CN-115456977-B - Wafer defect detection method, system and storage medium

CN115456977BCN 115456977 BCN115456977 BCN 115456977BCN-115456977-B

Abstract

The invention discloses a wafer defect detection method, a wafer defect detection system and a storage medium. The method comprises the steps of collecting images of a to-be-detected target saw-tooth area in a wafer, extracting angular points from the images, calculating angle values of the angular points, taking the angular points with the angle values in a preset interval as first alternative profile angular points, obtaining coordinate values of all first alternative profile angular points, determining tooth peak areas and tooth valley areas of the to-be-detected target saw-tooth area according to distribution of all first alternative profile angular points in a coordinate system, taking the first alternative profile angular points located in the tooth peak areas and the tooth valley areas as detection profile points of the to-be-detected target saw-tooth area, matching the detection profile points of the to-be-detected target saw-tooth area with profile points of a template saw-tooth area, and determining saw-tooth defect types according to matching results. The invention is suitable for detecting various saw-tooth targets of the wafer, in particular for detecting targets in a sparse period rough process, such as wafer substrate colloid or crystal grains.

Inventors

ZHANG HU
Xiong Baitai

Assignees

武汉精立电子技术有限公司
武汉精测电子集团股份有限公司

Dates

Publication Date: 20260505
Application Date: 20220901

Claims (9)

1. A method for detecting wafer defects, comprising: Collecting an image of a target sawtooth area to be detected in a wafer, extracting corner points from the image, and calculating the angle value of each corner point; Taking angular points with angle values in a preset interval as first alternative contour angular points, acquiring coordinate values of all the first alternative contour angular points, determining tooth peak areas and tooth valley areas of the to-be-detected target saw-tooth areas according to distribution of all the first alternative contour angular points in a coordinate system, and taking the first alternative contour angular points positioned in the tooth peak areas and the tooth valley areas as detection contour points of the to-be-detected target saw-tooth areas; matching the detection contour points of the sawtooth region of the target to be detected with the contour points of the sawtooth region of the template, and determining the sawtooth defect type of the target to be detected according to a matching result; Extracting corner points from the image and calculating an angle value of each corner point comprises: for any pixel p in the image, acquiring gray values of 16 pixels on the circumference with the pixel p as a circle center and the radius of 3 pixels, numbering the 16 pixels from 1 to 16 in turn according to clockwise or anticlockwise, and if the gray values of n continuous pixels in the 16 pixels are all larger than the gray value Ip of the pixel p by a preset threshold t, or the gray values of n continuous pixels are all smaller than the gray value Ip of the pixel p by the preset threshold t, and n is larger than the preset threshold M,1< M <16, extracting the pixel p as a corner point; the mapping relation between the n value and the angle value is predetermined, and the angle value of each angular point is obtained according to the n value in the angular point extraction process.
2. The method of claim 1, wherein the target is a wafer substrate gel or a die.
3. The method according to claim 1, wherein the preset interval is set according to the shape of the target to be detected, and if the standard saw tooth of the target to be detected is a regular triangle, the preset interval is set to be [60 ° - α,60 ° +α ] [300 ° - α,300 ° +α ], α being an angle error threshold.
4. The wafer defect detection method of claim 1, wherein the taking the first candidate contour corner points located in the tooth peak area and the tooth valley area as the detected target saw tooth area contour points comprises: And processing the first alternative contour angular points positioned in the tooth peak area and the tooth valley area by adopting a non-maximum value suppression algorithm, removing the non-maximum value to obtain a second alternative contour angular point, and taking the second alternative contour angular point as a detection contour point of the to-be-detected target sawtooth area.
5. The wafer defect detection method of claim 1, wherein the determining the tooth peak areas and the tooth valley areas of the target saw-tooth area to be detected according to the distribution of all the first candidate contour corner points in the coordinate system comprises: determining whether the saw tooth arrangement direction of the object to be detected is parallel to the x axis of the coordinate system according to the distribution of all the first alternative contour corner points in the coordinate system; if the arrangement direction of the saw teeth of the object to be detected is parallel to the x axis, determining a tooth peak area and a tooth valley area according to the y axis coordinate value of the first alternative contour angular point; If the sawtooth arrangement direction of the object to be detected is not parallel to the x axis, determining an included angle between the sawtooth arrangement direction of the object to be detected and the x axis, correcting coordinate values of all first alternative contour angular points according to the included angle, enabling the corrected sawtooth arrangement direction to be parallel to the x axis, and determining a tooth peak area and a tooth valley area according to the corrected y-axis coordinate values of the first alternative contour angular points.
6. The wafer defect detection method of claim 1 wherein the template saw tooth area image is input to a trained neural network model to generate template saw tooth area contour points.
7. The method for detecting wafer defects according to claim 1, wherein the matching the detected contour points of the target saw tooth area to be detected with the contour points of the template saw tooth area comprises: counting the angle value of the detected contour points of the target sawtooth region to be detected and the interval between the detected contour points, and counting the angle value of the contour points of the template sawtooth region and the interval between the contour points; And matching is carried out according to the statistical information of the detection contour points of the target sawtooth region to be detected and the statistical information of the contour points of the template sawtooth region.
8. A wafer defect inspection system for implementing the wafer defect inspection method of any one of claims 1 to 7, comprising: The preprocessing module is used for collecting images of the target sawtooth area to be detected in the wafer, extracting corner points from the images and calculating the angle value of each corner point; The profile point detection module is used for taking angular points with angle values in a preset interval as first alternative profile angular points, acquiring coordinate values of all the first alternative profile angular points, determining a tooth peak area and a tooth valley area of the to-be-detected target saw-tooth area according to distribution of all the first alternative profile angular points in a coordinate system, and taking the first alternative profile angular points positioned in the tooth peak area and the tooth valley area as detection profile points of the to-be-detected target saw-tooth area; and the matching module is used for matching the detection contour point of the target sawtooth region to be detected with the contour point of the template sawtooth region, and determining the sawtooth defect type of the target to be detected according to a matching result.
9. A storage medium storing a computer program which, when run on a processor, causes the processor to perform the steps of the method of any one of claims 1 to 7.

Description

Wafer defect detection method, system and storage medium Technical Field The present application relates to the field of wafer inspection technology, and in particular, to a wafer defect inspection method, system and storage medium. Background In the field of semiconductor package test process detection, the detection of target saw teeth in various sparse periodic rough processes is often involved. Because the manufacturing process of the target to be detected is unstable, the randomness of the related specification size is strong, the stable positioning can not be performed by using a specific template, various defects are generated, the fluctuation randomness is large, the periodic distribution rule is not generated, available effective image information is limited, and enough effective information can not be provided for positioning the target, so that the conventional defect detection method can not obtain a good detection effect. For example, in the wafer manufacturing process flow for CIS (cmos imaging sensor) chips, the glue injection process has a great yield challenge, because the process quality directly affects the quality of the subsequent bonding process and the finished product, and it is important to detect the glue on the glass substrate in the previous process. However, since the colloid is freely diffused, the actual production effect and the theoretical process standard have great fluctuation and randomness. For defect detection in this process, conventional golden template matching (die-to-goldendie) and die matching (die-to-die) algorithm schemes based on gray scale comparison are no longer applicable. Disclosure of Invention In view of at least one defect or improvement requirement in the prior art, the invention provides a wafer defect detection method, a system and a storage medium, which are suitable for detecting various saw-tooth targets of a wafer, and are particularly suitable for detecting targets of a sparse periodic rough process, such as wafer substrate colloid or crystal grains. In order to achieve the above object, according to a first aspect of the present invention, there is provided a wafer defect detecting method comprising: Collecting an image of a target sawtooth area to be detected in a wafer, extracting corner points from the image, and calculating the angle value of each corner point; Taking angular points with angle values in a preset interval as first alternative contour angular points, acquiring coordinate values of all the first alternative contour angular points, determining tooth peak areas and tooth valley areas of the to-be-detected target saw-tooth areas according to distribution of all the first alternative contour angular points in a coordinate system, and taking the first alternative contour angular points positioned in the tooth peak areas and the tooth valley areas as detection contour points of the to-be-detected target saw-tooth areas; And matching the detection contour point of the sawtooth region of the target to be detected with the contour point of the sawtooth region of the template, and determining the sawtooth defect type of the target to be detected according to a matching result. Further, the target to be detected is wafer substrate colloid or crystal grain. Further, extracting corner points from the image and calculating an angle value of each corner point includes: for any pixel p in the image, acquiring gray values of 16 pixels on the circumference with the pixel p as a circle center and the radius of 3 pixels, numbering the 16 pixels from 1 to 16 in turn according to clockwise or anticlockwise, and if the gray values of n continuous pixels in the 16 pixels are all larger than the gray value Ip of the pixel p by a preset threshold t, or the gray values of n continuous pixels are all smaller than the gray value Ip of the pixel p by the preset threshold t, and n is larger than the preset threshold M,1< M <16, extracting the pixel p as a corner point; the mapping relation between the n value and the angle value is predetermined, and the angle value of each angular point is obtained according to the n value in the angular point extraction process. Further, the preset interval is set according to the shape of the object to be detected, if the standard saw tooth of the object to be detected is a regular triangle, the preset interval is set to be [60 ° -alpha, 60 ° +alpha ] [300 ° -alpha, 300 ° +alpha ], and alpha is an angle error threshold. Further, the step of taking the first alternative contour corner points located in the tooth peak area and the tooth valley area as the detected target sawtooth area contour points comprises the following steps: And processing the first alternative contour angular points positioned in the tooth peak area and the tooth valley area by adopting a non-maximum value suppression algorithm, removing the non-maximum value to obtain a second alternative contour angular point, and taking the second alternative cont