordinate and the second center point coordinate are extracted from the two electrode profile images respectively through a template matching algorithm. After obtaining the two point coordinates, the difference in longitudinal coordinates on the plane is calculated. If the longitudinal coordinates of the two points are the same, it indicates that the wafer has been accurately positioned to the limited position of the wafer standard template and does not need to be further adjusted. If the longitudinal coordinates are different, the angle between the line connecting the first center point and the second center point and the cutting path on the wafer is calculated. Based on the angle value, the glass stage is controlled to rotate the wafer based on the first center point as the reference until the longitudinal coordinates of the two center points match, thereby completing the accurate positioning of the wafer.

[0077] By accurately determining whether the longitudinal coordinates of the first center point and the second center point are the same, it can be effectively determined whether the wafer has been correctly aligned. If not, the wafer is rotated based on the calculated angle, further ensuring accurate positioning of the wafer during processing. This method can significantly reduce the positional deviation of the wafer during subsequent processing, ensuring high-precision cutting and processing.

[0078] In an embodiment of the present application, the focal length adjustment model is constructed using a convolutional neural network.

[0079] By using a convolutional neural network, the convolutional neural network can automatically extract and learn multi-level features in the image when processing the image. It has strong precision and robustness in image recognition and processing, can cope with complex environmental changes such as different light, background, texture, etc., and can accurately adjust the small changes in focal length, facilitating the construction of the focal length adjustment model.

[0080] In an embodiment of the present application, after the accurate positioning of the wafer is completed, the method further comprises:

[0081] The scribe lane in the wafer standard template is mapped to the actual scribe lane of the wafer to determine the horizontal and vertical coordinates of the scribe lane in the wafer when the wafer is completely flat; the horizontal, vertical, and height coordinates of the scribe lane in the wafer are obtained by fitting the 3D topography map according to the horizontal and vertical coordinates of the scribe lane in the wafer, and accurate scribe lane planning is achieved.

[0082] Specifically, the scribe lane coordinates in the preset wafer standard template are mapped to the scribe lane of the actual wafer. During the mapping process, the center point coordinates and direction angle of the actual wafer are obtained through the fine spin positioning result of the wafer, and the horizontal and vertical coordinates of the scribe lane are adjusted in combination with the scribe lane information in the wafer standard template to make them completely aligned with the position of the actual wafer, so as to ensure that the mapped scribe lane has consistent geometric distribution with the scribe lane on the actual wafer.

[0083] The mapped scribe lane of the wafer standard template is fitted with the coordinate of the three-dimensional topography map of the actual wafer. First, the corresponding positions of the scribe lane in the horizontal coordinate (X-axis) and the vertical coordinate (Y-axis) are extracted according to the wafer surface information in the three-dimensional topography map. Then, the height coordinate (Z-axis) of each point of the scribe lane is calculated in combination with the height information of the three-dimensional topography map. Through fitting calculation, the accurate position coordinates of the scribe lane in the three-dimensional space are obtained, including the horizontal coordinate, the vertical coordinate, and the height coordinate.

[0084] According to the three-dimensional space coordinates of the scribe lane obtained by fitting, the cutting path is accurately planned. Based on the height coordinate information of the scribe lane, the local tilt or deformation that may exist in the wafer during the cutting process is calculated. By adjusting the height parameter of the cutting path, it is ensured that the cutting tool always maintains the ideal relative position with the wafer surface during the machining process, so as to realize accurate cutting.

[0085] The above technical scheme realizes the fine spin positioning of the wafer to the three-dimensional coordinate accurate planning of the scribe lane, ensures that the wafer cutting path is completely consistent with the standard scribe lane, and significantly improves the quality and precision of wafer processing.

[0086] The above only describes the preferred embodiments of the present application, and does not limit the patent scope of the present application, and any equivalent structural transformation made by using the content of the present application specification and drawings, or direct / indirect application in other related technical fields is included in the patent protection scope of the present application.

Claims

1. A wafer laser hidden grooving positioning method, characterized in that, Includes the following steps: A pre-set wafer standard template is identical to the wafer to be cut and has dicing channels distributed on it. A complete 3D topographic image of the wafer surface is generated by scanning with a 3D camera. The coordinates of the flat edge and center point of the wafer are obtained from the generated 3D topography image to determine the actual position of the wafer. Based on the actual position of the wafer and the position of the wafer standard template, the first rotation angle that the wafer needs to rotate is determined. The glass stage is rotated according to the first rotation angle to complete the coarse rotation positioning of the wafer; Multiple electrodes are arranged in an array below the wafer. The first electrode contour image of the electrodes below the wafer is acquired. Based on the acquired first electrode contour image, the coordinates of the center point of the wafer and the orientation angle are obtained to determine the current position of the wafer. Based on the current position of the wafer and the position of the wafer standard template, the second rotation angle that the wafer needs to rotate is determined. The glass stage is rotated according to the second rotation angle to complete the precise rotation and positioning of the wafer. Based on the first electrode contour image, obtain the coordinates of the first center point of the wafer; The glass stage is translated laterally along the plane to obtain the second electrode contour image of the electrode below the wafer; the coordinates of the second center point of the wafer are obtained based on the second electrode contour image. Determine whether the ordinate of the first center point and the second center point are the same in plane. If they are the same, it means that the wafer has moved to the defined position of the wafer standard template. Determine whether the plane ordinates of the coordinates of the first center point and the second center point are the same. If they are not the same, obtain the angle value between the line connecting the first center point and the second center point and the dicing track on the wafer, and rotate the wafer according to the angle value and with the first center point as the reference. Map the dicing traces in the standard wafer template to the actual dicing traces on the wafer to determine the horizontal and vertical coordinates of the dicing traces on the wafer when the wafer is in a completely flat state. By fitting the horizontal and vertical coordinates of the dicing channels within the wafer to the 3D topography image, the horizontal, vertical, and height coordinates of the dicing channels within the wafer are obtained, enabling precise dicing channel planning. Construct a focal length adjustment model; The wafer is photographed at different focal lengths to obtain original images of the wafer at different focal lengths. Each original image is labeled to generate training images. The focal length adjustment model is trained using the training images so that the focal length adjustment model can automatically adjust the position between the coaxial camera and the wafer so that the coaxial camera is always focused on the upper surface of the wafer. After the coaxial camera focuses on the upper surface of the wafer, the coaxial camera is controlled to move downward by a standard thickness so that the coaxial camera can acquire the first electrode profile image of the electrode below the wafer; where the standard thickness is equal to the wafer thickness.

2. The wafer laser etched positioning method as described in claim 1, characterized in that, The wafer's flat edge and center point coordinates are obtained from the generated 3D topography image; the wafer's center point coordinates and orientation angle are obtained from the obtained electrode contour image; the wafer's first center point coordinates are obtained from the obtained electrode contour image; and the wafer's second center point coordinates are obtained from the second electrode contour image; all of these are accomplished using a template matching algorithm.

3. The wafer laser etched positioning method as described in claim 1, characterized in that, The focal length adjustment model is constructed using a convolutional neural network.

Images