CN115035122B - Artificial intelligence-based integrated circuit wafer surface defect detection method - Google Patents

Artificial intelligence-based integrated circuit wafer surface defect detection method Download PDF

Info

Publication number
CN115035122B
CN115035122B CN202210965350.6A CN202210965350A CN115035122B CN 115035122 B CN115035122 B CN 115035122B CN 202210965350 A CN202210965350 A CN 202210965350A CN 115035122 B CN115035122 B CN 115035122B
Authority
CN
China
Prior art keywords
wafer
surface image
detected
gradient
area
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202210965350.6A
Other languages
Chinese (zh)
Other versions
CN115035122A (en
Inventor
刘卫卫
金琼洁
欧阳一冉
邹阳春
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ningbo Xinxin Microelectronics Technology Co ltd
Original Assignee
Ningbo Xinxin Microelectronics Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Ningbo Xinxin Microelectronics Technology Co ltd filed Critical Ningbo Xinxin Microelectronics Technology Co ltd
Priority to CN202210965350.6A priority Critical patent/CN115035122B/en
Publication of CN115035122A publication Critical patent/CN115035122A/en
Application granted granted Critical
Publication of CN115035122B publication Critical patent/CN115035122B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/20—Image enhancement or restoration by the use of local operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/40—Image enhancement or restoration by the use of histogram techniques
    • G06T5/70—
    • G06T5/90—
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/761—Proximity, similarity or dissimilarity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20004—Adaptive image processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30148—Semiconductor; IC; Wafer

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Data Mining & Analysis (AREA)
  • Geometry (AREA)
  • Multimedia (AREA)
  • Medical Informatics (AREA)
  • Databases & Information Systems (AREA)
  • Quality & Reliability (AREA)
  • Testing Or Measuring Of Semiconductors Or The Like (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

The invention discloses an integrated circuit wafer surface defect detection method based on artificial intelligence, and relates to the field of defect detection. The method comprises the following steps: acquiring a surface image of a wafer to be detected; coding the gradient variation of the pixel points to obtain a gradient variation chain code of the surface image of the wafer to be detected and the surface image of the standard wafer; calculating gradient similarity to divide the area of the surface image of the wafer to be detected; calculating the pixel characteristic value of each pixel point; calculating the suspected defect degree of each area according to the frequency of each characteristic value; dividing the characteristic histogram according to the suspected defect degree; and obtaining a defect area of the wafer to be detected by performing edge detection after obtaining the enhanced image of the surface of the wafer to be detected. According to the method, the gradient change of the pixel points of the surface image of the wafer to be detected is calculated, the image to be detected is subjected to region division through the gradient change and the gray difference, the contrast of the defect is improved through self-adaption enhancement of different defect degrees, and the precision of defect detection can be effectively improved.

Description

Artificial intelligence-based integrated circuit wafer surface defect detection method
Technical Field
The application relates to the field of defect detection, in particular to an integrated circuit wafer surface defect detection method based on artificial intelligence.
Background
The wafer fabrication process requires many steps such as doping, etching, photolithography, dicing, etc., before final packaging to form an integrated circuit. During these steps, it is important to detect faulty wafers before packaging is completed, since wafers are inevitably damaged and later integrated circuit performance is affected.
There are many surface defects that affect the acceptability of a wafer product, and among the defect types of wafers, unpatterned wafers and patterned wafers are the most common two types of wafers. Wafer surface redundancy, crystal defects, mechanical damage (scratch pattern) are more common defects. The redundancy is a common defect type on the surface of the wafer, and mainly comprises nanometer-sized micro particles, micron-sized dust and residues of related processes. With the smaller and smaller semiconductor size, wafers are manufactured more and more finely, and the problems of difficulty in distinguishing fine defects and non-defects, similarity of defect shapes and background patterns, low recognition accuracy and the like exist in the aspect of defect detection.
In the prior art, a chip to be detected is compared with a standard chip between virtual layers, and when the chip to be detected is obviously different from a standard wafer, the chip to be detected can be detected. In the wafer defect detection, the position of the wafer can be found under the conditions of being blocked or unfilled corner and uneven illumination, and the gray value-based template matching algorithm cannot process the types of interference; therefore, the method has large limitation, is not suitable for detecting objects with non-fixed shapes, and cannot detect the fine defects on the surface of the wafer.
Disclosure of Invention
In order to solve the technical problems, the invention provides an artificial intelligence-based integrated circuit wafer surface defect detection method, which comprises the following steps:
acquiring a surface image of a wafer to be detected;
by using
Figure DEST_PATH_IMAGE002
The operator respectively calculates the gradient variation of each pixel point in the surface image of the wafer to be detected and the standard surface image of the wafer, and codes according to the gradient variation to respectively obtain the gradient variation chain code of the surface image of the wafer to be detected and the gradient variation chain code of the surface image of the standard wafer;
calculating the gradient similarity of the surface image of the wafer to be detected and the surface image of the standard wafer according to the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer;
dividing the area of the surface image of the wafer to be detected according to the gradient similarity and the defect threshold, and taking the gradient similarity corresponding to each area as the abnormal degree of the area;
calculating the pixel characteristic value of each pixel point according to the gray value of each pixel point and the area in which the pixel point is located, and calculating the area characteristic value of each area according to the pixel characteristic value of each pixel point in each area;
calculating the suspected defect degree of each area by using the frequency of the pixel characteristic value of each pixel point in each area;
constructing a characteristic histogram by using the frequency of the characteristic value of each region, and dividing the characteristic histogram by using the suspected defect degree of each region to obtain different types of histogram regions;
carrying out self-adaptive enhancement on the surface image of the wafer to be detected by utilizing the obtained histogram areas of different types to obtain a self-adaptively enhanced surface enhanced image of the wafer to be detected;
and performing edge detection on the enhanced image of the surface of the wafer to be detected to obtain a defect area of the wafer to be detected.
The method for respectively obtaining the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer by coding according to the gradient change comprises the following steps:
setting code elements in eight neighborhood directions;
if the gradient amplitude of the gradient variable quantity of the pixel point in a certain direction of the eight neighborhood directions changes, taking the gradient in the direction as the change direction of the gradient of the pixel point, and coding according to a code element in the gradient change direction;
if the gradient amplitude variation of the gradient variation of the pixel point in the eight neighborhood directions is the same, taking 0 as the code element of the pixel point;
and coding the surface image of the wafer to be detected and the surface image of the standard wafer according to the method to obtain the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer.
The process of calculating the gradient similarity between the surface image of the wafer to be detected and the surface image of the standard wafer according to the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer is as follows:
matching the gradient change chain code of the surface image of the wafer to be detected with the gradient change chain code of the surface image of the standard wafer, calculating the difference degree between the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer according to the matching result, and taking the obtained difference degree as the gradient similarity of the surface image of the wafer to be detected and the surface image of the standard wafer, wherein the calculation formula of the gradient similarity of the surface image of the wafer to be detected and the surface image of the standard wafer is as follows:
Figure DEST_PATH_IMAGE004
in the formula:
Figure DEST_PATH_IMAGE006
representing the gradient similarity between the detected wafer surface image and the standard wafer surface image,
Figure DEST_PATH_IMAGE008
the code length difference between the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer,
Figure DEST_PATH_IMAGE010
the minimum code length difference between the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer is represented,
Figure DEST_PATH_IMAGE012
and the maximum code length difference between the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer is represented.
The method for calculating the code length difference between the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer comprises the following steps:
the calculation formula of the code length difference of the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer is as follows:
Figure DEST_PATH_IMAGE014
in the formula:
Figure DEST_PATH_IMAGE016
the code element length of the shortest abnormal chain code in the gradient change chain codes of the surface image of the wafer to be detected and the gradient change chain codes of the surface image of the standard wafer is shown,
Figure DEST_PATH_IMAGE018
the second of the gradient chain codes representing the image of the surface of the standard wafer
Figure DEST_PATH_IMAGE020
The length of each of the code segments is,
Figure DEST_PATH_IMAGE022
drawing of the surface of wafer to be inspectedImage gradient change chain code
Figure DEST_PATH_IMAGE024
The length of a code segment;
the code segment dividing method comprises the following steps: and respectively carrying out code segment division on the gradient change chain codes of the surface image of the wafer to be detected and the gradient change chain codes of the surface image of the standard wafer by utilizing the matching results of the gradient change chain codes of the surface image of the wafer to be detected and the gradient change chain codes of the surface image of the standard wafer, and dividing the gradient change chain codes into abnormal chain codes and normal chain codes by taking the abnormal chain codes in the matching results as boundaries.
The method for dividing the area of the surface image of the wafer to be detected according to the gradient similarity and the defect threshold comprises the following steps:
dividing the surface image of the wafer to be detected into three types of regions according to the gradient similarity: a non-defective region, a lightly defective region, and a heavily defective region;
and setting a low defect threshold and a high defect threshold, taking an image area corresponding to the gradient similarity greater than or equal to the low defect threshold as a non-defective area, taking an image area corresponding to the gradient similarity between the low defect threshold and the high defect threshold as a light defective area, and taking an image area corresponding to the gradient similarity less than or equal to the high defect threshold as a heavy defective area.
The method for acquiring the low defect threshold and the high defect threshold comprises the following steps:
dividing an initial region according to a gradient similarity histogram of a wafer surface image to be detected and the gradient similarity histogram, establishing a gray level histogram of the initial light defect region, and calculating the defect area according to the frequency of each gray level value and the number of pixel points in the divided initial light defect region
Figure DEST_PATH_IMAGE026
And normal area
Figure DEST_PATH_IMAGE028
Area of defect
Figure 668944DEST_PATH_IMAGE026
The calculation formula of (a) is as follows:
Figure DEST_PATH_IMAGE030
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE032
the number of pixels of the wafer surface image to be detected,
Figure DEST_PATH_IMAGE034
is a gray scale of
Figure DEST_PATH_IMAGE036
The frequency of occurrence of the pixel points of (a),
Figure DEST_PATH_IMAGE038
the gray level corresponding to the first trough in the gray level histogram;
the area of the normal area is the area of the initial mild defect area minus the defect area, and then the defect threshold is low
Figure DEST_PATH_IMAGE040
Comprises the following steps:
Figure DEST_PATH_IMAGE042
and the high defect threshold is obtained by iterating the above process by using the initial severe defect area.
The process of calculating the region characteristic value of each region according to the pixel characteristic value of each pixel point in each region is as follows:
calculating the pixel characteristic value of each pixel point according to the gray values of each pixel point and all the pixel points in the area of the pixel point, wherein the calculation method comprises the following steps: for region
Figure DEST_PATH_IMAGE044
In the mean value of the intermediate gray values and the absolute value of the difference between the gray values of the pixelsMaximum value and region
Figure 845978DEST_PATH_IMAGE044
The average value of the intermediate gray values is subtracted from the minimum value of the absolute value of the difference value of the gray values of all the pixel points, and the obtained difference value is used as the pixel characteristic value of the pixel point;
calculating the area characteristic value of the area according to the pixel characteristic value of each pixel point and the pixel characteristic values of other pixel points in the area where the pixel point is located, wherein the calculation formula is as follows:
Figure DEST_PATH_IMAGE046
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE048
is a pixel point
Figure DEST_PATH_IMAGE050
The region characteristic value of the region in which the region is located,
Figure DEST_PATH_IMAGE052
is a pixel point
Figure 144236DEST_PATH_IMAGE050
The value of the characteristic of the pixel of (c),
Figure DEST_PATH_IMAGE054
is a pixel point
Figure 265776DEST_PATH_IMAGE050
In the area of
Figure 369998DEST_PATH_IMAGE044
The mean value of the gray values of all the pixel points in the image,
Figure DEST_PATH_IMAGE056
is a pixel point
Figure 995014DEST_PATH_IMAGE050
The gray value of (a).
The method for performing self-adaptive enhancement on the surface image of the wafer to be detected by utilizing the obtained histogram areas of different types comprises the following steps:
dividing the feature histogram according to the trough positions in the feature histogram to obtain two parts, and performing image enhancement according to the division result, wherein the method comprises the following steps:
Figure DEST_PATH_IMAGE058
wherein:
Figure DEST_PATH_IMAGE060
the enhanced image is represented by a representation of the image,
Figure DEST_PATH_IMAGE062
indicating area
Figure 995331DEST_PATH_IMAGE044
To middle
Figure DEST_PATH_IMAGE064
The characteristic value of each pixel point is calculated,
Figure DEST_PATH_IMAGE066
is a region
Figure 73009DEST_PATH_IMAGE044
The number of the middle pixel points is increased,
Figure 980922DEST_PATH_IMAGE064
indicating area
Figure 726024DEST_PATH_IMAGE044
To middle
Figure 693980DEST_PATH_IMAGE064
One of the pixel points is selected from the group consisting of,
Figure DEST_PATH_IMAGE068
representing pixel points
Figure 256024DEST_PATH_IMAGE050
The degree of suspected defects in the area of the defect,
Figure DEST_PATH_IMAGE070
、
Figure DEST_PATH_IMAGE072
respectively the abscissa of the feature histogram boundary,
Figure 702049DEST_PATH_IMAGE048
is a pixel point
Figure 301657DEST_PATH_IMAGE050
In the area of
Figure 174935DEST_PATH_IMAGE044
The area feature value of (1).
The calculation formula for calculating the suspected defect degree of each area by using the occurrence frequency of the pixel characteristic value of each pixel point in each area is as follows:
Figure DEST_PATH_IMAGE074
in the formula:
Figure 227205DEST_PATH_IMAGE068
representing pixels
Figure 211341DEST_PATH_IMAGE050
In the area of
Figure 196615DEST_PATH_IMAGE044
The degree of suspected defects of (a) is,
Figure DEST_PATH_IMAGE076
is a region
Figure 240794DEST_PATH_IMAGE044
To middle
Figure DEST_PATH_IMAGE078
The frequency of the seed pixel characteristic value is,
Figure DEST_PATH_IMAGE080
is a region
Figure 780360DEST_PATH_IMAGE044
Of different pixel characteristic values.
Compared with the prior art, the embodiment of the invention has the beneficial effects that: the gradient change of the pixel points of the surface image of the wafer to be detected and the surface image of the standard wafer is calculated, the gradient change chain codes of the surface image of the wafer to be detected and the surface image of the standard wafer are obtained according to the gradient change of the pixel points of the surface image of the wafer to be detected and the surface image of the standard wafer, the gradient similarity of the wafer to be detected and the standard wafer is obtained according to the gradient change chain codes of the surface image of the wafer to be detected and the surface image of the standard wafer, and the gradient similarity reflects the difference between the surface image of the wafer to be detected and the surface image of the standard wafer; the abnormal degree of the surface defect is obtained according to the gradient similarity, the self-adaptive image enhancement is carried out according to the abnormal degree, the defect detection is carried out according to the enhanced image, the defect on the surface of the wafer can be more obvious, and the accuracy of the wafer defect detection is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a flowchart of a system for detecting surface defects of an integrated circuit wafer based on artificial intelligence according to an embodiment of the present invention;
FIG. 2 is a flowchart of the method steps provided by the method for detecting surface defects of an IC wafer based on artificial intelligence in accordance with an embodiment of the present invention;
FIG. 3 is a schematic diagram of an image acquisition environment provided by an artificial intelligence based method for detecting surface defects of an integrated circuit wafer according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a wafer defect provided by an artificial intelligence based integrated circuit wafer surface defect detection method according to an embodiment of the present invention;
FIG. 5 is a code element diagram provided by an artificial intelligence based integrated circuit wafer surface defect detection method according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of the classification of characteristic values provided by the method for detecting surface defects of an integrated circuit wafer based on artificial intelligence according to the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, belong to the protection scope of the present invention.
In the following description, for purposes of explanation and not limitation, specific details are set forth such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. It will be apparent, however, to one skilled in the art that the present application may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.
The terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or to implicitly indicate the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature; in the description of the present embodiment, "a plurality" means two or more unless otherwise specified.
Example 1
An embodiment of the present invention provides an integrated circuit wafer surface defect detection method based on artificial intelligence, as shown in fig. 1 and 2, including:
s101, acquiring a surface image of a wafer to be detected
In this embodiment, the gradient change of the pixel points of the image on the surface of the wafer is analyzed to further determine whether the defect exists and the position of the defect, so that the image on the surface of the wafer to be detected needs to be obtained first for subsequent analysis of the wafer to be detected.
S102, obtaining a gradient change chain code of the surface image of the wafer to be detected and a gradient change chain code of the surface image of the standard wafer
And coding the gradient variation according to the gradient variation of the pixel points of the acquired surface image of the wafer to be detected to obtain a gradient variation chain code, and reflecting the size of the defect area on the surface of the wafer to be detected by comparing the gradient variation chain code with the gradient variation chain code of the standard surface image of the wafer.
S103, calculating the gradient similarity between the surface image of the wafer to be detected and the surface image of the standard wafer
The code segment of the abnormal pixel, namely the abnormal chain code, is obtained by comparing the gradient change chain code of the surface image of the wafer to be detected with the gradient change chain code of the surface image of the standard wafer, the gradient similarity of the surface image of the wafer to be detected and the surface image of the standard wafer is further calculated according to the abnormal chain code, the difference between the gray levels is expressed by the gradient similarity, and the defect area of the surface of the wafer to be detected can be more obviously represented.
S104, carrying out area division on the surface image of the wafer to be detected
The larger the gradient similarity is, the smaller the defect area exists or the defect does not exist; the smaller the gradient similarity is, the larger the defect area exists, the area division is carried out on the surface image of the wafer to be detected according to the gradient similarity, the image enhancement of different degrees is carried out on the divided areas according to the self characteristics of each area, the characteristics of the defect area are more obvious, and the final defect detection efficiency and accuracy can be improved.
S105, calculating the suspected defect degree of each area
The region division is carried out according to the gradient similarity, light defects in the heavy defect region are not divided, and when the contrast of the defect region is not greatly different from that of a normal background region, the statistics of the histogram can be influenced, and brightness errors can occur after the enhancement. Therefore, the image cannot be well enhanced by the traditional gray stretching enhancement, the embodiment performs local self-adaptive image enhancement according to the gray change in the image, calculates the pixel characteristic value of each pixel point according to the gray value of each pixel point, and calculates the suspected defect degree of the region according to the pixel characteristic value of the pixel point in the region.
S106, obtaining the self-adaptively enhanced surface enhanced image of the wafer to be detected
Because only the pixel points of the area with the changed gradient in the wafer surface image to be detected can be described according to the gradient change, the initial area of the suspected defect is obtained, but whether the defect is a true defect or not cannot be distinguished. Therefore, the surface image of the wafer to be detected needs to be enhanced, so that the defect area is more obvious, and the defect on the surface of the wafer can be better detected.
S107, acquiring a defect area of the wafer to be detected
Because the obtained enhanced image of the surface of the wafer to be detected can obviously reflect the suspected defect area of the surface of the wafer to be detected, the more accurate defect area of the surface of the wafer to be detected can be obtained by utilizing edge detection according to the adaptively enhanced image of the surface of the wafer to be detected.
Example 2
The embodiment of the invention provides an integrated circuit wafer surface defect detection method based on artificial intelligence, which comprises the following specific contents as shown in figures 1 and 2:
s201, acquiring surface image of wafer to be detected
Since the wafer is a very tiny integrated circuit device, an industrial high-resolution camera is required to obtain a clear image of the wafer surface when acquiring the image, and the illumination of the light source needs to be uniform, and cannot be too strong or too weak, otherwise the gray scale value of the wafer surface is changed, so that some defects are covered by the light, thereby causing inaccurate detection. A schematic of the environment in which the image is acquired is shown in fig. 3.
1. The collected wafer image is preprocessed, and because the wafer components are tiny, noise is formed when the image is collected, the accuracy of subsequent wafer surface defects is influenced, and therefore denoising processing is carried out. In this embodiment, the noise reduction of the image is performed by using a mean filtering method, and a 3 × 3 filtering window is used.
2. During defect detection, only defect detection is needed to be performed on the region of the wafer surface, after denoising processing is performed, semantic segmentation is performed on the denoised wafer image, background pixels are removed, and only the wafer surface image to be detected including the wafer is obtained, wherein the wafer defect schematic diagram is shown in fig. 4.
The relevant content of the DNN network is as follows:
the data set used is an RGB image data set of the wafer to be detected acquired in a overlooking mode.
The pixels needing to be segmented are divided into two types, namely the labeling process of the corresponding labels of the training set is as follows: in the semantic label of the single channel, the label of the pixel at the corresponding position belonging to the background class is 0, and the label of the pixel belonging to the wafer area is 1.
The task of the network is to classify, and all the used loss functions are cross entropy loss functions.
The 0-1 mask image obtained by semantic segmentation is multiplied by the original image, and the obtained image only contains the image of the wafer area, so that the interference of the background is removed.
And deducing through the trained DNN according to the acquired RGB image to obtain an image of the wafer region in the image, and performing graying processing on the image of the wafer region to obtain a corresponding wafer surface image to be detected.
S202, obtaining the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer
1. Calculating the gradient variation of each pixel point in the surface image of the wafer to be detected and the standard wafer surface image
The uniformity of pixel points in the surface image of the wafer to be detected is represented according to the gradient direction and the amplitude value in the surface image of the wafer to be detected, then the surface image of the wafer to be detected is subjected to gradient change coding, and the abnormal code segment of the surface image of the wafer to be detected is obtained by matching with the gradient change chain code of the surface image of the standard wafer.
In order to obtain gradient change chain codes, the horizontal gradient and the vertical gradient of each pixel point need to be calculated first, and a kernel with the size of 1 can be used
Figure DEST_PATH_IMAGE082
Operators to obtain the same result. About
Figure 302608DEST_PATH_IMAGE082
The details of calculating the gradient are well known and will not be described herein.
Then recalculate
Figure DEST_PATH_IMAGE084
And
Figure DEST_PATH_IMAGE086
the resultant gradient of the directional gradient, including magnitude and direction:
Figure DEST_PATH_IMAGE088
Figure DEST_PATH_IMAGE090
in the formula (I), the compound is shown in the specification,
Figure DEST_PATH_IMAGE092
which is representative of the magnitude of the gradient,
Figure DEST_PATH_IMAGE094
which is representative of the direction of the gradient,
Figure DEST_PATH_IMAGE096
wafer table for indicating to be detectedSerial numbers of pixel points of the face image. The gradient direction will take the absolute value, so the resulting angular range is
Figure DEST_PATH_IMAGE098
. At each pixel point, the gradient has a magnitude and a direction.
Figure 283334DEST_PATH_IMAGE084
The directional gradient map will emphasize the vertical edge features,
Figure 763994DEST_PATH_IMAGE086
the directional gradient map emphasizes horizontal edge features. This allows useful features (contours) to be preserved, with irrelevant unimportant information removed.
Figure DEST_PATH_IMAGE100
Representing the gradient amplitude and the change vector of the gradient direction of the pixel point on the surface image of the wafer to be detected, representing by the vector, and obtaining the gradient change quantity of the pixel point on the surface image of the wafer to be detected according to the gradient amplitude and the gradient direction:
Figure DEST_PATH_IMAGE102
，
Figure DEST_PATH_IMAGE104
and expressing the gradient variable quantity to obtain the gradient variable quantity of all pixel points on the surface image of the wafer to be detected. The gradient variation of the standard wafer surface image is obtained by the method
Figure DEST_PATH_IMAGE106
The modulus is
Figure DEST_PATH_IMAGE108
And coding the gradient variation of the standard wafer image and the gradient variation of the wafer image to be detected according to the gradient variation of all the pixel points on the surface image of the standard wafer.
2. Acquiring gradient change chain codes of surface images of wafers to be detected and gradient change chain codes of surface images of standard wafers
And obtaining the gradient variation of the standard wafer surface image and the gradient variation of the wafer surface image to be detected according to the steps, then coding the obtained gradient variations, and obtaining the gradient similarity of the two images according to the difference degree between the chain code of the gradient variation of the standard wafer surface image and the chain code of the gradient variation of the wafer surface image to be detected.
The encoding according to the gradient change amount is as follows:
setting the code elements in the eight neighborhood directions, wherein the code element diagram is shown in FIG. 5;
if the gradient amplitude of the gradient variable quantity of the pixel point in a certain direction of the eight neighborhood directions changes, taking the gradient in the direction as the change direction of the gradient of the pixel point, and coding according to a code element in the gradient change direction;
if the gradient amplitude variation of the gradient variation of the pixel point in the eight neighborhood directions is the same, taking 0 as the code element of the pixel point;
and coding the surface image of the wafer to be detected and the surface image of the standard wafer according to the method to obtain the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer.
Coding gradient change values of pixel points in the wafer surface image to be detected and the standard wafer surface image, and coding the images from left to right and from top to bottom, for example:
the gradient change chain code of the surface image of the wafer to be detected is { [1] < 0,1,3, D,2, C, D,4,3, A, B,1,4, 0}, wherein [1] of the chain code represents an identification code, [1] represents the wafer to be detected, [0] represents a standard wafer, and {0} represents that the gradient change in each direction of the pixel point is the same.
The gradient change chain code of the standard wafer surface image is obtained by encoding the standard wafer surface image through the method.
S203, calculating the gradient similarity between the surface image of the wafer to be detected and the surface image of the standard wafer
Comparing the gradient change chain code of the surface image of the wafer to be detected with the gradient change chain code of the surface image of the standard wafer, and when the gradient change chain code of the surface image of the wafer to be detected is completely matched with the gradient change chain code of the surface image of the standard wafer, indicating that no defect exists on the surface of the wafer to be detected; when the gradient change chain code of the to-be-detected wafer surface image is different from the gradient change chain code of the standard wafer surface image, calculating the difference degree between the gradient change chain code of the to-be-detected wafer surface image and the gradient change chain code of the standard wafer surface image, and taking the obtained difference degree as the gradient similarity of the to-be-detected wafer surface image and the standard wafer surface image.
The calculation formula of the gradient similarity between the surface image of the wafer to be detected and the surface image of the standard wafer is as follows:
Figure DEST_PATH_IMAGE004A
in the formula:
Figure DEST_PATH_IMAGE110
representing the gradient similarity between the detected wafer surface image and the standard wafer surface image,
Figure DEST_PATH_IMAGE112
the code length difference between the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer,
Figure DEST_PATH_IMAGE114
the minimum code length difference between the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer is shown,
Figure DEST_PATH_IMAGE116
and the maximum code length difference between the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer is represented.
The calculation formula of the code length difference of the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer is as follows:
Figure DEST_PATH_IMAGE118
in the formula:
Figure DEST_PATH_IMAGE120
the code element length of the shortest abnormal chain code in the gradient change chain codes of the surface image of the wafer to be detected and the gradient change chain codes of the surface image of the standard wafer is shown,
Figure DEST_PATH_IMAGE122
second of gradient change chain code representing image of standard wafer surface
Figure DEST_PATH_IMAGE124
The length of each of the code segments is,
Figure DEST_PATH_IMAGE126
the second step of the gradient change chain code representing the surface image of the wafer to be detected
Figure DEST_PATH_IMAGE128
The length of a code segment;
the code segment dividing method comprises the following steps: and respectively carrying out code segment division on the gradient change chain code of the wafer surface image to be detected and the gradient change chain code of the standard wafer surface image by utilizing the matching result of the gradient change chain code of the wafer surface image to be detected and the gradient change chain code of the standard wafer surface image, and dividing the gradient change chain code into an abnormal chain code and a normal chain code by taking the abnormal chain code in the matching result as a boundary.
S204, carrying out region division on the surface image of the wafer to be detected
The gradient similarity between the standard wafer and the wafer to be detected is obtained through the steps and is used for representing the change of the gray value of the pixel point on the surface of the wafer, and the larger the gradient similarity is, the smaller the defect area exists or the defect does not exist; the smaller the gradient similarity is, the larger the defect area exists, the size of the defect area is divided according to the gradient similarity, and three conditions exist, namely a non-defect area, a light defect area and a heavy defect area.
Setting a low defect threshold
Figure DEST_PATH_IMAGE130
And high defect threshold
Figure DEST_PATH_IMAGE132
The image area corresponding to the gradient similarity greater than or equal to the low defect threshold is set as a non-defective area, the image area corresponding to the gradient similarity between the low defect threshold and the high defect threshold is set as a light-defective area, and the image area corresponding to the gradient similarity less than or equal to the high defect threshold is set as a heavy-defective area. And taking the gradient similarity of each region as the abnormal degree of the region.
Low defect threshold
Figure 335396DEST_PATH_IMAGE130
And high defect threshold
Figure 192493DEST_PATH_IMAGE132
The acquisition method comprises the following steps:
dividing an initial region according to a gradient similarity histogram of a wafer surface image to be detected and the gradient similarity histogram, establishing a gray level histogram of an initial light defect region, and calculating the defect area according to the frequency of each gray level value and the number of pixel points in the divided initial light defect region
Figure DEST_PATH_IMAGE134
And normal area
Figure DEST_PATH_IMAGE136
Area of defect
Figure 355622DEST_PATH_IMAGE134
The calculation formula of (c) is as follows:
Figure DEST_PATH_IMAGE030A
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE138
the number of pixels of the wafer surface image to be detected,
Figure DEST_PATH_IMAGE140
is a gray scale of
Figure DEST_PATH_IMAGE142
The frequency of occurrence of the pixel points of (a),
Figure DEST_PATH_IMAGE144
the gray level corresponding to the first trough in the gray level histogram;
the area of the normal area is the area of the initial mild defect area minus the defect area, and then the defect threshold is low
Figure 679287DEST_PATH_IMAGE130
Comprises the following steps:
Figure DEST_PATH_IMAGE042A
the high defect threshold is obtained by iterating the above process with the initial heavily defective region.
S205, calculating the suspected defect degree of each area
The region division is performed according to the gradient similarity, the light defects in the heavy defect region are not divided, and when the contrast of the defect region is not much different from that of the normal background region, the statistics of the histogram can be affected, and a brightness error can occur after the enhancement. Therefore, the image cannot be well enhanced by the traditional gray stretching enhancement, the embodiment performs local self-adaptive image enhancement according to the gray value change in the image, calculates the pixel characteristic value of each pixel point according to the gray value of each pixel point, calculates the suspected defect degree of the area according to the pixel characteristic value of the pixel point in the area, and re-assigns the pixel points in the area according to the suspected defect degree to obtain the characteristic value of the pixel point with more obvious difference.
1. Calculating the characteristic value of each pixel point in each region
Calculating the pixel characteristic value of each pixel point according to the gray values of each pixel point and all the pixel points in the area of the pixel point, wherein the calculation method comprises the following steps: for region
Figure DEST_PATH_IMAGE146
Maximum value and area in gray value difference absolute values of intermediate gray value mean value and each pixel point
Figure 193445DEST_PATH_IMAGE146
And (3) making a difference between the average value of the intermediate gray values and the minimum value in the absolute value of the gray value difference of each pixel point, and taking the obtained difference as the pixel characteristic value of the pixel point, wherein the calculation formula is as follows:
Figure DEST_PATH_IMAGE148
in the formula:
Figure DEST_PATH_IMAGE150
is a pixel point
Figure DEST_PATH_IMAGE152
The value of the characteristic of the pixel of (c),
Figure DEST_PATH_IMAGE154
is a pixel point
Figure 526337DEST_PATH_IMAGE152
In the region of
Figure 809551DEST_PATH_IMAGE146
The mean value of the gray values of all the pixel points,
Figure DEST_PATH_IMAGE156
is a pixel point
Figure 632013DEST_PATH_IMAGE152
The gray value of (a).
2. Calculating a region feature value of each region
Calculating the characteristic value of each pixel point according to the pixel characteristic value of each pixel point and the pixel characteristic values of other pixel points in the area of the pixel point, wherein the calculation formula is as follows:
Figure DEST_PATH_IMAGE158
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE160
is a pixel point
Figure 102309DEST_PATH_IMAGE152
The value of the characteristic of the area in which the area is located,
Figure 301209DEST_PATH_IMAGE150
is a pixel point
Figure 438929DEST_PATH_IMAGE152
The value of the characteristic of the pixel of (a),
Figure 432293DEST_PATH_IMAGE154
is a pixel point
Figure 186622DEST_PATH_IMAGE152
In the region of
Figure 658055DEST_PATH_IMAGE146
The mean value of the gray values of all the pixel points in the image,
Figure 915861DEST_PATH_IMAGE156
is a pixel point
Figure 345705DEST_PATH_IMAGE152
The gray value of (a).
3. Calculating the suspected defect degree of each area
Figure DEST_PATH_IMAGE074A
In the formula:
Figure DEST_PATH_IMAGE162
representing pixel points
Figure 525014DEST_PATH_IMAGE152
In the region of
Figure 800138DEST_PATH_IMAGE146
The degree of suspected defect of (a) is,
Figure 178029DEST_PATH_IMAGE076
is a region
Figure 247616DEST_PATH_IMAGE146
To middle
Figure DEST_PATH_IMAGE164
The frequency of the seed characteristic value is determined,
Figure DEST_PATH_IMAGE166
is a region
Figure 445380DEST_PATH_IMAGE146
Of different eigenvalues.
S206, obtaining the self-adaptive enhanced wafer surface enhanced image to be detected
Because only the pixel points of the gradient-changed area in the wafer surface image to be detected can be described according to the gradient change, the initial area of the suspected defect is obtained, but whether the defect is true or not cannot be distinguished. Therefore, the surface image of the wafer to be detected needs to be enhanced, so that the defect area is more obvious, and the defect on the surface of the wafer can be better detected.
Dividing the characteristic value according to the obtained characteristic value, and counting the histogram of the number of the characteristic value to obtain a more refined enhanced imageBecause the number of the defect areas is uncertain, the histogram contains uncertain number of peak values, and the characteristic value histogram is divided according to the characteristic value grades:
Figure DEST_PATH_IMAGE168
representing the boundaries of the partition, typically at the bottom of the histogram.
Figure DEST_PATH_IMAGE170
The number of the limits is indicated and,
Figure 990106DEST_PATH_IMAGE170
the boundaries can divide the image into
Figure DEST_PATH_IMAGE172
And (4) section. In this embodiment, the histogram is divided into two parts according to different feature value levels, and a schematic diagram of feature value level division is shown in fig. 6.
The enhancement process is as follows:
Figure DEST_PATH_IMAGE058A
wherein:
Figure DEST_PATH_IMAGE174
which represents the image after the enhancement, is,
Figure DEST_PATH_IMAGE176
indicating area
Figure 629029DEST_PATH_IMAGE146
To middle
Figure DEST_PATH_IMAGE178
The characteristic value of each pixel point is calculated,
Figure DEST_PATH_IMAGE180
is a region
Figure 869517DEST_PATH_IMAGE146
The number of the middle pixel points is increased,
Figure 820156DEST_PATH_IMAGE178
indicating area
Figure 437082DEST_PATH_IMAGE146
To middle
Figure 523987DEST_PATH_IMAGE178
The number of the pixel points is one,
Figure 466535DEST_PATH_IMAGE162
representing pixel points
Figure 638890DEST_PATH_IMAGE152
The degree of suspected defects in the area of the defect,
Figure DEST_PATH_IMAGE182
、
Figure DEST_PATH_IMAGE184
respectively the abscissa of the feature histogram boundary,
Figure 528349DEST_PATH_IMAGE160
is a pixel point
Figure 918DEST_PATH_IMAGE152
In the area of
Figure 583209DEST_PATH_IMAGE146
The area feature value of (1).
S207, acquiring the defect area of the wafer to be detected
The obtained enhanced image of the surface of the wafer to be detected contains defect areas with different degrees and different sizes, the suspected defect area of the enhanced image of the surface of the wafer to be detected is very obvious, and the enhanced image of the surface of the wafer to be detected is utilized
Figure DEST_PATH_IMAGE186
The operator carries out edge detection to obtain the surface of the wafer to be detectedAnd completing the defect detection of the wafer to be detected in the defect area.
And recycling or secondarily processing the wafer to be detected with the defects according to the defect area obtained by the defect detection and the defect position defect type of the defect area, so that the wafer to be delivered can reach the delivery quality standard.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (9)

1. An integrated circuit wafer surface defect detection method based on artificial intelligence is characterized by comprising the following steps:
acquiring a surface image of a wafer to be detected;
by using
Figure DEST_PATH_IMAGE001
The operator respectively calculates the gradient variation of each pixel point in the surface image of the wafer to be detected and the standard surface image of the wafer, and codes according to the gradient variation to respectively obtain the gradient variation chain code of the surface image of the wafer to be detected and the gradient variation chain code of the surface image of the standard wafer;
calculating the gradient similarity of the surface image of the wafer to be detected and the surface image of the standard wafer according to the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer;
dividing the area of the surface image of the wafer to be detected according to the gradient similarity and the defect threshold, and taking the gradient similarity corresponding to each area as the abnormal degree of the area;
calculating the pixel characteristic value of each pixel point according to the gray value of each pixel point in each pixel point and the area where the pixel point is located, and calculating the area characteristic value of each area according to the pixel characteristic value of each pixel point in each area;
calculating the suspected defect degree of each area by using the frequency of the pixel characteristic value of each pixel point in each area;
constructing a characteristic histogram by using the frequency of the characteristic value of each region, and dividing the characteristic histogram by using the suspected defect degree of each region to obtain different types of histogram regions;
carrying out self-adaptive enhancement on the surface image of the wafer to be detected by utilizing the obtained histogram areas of different types to obtain a self-adaptively enhanced surface enhanced image of the wafer to be detected;
and carrying out edge detection on the enhanced image of the surface of the wafer to be detected to obtain a defect area of the wafer to be detected.
2. The method for detecting the surface defects of the integrated circuit wafer based on the artificial intelligence of claim 1, wherein the method for respectively obtaining the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer by encoding according to the gradient change amount comprises the following steps:
setting code elements in eight neighborhood directions;
if the gradient amplitude of the gradient variable quantity of the pixel point in a certain direction of the eight neighborhood directions changes, taking the gradient in the direction as the change direction of the gradient of the pixel point, and encoding according to a code element in the gradient change direction;
if the gradient amplitude variation of the gradient variation of the pixel point in the eight neighborhood directions is the same, taking 0 as the code element of the pixel point;
and coding the surface image of the wafer to be detected and the surface image of the standard wafer according to the method to obtain the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer.
3. The method for detecting surface defects of an integrated circuit wafer based on artificial intelligence as claimed in claim 1, wherein the process of calculating the gradient similarity between the surface image of the wafer to be detected and the surface image of the standard wafer according to the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer is as follows:
matching the gradient change chain code of the surface image of the wafer to be detected with the gradient change chain code of the surface image of the standard wafer, calculating the difference degree between the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer according to the matching result, and taking the obtained difference degree as the gradient similarity of the surface image of the wafer to be detected and the surface image of the standard wafer, wherein the calculation formula of the gradient similarity of the surface image of the wafer to be detected and the surface image of the standard wafer is as follows:
Figure DEST_PATH_IMAGE003
in the formula:
Figure 801068DEST_PATH_IMAGE004
representing the gradient similarity between the detected wafer surface image and the standard wafer surface image,
Figure DEST_PATH_IMAGE005
the code length difference between the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer,
Figure 187050DEST_PATH_IMAGE006
the minimum code length difference between the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer is shown,
Figure DEST_PATH_IMAGE007
and the maximum code length difference between the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer is represented.
4. The method as claimed in claim 3, wherein the calculating method of the code length difference between the gradient change chain code of the image on the surface of the wafer to be detected and the gradient change chain code of the image on the surface of the standard wafer comprises:
the calculation formula of the code length difference of the gradient change chain code of the surface image of the wafer to be detected and the gradient change chain code of the surface image of the standard wafer is as follows:
Figure DEST_PATH_IMAGE009
in the formula:
Figure 435628DEST_PATH_IMAGE010
the code element length of the shortest abnormal chain code in the gradient change chain codes of the surface image of the wafer to be detected and the gradient change chain codes of the surface image of the standard wafer is shown,
Figure DEST_PATH_IMAGE011
second of gradient change chain code representing image of standard wafer surface
Figure 562329DEST_PATH_IMAGE012
The length of a segment of a code is,
Figure DEST_PATH_IMAGE013
the second of the gradient change chain codes representing the surface image of the wafer to be detected
Figure 111122DEST_PATH_IMAGE014
The length of a code segment;
the code segment dividing method comprises the following steps: and respectively carrying out code segment division on the gradient change chain code of the wafer surface image to be detected and the gradient change chain code of the standard wafer surface image by utilizing the matching result of the gradient change chain code of the wafer surface image to be detected and the gradient change chain code of the standard wafer surface image, and dividing the gradient change chain code into an abnormal chain code and a normal chain code by taking the abnormal chain code in the matching result as a boundary.
5. The method as claimed in claim 1, wherein the method for dividing the regions of the wafer surface image to be detected according to the gradient similarity and the defect threshold comprises:
dividing the surface image of the wafer to be detected into three types of areas according to the gradient similarity: a non-defective region, a lightly defective region, and a heavily defective region;
and setting a low defect threshold and a high defect threshold, taking an image area corresponding to the gradient similarity greater than or equal to the low defect threshold as a non-defective area, taking an image area corresponding to the gradient similarity between the low defect threshold and the high defect threshold as a light defective area, and taking an image area corresponding to the gradient similarity less than or equal to the high defect threshold as a heavy defective area.
6. The method of claim 5, wherein the obtaining of the low defect threshold and the high defect threshold comprises:
dividing an initial region according to a gradient similarity histogram of a wafer surface image to be detected and the gradient similarity histogram, establishing a gray level histogram of the initial light defect region, and calculating the defect area according to the frequency of each gray level value and the number of pixel points in the divided initial light defect region
Figure DEST_PATH_IMAGE015
And normal area
Figure 668005DEST_PATH_IMAGE016
Area of defect
Figure 669459DEST_PATH_IMAGE015
The calculation formula of (a) is as follows:
Figure 602780DEST_PATH_IMAGE018
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE019
the number of pixels of the wafer surface image to be detected,
Figure 6079DEST_PATH_IMAGE020
is a gray scale of
Figure DEST_PATH_IMAGE021
The frequency of occurrence of the pixel points of (a),
Figure 733864DEST_PATH_IMAGE022
the gray level corresponding to the first trough in the gray histogram;
the area of the normal region is the area of the initial mild defect region minus the defect area, and then the defect threshold is low
Figure DEST_PATH_IMAGE023
Comprises the following steps:
Figure DEST_PATH_IMAGE025
the high defect threshold is obtained by iterating the above process with the initial heavily defective region.
7. The method as claimed in claim 1, wherein the calculating of the region feature value of each region according to the pixel feature value of each pixel in each region comprises:
calculating the pixel characteristic value of each pixel point according to the gray values of each pixel point and all the pixel points in the area of the pixel point, wherein the calculation method comprises the following steps: to the region
Figure 957035DEST_PATH_IMAGE026
Maximum value and area in gray value difference absolute values of intermediate gray value mean value and each pixel point
Figure 428468DEST_PATH_IMAGE026
The average value of the intermediate gray values is subtracted from the minimum value of the absolute value of the difference value of the gray values of all the pixel points, and the obtained difference value is used as the pixel characteristic value of the pixel point;
calculating the area characteristic value of the area according to the pixel characteristic value of each pixel point and the pixel characteristic values of other pixel points in the area where the pixel point is located, wherein the calculation formula is as follows:
Figure 686274DEST_PATH_IMAGE028
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE029
is a pixel point
Figure 850539DEST_PATH_IMAGE030
The value of the characteristic of the area in which the area is located,
Figure DEST_PATH_IMAGE031
is a pixel point
Figure 826585DEST_PATH_IMAGE030
The value of the characteristic of the pixel of (a),
Figure 836129DEST_PATH_IMAGE032
is a pixel point
Figure 214021DEST_PATH_IMAGE030
In the region of
Figure 549187DEST_PATH_IMAGE026
The mean value of the gray values of all the pixel points in the image,
Figure DEST_PATH_IMAGE033
is a pixel point
Figure 746950DEST_PATH_IMAGE030
The gray value of (a).
8. The method as claimed in claim 1, wherein the method for adaptively enhancing the surface image of the wafer to be detected by using the histogram regions of different types comprises:
dividing the feature histogram according to the trough positions in the feature histogram to obtain two parts, and performing image enhancement according to the division result, wherein the method comprises the following steps:
Figure DEST_PATH_IMAGE035
wherein:
Figure 560186DEST_PATH_IMAGE036
which represents the image after the enhancement, is,
Figure DEST_PATH_IMAGE037
indicating area
Figure 792584DEST_PATH_IMAGE026
To middle
Figure 298652DEST_PATH_IMAGE038
The characteristic value of each pixel point is calculated,
Figure DEST_PATH_IMAGE039
is a region
Figure 983711DEST_PATH_IMAGE026
The number of the middle pixel points is larger,
Figure 600637DEST_PATH_IMAGE038
indicating area
Figure 687542DEST_PATH_IMAGE026
To middle
Figure 630090DEST_PATH_IMAGE038
The number of the pixel points is one,
Figure 802445DEST_PATH_IMAGE040
representing pixel points
Figure 223062DEST_PATH_IMAGE030
The extent of suspected defects in the area of the defect,
Figure DEST_PATH_IMAGE041
、
Figure 164473DEST_PATH_IMAGE042
respectively the abscissa of the feature histogram boundary,
Figure 277923DEST_PATH_IMAGE029
is a pixel point
Figure 937574DEST_PATH_IMAGE030
In the region of
Figure 896303DEST_PATH_IMAGE026
The area feature value of (2).
9. The method as claimed in claim 1, wherein the calculation formula for calculating the suspected defect level of each region using the frequency of occurrence of the pixel feature value of each pixel point in each region is as follows:
Figure 957800DEST_PATH_IMAGE044
in the formula:
Figure 976572DEST_PATH_IMAGE040
representing pixels
Figure 123519DEST_PATH_IMAGE030
In the region of
Figure 151518DEST_PATH_IMAGE026
The degree of suspected defects of (a) is,
Figure DEST_PATH_IMAGE045
is a region
Figure 801942DEST_PATH_IMAGE026
To middle
Figure 991615DEST_PATH_IMAGE046
The frequency of the seed pixel characteristic value is,
Figure DEST_PATH_IMAGE047
is a region
Figure 357350DEST_PATH_IMAGE026
The types of different pixel characteristic values.
CN202210965350.6A 2022-08-12 2022-08-12 Artificial intelligence-based integrated circuit wafer surface defect detection method Active CN115035122B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210965350.6A CN115035122B (en) 2022-08-12 2022-08-12 Artificial intelligence-based integrated circuit wafer surface defect detection method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210965350.6A CN115035122B (en) 2022-08-12 2022-08-12 Artificial intelligence-based integrated circuit wafer surface defect detection method

Publications (2)

Publication Number Publication Date
CN115035122A CN115035122A (en) 2022-09-09
CN115035122B true CN115035122B (en) 2022-11-11

Family

ID=83130743

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210965350.6A Active CN115035122B (en) 2022-08-12 2022-08-12 Artificial intelligence-based integrated circuit wafer surface defect detection method

Country Status (1)

Country Link
CN (1) CN115035122B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115661105B (en) * 2022-11-05 2023-06-20 东莞市蒂安斯实业有限公司 Automobile model vision detection method based on artificial intelligence
CN116758056B (en) * 2023-08-09 2023-12-26 广东乾威精密连接器有限公司 Electrical terminal production defect detection method
CN116912242B (en) * 2023-09-12 2023-12-15 青岛天仁微纳科技有限责任公司 Method for positioning defects of nanoimprint wafer
CN116958125B (en) * 2023-09-18 2023-12-26 惠州市鑫晖源科技有限公司 Electronic contest host power supply element defect visual detection method based on image processing
CN117114420B (en) * 2023-10-17 2024-01-05 南京启泰控股集团有限公司 Image recognition-based industrial and trade safety accident risk management and control system and method
CN117252878B (en) * 2023-11-17 2024-02-02 青岛天仁微纳科技有限责任公司 Image defect detection method of nano-imprint mold

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114219805A (en) * 2022-02-22 2022-03-22 武汉旺佳玻璃制品有限公司 Intelligent detection method for glass defects
CN114581428A (en) * 2022-03-13 2022-06-03 江苏涂博士新材料有限公司 Powder coating adhesion degree detection method based on optical means
CN114757949A (en) * 2022-06-15 2022-07-15 济宁市海富电子科技有限公司 Wire and cable defect detection method and system based on computer vision

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101214806B1 (en) * 2010-05-11 2012-12-24 가부시키가이샤 사무코 Apparatus and method for defect inspection of wafer
JP2013160629A (en) * 2012-02-06 2013-08-19 Hitachi High-Technologies Corp Defect inspection method, defect inspection apparatus, program, and output unit
CN114723701B (en) * 2022-03-31 2023-04-18 厦门力和行自动化有限公司 Gear defect detection method and system based on computer vision

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114219805A (en) * 2022-02-22 2022-03-22 武汉旺佳玻璃制品有限公司 Intelligent detection method for glass defects
CN114581428A (en) * 2022-03-13 2022-06-03 江苏涂博士新材料有限公司 Powder coating adhesion degree detection method based on optical means
CN114757949A (en) * 2022-06-15 2022-07-15 济宁市海富电子科技有限公司 Wire and cable defect detection method and system based on computer vision

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
Crack detection in magnetic tile images using nonsubsampled shearlet transform and envelope gray level gradient;Chengli Yang等;《Optics & Laser Technology》;20170501;第90卷;第7-17页 *
Surface Defect Detection Based on Gradient LBP;Xiaojing Liu等;《2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC)》;20181018;第133-137页 *
铁轨表面缺陷的机器视觉检测系统;吴迪;《工程科技Ⅱ辑》;20180715(第07期);第7-58页 *
面向带钢表面缺陷图像的特征提取算法研究;周青山;《信息科技辑》;20180115(第01期);第7-57页 *

Also Published As

Publication number Publication date
CN115035122A (en) 2022-09-09

Similar Documents

Publication Publication Date Title
CN115035122B (en) Artificial intelligence-based integrated circuit wafer surface defect detection method
Zhou et al. Automated visual inspection of glass bottle bottom with saliency detection and template matching
JP4095860B2 (en) Defect inspection method and apparatus
CN111383209B (en) Unsupervised flaw detection method based on full convolution self-encoder network
CN109978839B (en) Method for detecting wafer low-texture defects
CN115147414B (en) Surface breakdown defect detection method for bipolar power transistor
CN115020267B (en) Semiconductor surface defect detection method
CN115861291B (en) Chip circuit board production defect detection method based on machine vision
CN115311282B (en) Wafer defect detection method based on image enhancement
CN115791822A (en) Visual detection algorithm and detection system for wafer surface defects
CN115205223B (en) Visual inspection method and device for transparent object, computer equipment and medium
CN116883408B (en) Integrating instrument shell defect detection method based on artificial intelligence
CN113252695B (en) Plastic packaging film defect detection method and detection device based on image processing
CN115775250A (en) Golden finger circuit board defect rapid detection system based on digital image analysis
CN115690670A (en) Intelligent identification method and system for wafer defects
CN116309565A (en) High-strength conveyor belt deviation detection method based on computer vision
Ghosh et al. Counterfeit IC detection by image texture analysis
CN117094975A (en) Method and device for detecting surface defects of steel and electronic equipment
CN114693651A (en) Rubber ring flow mark detection method and device based on image processing
CN115713480A (en) Integrated circuit metal packaging deep learning defect detection method
CN116523922B (en) Bearing surface defect identification method
CN116485779B (en) Adaptive wafer defect detection method and device, electronic equipment and storage medium
WO2005008223A2 (en) Methods for defect detection and process monitoring based on sem images
CN116228780A (en) Silicon wafer defect detection method and system based on computer vision
CN115239595A (en) Method for detecting qualification of two-dimensional code of packaging printed matter

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant