CN-119757348-B - Defect detection method and related device

CN119757348BCN 119757348 BCN119757348 BCN 119757348BCN-119757348-B

Abstract

The embodiment of the invention provides a defect detection method and a related device, which are used for rapidly acquiring the position coordinates of a defect relative to a second reference coordinate system according to the first position coordinates of the defect relative to the first reference coordinate system. The method comprises the steps of obtaining a first position coordinate of a defect in a detection object relative to a first reference coordinate system, inputting the first position coordinate of the defect into a coordinate conversion neural network model to obtain a second position coordinate of the defect output by the coordinate conversion neural network model relative to a second reference coordinate system, wherein the second reference coordinate system is arranged in a bearing device of the detection object, the coordinate conversion neural network model is used for converting the first position coordinate into the second position coordinate, and detecting the defect according to the second position coordinate of the defect.

Inventors

CHEN LU
MA YANZHONG
WEN LANGFENG
ZHANG SONG

Assignees

深圳中科飞测科技股份有限公司

Dates

Publication Date: 20260508
Application Date: 20230928

Claims (9)

1. A defect detection method, comprising: acquiring a plurality of first position coordinates of a plurality of defects in a detection object relative to a first reference coordinate system, wherein the first reference coordinate system is arranged in the detection object; Inputting a plurality of first position coordinates of the plurality of defects into a coordinate conversion neural network model to obtain a plurality of second position coordinates of the plurality of defects, which are output by the coordinate conversion neural network model, relative to a second reference coordinate system, wherein the second reference coordinate system is arranged in a bearing device of the detection object, and the coordinate conversion neural network model is used for realizing conversion from the first position coordinates to the second position coordinates; detecting the defects according to a plurality of second position coordinates of the defects; The coordinate conversion neural network model comprises a convolution module, a batch normalization layer, a long-period memory network layer, a full-connection layer and a shape changing layer; Inputting the first position coordinates of the defects into a coordinate conversion neural network model to obtain second position coordinates of the defects output by the coordinate conversion neural network model relative to a second reference coordinate system, wherein the first position coordinates comprise: inputting a plurality of first position coordinates of the plurality of defects to the convolution module to extract coordinate features of the plurality of first position coordinates; inputting the coordinate features of the plurality of first position coordinates to the batch normalization layer to normalize the coordinate features of the plurality of first position coordinates; inputting the normalized coordinate features of the plurality of first position coordinates into the long-short-period memory network to be used for ignoring first part features of the coordinate features of the plurality of first position coordinates and emphasizing second part features of the coordinate features of the plurality of first position coordinates so as to obtain output of the long-short-period memory network layer; Inputting the output of the long-short-term memory network layer to the full-connection layer for fitting the coordinate features of the plurality of first position coordinates; And inputting the coordinate features of the fitted first position coordinates to the shape changing layer to obtain a preset number of second position coordinates, wherein the preset number is equal to the number of the first position coordinates input to the convolution module.
2. The method of claim 1, wherein inputting the first plurality of location coordinates of the plurality of defects into a coordinate conversion neural network model to obtain the second plurality of location coordinates of the plurality of defects output by the coordinate conversion neural network model relative to a second reference frame comprises: and respectively inputting the first position coordinates of the defects into a coordinate conversion neural network model by utilizing multithreading to obtain a plurality of second position coordinates of the defects output by the coordinate conversion neural network model relative to the second reference coordinate system.
3. The method of claim 1, wherein prior to inputting the first plurality of location coordinates of the plurality of defects into the coordinate conversion neural network model, the method further comprises: The method comprises the steps of obtaining a training sample, wherein the training sample comprises a plurality of first position coordinates of a plurality of defects in a plurality of detection objects relative to a first reference coordinate system and a plurality of second position coordinates of the plurality of defects relative to a second reference coordinate system, and the first position coordinates and the second position coordinates of each defect are in one-to-one correspondence; Inputting a plurality of first position coordinates of the defects into the coordinate conversion neural network model to obtain a plurality of corresponding third position coordinates output by the coordinate conversion neural network model; Calculating losses between a third position coordinate and a plurality of second position coordinates of the defects according to a preset loss function; training the coordinate conversion neural network model by using the loss and back propagation algorithm until the loss meets a target loss.
4. The method of claim 3, wherein prior to inputting the first plurality of location coordinates of the plurality of defects into the coordinate conversion neural network model, the method further comprises: And performing data cleaning on the training sample, wherein the data cleaning at least comprises at least one of deleting repeated values, deleting abnormal values and filling missing values.
5. The method of claim 3, wherein the training the coordinate conversion neural network model using the loss and back propagation algorithm comprises: When model parameters of the coordinate conversion neural network model are updated by using the loss and back propagation algorithm, a Adamw optimizer is used for optimizing the learning rate of parameter update.
6. The method of claim 3, wherein the coordinate conversion neural network model comprises a convolution module, a batch normalization layer, a long and short term memory network layer, a full connection layer, and a shape change layer; Inputting the first position coordinates of the defects into the coordinate conversion neural network model to obtain corresponding third position coordinates output by the coordinate conversion neural network model, wherein the first position coordinates comprise: inputting a plurality of first position coordinates of the plurality of defects to the convolution module to extract coordinate features of the plurality of first position coordinates; inputting the coordinate features of the plurality of first position coordinates to the batch normalization layer to normalize the coordinate features of the plurality of first position coordinates; inputting the normalized coordinate features of the plurality of first position coordinates into the long-short-period memory network to be used for ignoring first part features of the coordinate features of the plurality of first position coordinates and emphasizing second part features of the coordinate features of the plurality of first position coordinates so as to obtain output of the long-short-period memory network layer; Inputting the output of the long-short-term memory network layer to the full-connection layer for fitting the coordinate features of the plurality of first position coordinates; And inputting the coordinate features of the fitted first position coordinates to the shape changing layer to obtain a preset number of third position coordinates, wherein the preset number is equal to the number of the first position coordinates input to the convolution module.
7. A defect detection apparatus, comprising: An acquisition unit configured to acquire a plurality of first position coordinates of a plurality of defects in a detection object with respect to a first reference coordinate system, where the first reference coordinate system is set in the detection object; the input/output unit is used for inputting a plurality of first position coordinates of the plurality of defects into the coordinate conversion neural network model to obtain a plurality of second position coordinates of the plurality of defects output by the coordinate conversion neural network model relative to a second reference coordinate system, wherein the second reference coordinate system is arranged in the bearing device of the detection object; The coordinate conversion neural network model comprises a convolution module, a batch normalization layer, a long-period memory network layer, a full-connection layer and a shape changing layer; the input/output unit is specifically configured to: inputting a plurality of first position coordinates of the plurality of defects to the convolution module to extract coordinate features of the plurality of first position coordinates; inputting the coordinate features of the plurality of first position coordinates to the batch normalization layer to normalize the coordinate features of the plurality of first position coordinates; inputting the normalized coordinate features of the plurality of first position coordinates into the long-short-period memory network to be used for ignoring first part features of the coordinate features of the plurality of first position coordinates and emphasizing second part features of the coordinate features of the plurality of first position coordinates so as to obtain output of the long-short-period memory network layer; Inputting the output of the long-short-term memory network layer to the full-connection layer for fitting the coordinate features of the plurality of first position coordinates; And inputting the coordinate features of the fitted first position coordinates to the shape changing layer to obtain a preset number of second position coordinates, wherein the preset number is equal to the number of the first position coordinates input to the convolution module.
8. A computer device comprising a processor for implementing the defect detection method according to any of claims 1 to 6 when executing a computer program stored on a memory.
9. A computer readable storage medium having stored thereon a computer program for implementing the defect detection method according to any of claims 1 to 6 when executed by a processor.

Description

Defect detection method and related device Technical Field The present invention relates to the field of data processing technologies, and in particular, to a defect detection method and a related device. Background As a defect detecting device for a wafer, a detecting device is used to detect the surface of the wafer, a computer is used to process the detected image and extract defects, and then the processed image data in the images is stored in the computer, which is already implemented in a wafer detecting instrument. That is, at present, a detection image of a wafer defect can be displayed quickly in a computer, and the distribution of the defect on the wafer can be intuitively known according to the detection image. However, since the coordinates of each location on the wafer generally consist of two coordinates, one is the coordinates of the target location on the wafer relative to the first reference frame on the wafer, and the other is the coordinates of the wafer relative to the second reference frame on the carrier device (the carrier device is generally used for carrying the wafer, such as a flat panel or a panel, etc.), in order to process the defect on the wafer, it is currently required to first obtain the position coordinates of the defect relative to the second reference frame, and process the defect on the wafer according to the position coordinates of the defect relative to the second reference frame. However, only the position coordinates of the defect relative to the first reference frame are known, and the position coordinates of the defect relative to the second reference frame cannot be obtained. Disclosure of Invention The embodiment of the invention provides a defect detection method and a related device, which are used for rapidly acquiring the position coordinates of a defect relative to a second reference coordinate system according to the first position coordinates of the defect relative to the first reference coordinate system. An embodiment of the present application provides a defect detection method, including: acquiring a first position coordinate of a defect in a detection object relative to a first reference coordinate system, wherein the first reference coordinate system is arranged in the detection object; inputting the first position coordinates of the defects into a coordinate conversion neural network model to obtain second position coordinates of the defects, which are output by the coordinate conversion neural network model, relative to a second reference coordinate system, wherein the second reference coordinate system is arranged in a bearing device of the detection object, and the coordinate conversion neural network model is used for realizing conversion from the first position coordinates to the second position coordinates; and detecting the defect according to the second position coordinates of the defect. Preferably, the acquiring the first position coordinates of the defect in the detected object relative to the first reference coordinate system includes: Acquiring a plurality of first position coordinates of a plurality of defects in the detection object relative to the first reference coordinate system respectively; inputting the first position coordinates of the defect into a coordinate conversion neural network model to obtain second position coordinates of the defect output by the coordinate conversion neural network model relative to a second reference coordinate system, wherein the first position coordinates comprise: and respectively inputting the first position coordinates of the defects into a coordinate conversion neural network model by utilizing multithreading to obtain a plurality of second position coordinates of the defects output by the coordinate conversion neural network model relative to the second reference coordinate system. Preferably, the coordinate conversion neural network model comprises a convolution module, a batch normalization layer, a long-term and short-term memory network layer, a full-connection layer and a shape changing layer; Inputting the first position coordinates of the defects into a coordinate conversion neural network model by using multithreading respectively to obtain second position coordinates of the defects output by the coordinate conversion neural network model relative to the second reference coordinate system respectively, wherein the method comprises the following steps: inputting a plurality of first position coordinates of the plurality of defects to the convolution module to extract coordinate features of the plurality of first position coordinates; inputting the coordinate features of the plurality of first position coordinates to the batch normalization layer to normalize the coordinate features of the plurality of first position coordinates; inputting the normalized coordinate features of the plurality of first position coordinates into the long-short-period memory network to be used for ignoring