Defect detection method and system for semiconductor sample

Abstract

The invention discloses a defect detection method and system for a semiconductor sample, relates to the technical field of semiconductors, and solves the technical problem of low recognition efficiency of partial defects on the surface of a semiconductor caused by an inherent detection scheme of existing semiconductor surface defect detection. Comprising the following steps: acquiring a first image group of a to-be-detected semiconductor sample, and sequentially inputting each detection image in the first image group into a defect identification model to obtain a corresponding identification type and confidence; the defect identification model is obtained through artificial intelligence model training; generating a secondary acquisition scheme based on the identification type and confidence of each detection image; the secondary acquisition scheme comprises a plurality of second camera coordinates and a second shooting direction; acquiring a second image group of the semiconductor sample to be detected; generating an identification result based on the detection images in the first image group and the second image group; and the accuracy of semiconductor surface defect detection is improved.

Description

Technical Field

[0001] The present application belongs to the field of semiconductor technology, and specifically relates to a method and system for defect detection of semiconductor samples. Background Art

[0002] Semiconductor defects are primarily categorized as surface defects and internal defects. Surface defects include scratches, foreign matter, cracks, bubbles, and stains, while internal defects include grain fracture, grain dislocation, metal wire fracture, and metal wire dislocation. Semiconductor defects are diverse, complex, random, and minute, posing challenges to detection. Semiconductor defect detection methods are primarily categorized as contact and non-contact. Contact detection utilizes a probe in contact with the semiconductor surface or interior, detecting the presence of defects through electrical or force signals.

[0003] Existing semiconductor surface defect detection methods often place the semiconductor in a specific position, use a camera to photograph its surface, and then use a trained model to identify and analyze the photographed images to determine defects on the semiconductor device; and image acquisition of the semiconductor is often only performed at a set position; because defects on semiconductors have different characteristics in different directions, image acquisition and recognition in different directions will result in different recognition confidence or results; these defects are not obvious in these set positions; this will result in low recognition accuracy for such defects; therefore, a defect detection method for semiconductor samples is needed. Summary of the Invention

[0004] The present application aims to solve at least one of the technical problems existing in the prior art; to this end, the present application proposes a defect detection method and system for semiconductor samples, which is used to solve the technical problem of low efficiency in identifying partial defects on the semiconductor surface caused by the inherent detection scheme of the existing semiconductor surface defect detection.

[0005] To achieve the above objectives, the first aspect of the present application provides a method for detecting defects in a semiconductor sample, comprising: Acquire a first image group of the semiconductor sample to be inspected, wherein the first image group includes inspection images of the semiconductor sample to be inspected captured by a camera at a plurality of set positions; Inputting each detection image in the first image group into a defect recognition model in sequence to obtain its corresponding recognition type and confidence level; the defect recognition model is obtained by training an artificial intelligence model; generating a secondary acquisition plan based on the recognition type and confidence of each detection image; the secondary acquisition plan includes a plurality of second camera coordinates and a second shooting direction; Acquire a second image group of the semiconductor sample to be inspected, wherein the second image group includes inspection images of the semiconductor to be inspected in a plurality of set directions in the secondary acquisition scheme; A recognition result is generated based on the detection images in the first image group and the second image group.

[0006] The present application obtains a first image group of the semiconductor sample to be inspected, inputs each inspection image in the first image group into the defect recognition model in sequence to obtain its corresponding recognition type and confidence; generates a secondary acquisition plan based on the recognition type and confidence of each inspection image; obtains a second image group of the semiconductor sample to be inspected, and generates a recognition result based on the inspection images in the first image group and the second image group; performs a secondary inspection by analyzing the first recognition result, so that some defects with low defect reliability detected at a set position will have a targeted secondary inspection, thereby improving the accuracy of defect recognition; and further increasing the accuracy of semiconductor surface defect detection.

[0007] In combination with the first aspect above, in a possible implementation, acquiring a first image group of the semiconductor sample to be inspected includes: Constructing a three-dimensional coordinate system centered on the center of the semiconductor sample to be inspected; since the robotic arm that places the semiconductor sample to be inspected is in a fixed position during inspection, that is, the center of the semiconductor sample to be inspected is placed in a fixed position, the three-dimensional coordinate system is generally fixed; obtaining a number of set first camera coordinates and a first shooting direction; capturing an image of the semiconductor sample to be inspected in the camera coordinates and in the capture direction, and obtaining a detection image corresponding to the first camera coordinates and the first shooting direction; The detection images at the set first camera coordinates and the first shooting direction are integrated into a first image group.

[0008] In conjunction with the first aspect above, in one possible implementation, the defect recognition model is obtained through artificial intelligence model training, including: Acquire a plurality of semiconductor samples corresponding to various defect types, perform image acquisition on each of the semiconductor samples to obtain a plurality of inspection images, and generate a plurality of training data and inspection data based on the inspection images; The artificial intelligence model is trained using training images; the trained artificial intelligence model is tested using test images to obtain an artificial intelligence model whose input is the test image and whose output is the corresponding defect type; the defect type is output as the recognition type, and the confidence level corresponding to the defect type is output together, and finally a defect recognition model is obtained whose input is the test image and whose output is the recognition type and its corresponding confidence level.

[0009] In conjunction with the first aspect above, in one possible implementation, generating a plurality of training data and test data based on the detection image includes: The corresponding recognition types are calibrated for several defective parts in the detection image, and each detection image and its corresponding recognition types are integrated into training data and test data.

[0010] In combination with the first aspect above, in one possible implementation, generating a secondary acquisition plan based on the recognition type and confidence level of each detection image includes: Obtaining the locations of defects corresponding to each identification type on the semiconductor to be inspected; integrating identification types and confidence levels with the same location into a defect identification group; Determine whether the maximum value of the confidence in the defect identification group is greater than a set confidence threshold. If yes, mark the identification type as a defect type; if not, generate a secondary acquisition plan based on the confidence in the defect identification group. Obtain the secondary acquisition plan corresponding to each defect identification group in turn.

[0011] In combination with the first aspect above, in a possible implementation, generating the secondary acquisition plan according to the confidence level in the defect identification group includes: When there is only one confidence level in the defect recognition group, the first camera coordinates corresponding to the confidence level are obtained, and the first camera coordinates are used as the detection coordinates. The detection coordinates are the center coordinates of the camera when the second image acquisition is performed; based on the detection coordinates, a number of second camera coordinates and corresponding second shooting directions are generated. When there are multiple confidence levels in the defect recognition group, the first camera coordinates corresponding to each confidence level are obtained, and the detection coordinates are calculated based on each first camera coordinate and the confidence level. Based on the detection coordinates, a number of second camera coordinates and corresponding second shooting directions are generated; The second camera coordinates and their corresponding second shooting directions are integrated into a secondary acquisition solution.

[0012] In combination with the first aspect above, in one possible implementation, calculating the detection coordinates based on the first camera coordinates and the confidence level includes: Get each first camera coordinate and its corresponding confidence, convert each first camera coordinate into spherical coordinate, marked as , and its corresponding confidence is marked as ZDi; where i is the number of the first camera coordinate; through the formula:

[0013] Calculate the detection coordinates ; Wherein, i=1, 2,…, I; I is the total number of the first camera coordinates.

[0014] In combination with the first aspect above, in one possible implementation, generating a plurality of second camera coordinates and corresponding second shooting directions based on the detected coordinates includes: The detected coordinates are used as the second camera coordinates of the target, and the direction of the detected coordinates pointing to the origin is used as the second shooting direction of the target; Obtain a set reference offset angle; obtain a number of directions whose included angle with the target second shooting direction is the reference offset angle as a reference direction set, and randomly select a set number of directions from the reference direction set as reference second shooting directions; Exemplarily, the detection coordinates are converted into a three-dimensional Cartesian coordinate system to obtain the detection coordinates, that is, the target second camera coordinates:

[0015] The direction of the detection coordinate pointing to the origin is taken as the second shooting direction of the target, that is, the direction of the target second camera coordinate pointing to the origin is the second shooting direction of the target; that is, the vector direction;

[0016] Reference the second shooting direction as vector direction; ;vector The following conditions are met:

[0017] Where, ω is the reference offset angle; Calculating the distance between the target second camera coordinates and the origin; selecting the coordinates of a point in the reference second shooting direction that is equal to the distance from the origin as the reference second camera coordinates; Sequentially obtain the reference second camera coordinates corresponding to each reference second shooting direction; The second camera coordinates include target second camera coordinates and reference second camera coordinates; the second shooting direction includes target second shooting direction and reference second shooting direction.

[0018] In combination with the first aspect above, in a possible implementation, generating the recognition result based on the detection images in the first image group and the second image group includes: Acquire a defect type identified based on the first image group, and a detection image corresponding to the defect type; Extracting a plurality of detection images from the second image group; inputting the detection images into a defect recognition model to obtain corresponding recognition types and confidence levels; Determine whether the maximum confidence level in the second recognition group is greater than a set confidence threshold; if so, use the recognition type corresponding to the maximum confidence level as the defect type corresponding to the second image group; if not, mark the defect type as unrecognized; Sequentially acquiring the defect type corresponding to each second image group and the corresponding plurality of inspection images; Each defect type and the corresponding inspection image are integrated into the recognition result.

[0019] Another aspect of the present application provides a semiconductor sample defect detection system, comprising: an image acquisition module, a data processing module, and a data storage module; The image acquisition module is configured to acquire a first image group and a second image group of the semiconductor sample to be inspected; the first image group and the second image group both include inspection images of the semiconductor sample to be inspected captured by a plurality of cameras at a plurality of set positions; The data processing module sequentially inputs each inspection image in the first image group into a defect recognition model to obtain its corresponding recognition type and confidence level; the defect recognition model is trained by an artificial intelligence model; and generates a secondary acquisition plan based on the recognition type and confidence level of each inspection image; the secondary acquisition plan includes a plurality of second camera coordinates and a second shooting direction. Acquire a second image group of the semiconductor sample to be inspected, the second image group including inspection images of the semiconductor to be inspected in a plurality of set directions in the secondary acquisition scheme; generate a recognition result based on the inspection images in the first image group and the second image group; The data storage module is used to store data such as the first image group, the second image group and the recognition result.

[0020] Compared with the prior art, the present invention has the following advantages: The present application obtains a first image group of the semiconductor sample to be inspected, inputs each inspection image in the first image group into the defect recognition model in sequence to obtain its corresponding recognition type and confidence; generates a secondary acquisition plan based on the recognition type and confidence of each inspection image; obtains a second image group of the semiconductor sample to be inspected, and generates a recognition result based on the inspection images in the first image group and the second image group; performs a secondary inspection by analyzing the first recognition result, so that some defects with low defect reliability detected at a set position will have a targeted secondary inspection, thereby improving the accuracy of defect recognition; and further increasing the accuracy of semiconductor surface defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0022] Figure 1 A schematic diagram of the method steps for defect detection of semiconductor samples in this application; Figure 2This is a schematic diagram of the module connection of the semiconductor sample defect detection system in this application. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions of this application in conjunction with the embodiments. Obviously, the embodiments described are only a part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0024] See also Figure 1 The first aspect of the present application provides a method for detecting defects in a semiconductor sample, comprising: Acquire a first image group of the semiconductor sample to be inspected, the first image group including inspection images of the semiconductor sample to be inspected captured by the camera at a plurality of set positions; the set positions are manually set fixed positions, such as the front, back, left, right, top, and bottom positions of the semiconductor to be inspected; Inputting each inspection image in the first image group into a defect recognition model in sequence to obtain its corresponding recognition type and confidence level; the defect recognition model is obtained by training an artificial intelligence model; the recognition type is the possible defect type corresponding to the defective portion on the inspection image identified by the trained artificial intelligence model, and the confidence level can be understood as the accuracy of the recognition of the corresponding defect type; A secondary acquisition scheme is generated based on the recognition type and confidence level of each inspection image; the secondary acquisition scheme includes a plurality of second camera coordinates and a second shooting direction; when the accuracy of defect recognition in the inspection image captured at the set acquisition position is low, the defect at the corresponding position is further confirmed by setting a secondary acquisition to increase the accuracy of defect recognition; Acquire a second image group of the semiconductor sample to be inspected, the second image group including inspection images of the semiconductor to be inspected in a plurality of set directions in the secondary acquisition scheme; A recognition result is generated based on the detection images in the first image group and the second image group; the recognition result is the final recognition result of the semiconductor to be detected, including normal, various defects and other results.

[0025] Since defects on semiconductors exhibit different characteristics in different directions, different recognition confidences or results will appear when performing image acquisition and recognition in different directions. It can be understood that there will be an angle where the image of the defective part acquired has the best recognition effect. In this embodiment, a first image group of the semiconductor sample to be inspected is acquired, and each detection image in the first image group is input into the defect recognition model in turn to obtain its corresponding recognition type and confidence. A secondary acquisition scheme is generated based on the recognition type and confidence of each detection image. A second image group of the semiconductor sample to be inspected is acquired, and a recognition result is generated based on the detection images in the first image group and the second image group. A secondary detection is performed by analyzing the first recognition result, so that some defects with low defect reliability detected at the set position will be targeted for secondary detection, thereby improving the accuracy of defect recognition and further increasing the accuracy of semiconductor surface defect detection.

[0026] In a possible implementation, acquiring a first image group of the semiconductor sample to be inspected includes: Constructing a three-dimensional coordinate system centered on the center of the semiconductor sample to be inspected; since the robotic arm that places the semiconductor sample to be inspected is in a fixed position during inspection, that is, the center of the semiconductor sample to be inspected is placed in a fixed position, the three-dimensional coordinate system is generally fixed; obtaining a number of set first camera coordinates and a first shooting direction; capturing an image of the semiconductor sample to be inspected in the camera coordinates and in the capture direction, and obtaining a detection image corresponding to the first camera coordinates and the first shooting direction; The detection images at the set first camera coordinates and the first shooting direction are integrated into a first image group.

[0027] Exemplarily, there are six groups of first camera coordinates and first shooting directions set in this embodiment; the first camera coordinates are (0, 0, z), and the first shooting direction is the negative direction of the Z axis; the first camera coordinates are (0, 0, -z), and the first shooting direction is the positive direction of the Z axis; the first camera coordinates are (x, 0, 0), and the first shooting direction is the negative direction of the X axis; the first camera coordinates are (-x, 0, 0), and the first shooting direction is the positive direction of the X axis; the first camera coordinates are (0, y, 0), and the first shooting direction is the negative direction of the Y axis; the first camera coordinates are (0, y, 0), and the first shooting direction is the positive direction of the Y axis; x, y and z are set positive numbers; it can be understood that the three-dimensional coordinate system that can be constructed here can be any one of the three-dimensional Cartesian coordinate system and the spherical coordinate system; this embodiment takes the three-dimensional Cartesian coordinate system as an example. If other coordinate systems are used, the corresponding coordinates only need to be converted to the corresponding coordinate system through the existing coordinate conversion method.

[0028] It can be understood that the first camera coordinates correspond to the first shooting direction in a one-to-one manner, and the first shooting direction is a direction pointing to the origin of the three-dimensional coordinate system with the corresponding first camera coordinates as the origin.

[0029] In one possible implementation, the defect recognition model is obtained by training an artificial intelligence model, including: Acquire a plurality of semiconductor samples corresponding to various defect types, perform image acquisition on each of the semiconductor samples to obtain a plurality of inspection images, and generate a plurality of training data and inspection data based on the inspection images; The artificial intelligence model is trained using training images; the trained artificial intelligence model is tested using test images to obtain an artificial intelligence model whose input is the test image and whose output is the corresponding defect type; the defect type is output as the recognition type, and the confidence level corresponding to the defect type is output together, and finally a defect recognition model whose input is the test image and whose output is the recognition type and its corresponding confidence level is obtained; wherein the artificial intelligence model includes a recurrent neural network model, etc.; it can be understood that when the defect type is a specific defect, its corresponding recognition type is the corresponding defect type, such as scratches, cracks, and chipped edges; when the defect type is none, its corresponding recognition type is normal; the training data includes images with defect type none.

[0030] In a possible implementation, generating a plurality of training data and test data based on the detection image includes: The corresponding recognition types are calibrated for several defective parts in the detection image, and each detection image and its corresponding recognition types are integrated into training data and test data.

[0031] In a possible implementation, generating a secondary acquisition plan based on the recognition type and confidence level of each detection image includes: Obtaining the locations of defects corresponding to each identification type on the semiconductor to be inspected; integrating identification types and confidence levels with the same location into a defect identification group; Determine whether the maximum value of the confidence in the defect identification group is greater than a set confidence threshold. If yes, mark the identification type as a defect type; if not, generate a secondary acquisition plan based on the confidence in the defect identification group; Obtain the secondary acquisition plan corresponding to each defect identification group in turn.

[0032] In a possible implementation, generating the secondary acquisition plan according to the confidence level in the defect identification group includes: When there is only one confidence level in the defect recognition group, the first camera coordinates corresponding to the confidence level are obtained and used as the detection coordinates. The detection coordinates are the center coordinates of the camera during the second image acquisition. Based on the detection coordinates, several second camera coordinates and corresponding second shooting directions are generated. When there are multiple confidence levels in the defect recognition group, the first camera coordinates corresponding to each confidence level are obtained, the detection coordinates are calculated based on each first camera coordinate and the confidence level, and a number of second camera coordinates and corresponding second shooting directions are generated based on the detection coordinates; The second camera coordinates and their corresponding second shooting directions are integrated into a secondary acquisition solution.

[0033] In one possible implementation, the detection coordinates are calculated based on the first camera coordinates and the confidence level, including: Get each first camera coordinate and its corresponding confidence, convert each first camera coordinate into spherical coordinate, marked as , and its corresponding confidence is marked as ZDi; where i is the number of the first camera coordinate; through the formula:

[0034] Calculate the detection coordinates ; Wherein, i=1, 2,…, I; I is the total number of the first camera coordinates.

[0035] In this embodiment, the confidence level and the corresponding first camera coordinate are calculated using the above formula to obtain the detection coordinates. When the confidence level in a certain camera coordinate direction is much greater than the confidence levels in other directions, it indicates that the defects in the shooting direction corresponding to the camera coordinate are more obvious. At this time, the closer the final detection coordinates are to the camera coordinates, that is, the images are recaptured around the camera coordinates, and the greater the probability of obtaining an image with obvious defects, which is convenient for improving the accuracy of subsequent recognition. When the confidence levels in the directions of various camera coordinates are similar, it indicates that a direction in which defects are more obvious will appear in the area surrounded by the various camera coordinates. At this time, the detection coordinates within the area are generated according to the confidence level, and the images are recaptured, which is convenient for improving the accuracy of subsequent recognition.

[0036] In one possible implementation, generating a plurality of second camera coordinates and corresponding second shooting directions based on the detected coordinates includes: The detected coordinates are used as the second camera coordinates of the target, and the direction of the detected coordinates pointing to the origin is used as the second shooting direction of the target; Get the set reference offset angle; the reference offset angle is artificially set, and the reference offset angle set in this embodiment is 10°; get a number of directions whose included angle with the target second shooting direction is the reference offset angle as a reference direction set, and randomly select a set number of directions from the reference direction set as reference second shooting directions; in this embodiment, by setting an angle limit between the reference direction and the reference direction, randomly select a set number of directions from the reference direction set as reference second shooting directions, that is, arbitrarily select one reference direction as the reference second shooting direction, and the reference directions in the reference direction set whose included angle with the reference second shooting direction is the set angle limit are also used as reference second shooting directions, and finally the set number of reference second shooting directions are obtained by iteration; the specific number is artificially set. For setting, the set number in this embodiment is 10; in another embodiment, the distance between the corresponding starting points of the unit vectors in the reference direction is set as a restriction condition; a set number of directions are randomly selected in the reference direction set as reference second shooting directions; that is, any reference direction is selected as the reference second shooting direction, and the starting point coordinates of the unit vector of the reference second shooting direction are obtained. It can be understood that the end point coordinates of the unit vector are the origin; the starting point coordinates of the unit vectors corresponding to each reference direction in the reference direction set are obtained; the distance between the starting point coordinates and the starting point coordinates of the unit vector corresponding to the reference second shooting direction and the reference direction with the set distance limit are also used as the reference second shooting direction, and the set number of reference second shooting directions are finally obtained by iteration; the specific number is set artificially.

[0037] Exemplarily, the detection coordinates are converted into a three-dimensional Cartesian coordinate system to obtain the detection coordinates, that is, the target second camera coordinates:

[0038] The direction of the detection coordinate pointing to the origin is taken as the second shooting direction of the target, that is, the direction of the target second camera coordinate pointing to the origin is the second shooting direction of the target; that is, the vector direction;

[0039] Reference the second shooting direction as vector direction; ;vector The following conditions are met:

[0040] Where ω is the reference offset angle; by substituting the reference offset angle, we can obtain several vectors ; Each vector The direction is taken as the second shooting direction; The following conditions are met:

[0041] Calculate the distance between the target second camera coordinates and the origin; select the coordinates of the point in the reference second shooting direction that is equal to the distance from the origin as the reference second camera coordinates; that is, there is a vector The reference second camera coordinates corresponding to the determined second shooting direction; Sequentially obtain the reference second camera coordinates corresponding to each reference second shooting direction; The second camera coordinates include the target second camera coordinates and the reference second camera coordinates; the second shooting direction includes the target second shooting direction and the reference second shooting direction. It is understood that the distances between each reference second camera coordinate and the target second camera coordinate are equal and distributed around the target second camera coordinate.

[0042] In a possible implementation, generating the recognition result based on the detection images in the first image group and the second image group includes: Acquire a defect type identified based on the first image group, and a detection image corresponding to the defect type; Extracting a plurality of detection images from the second image group; inputting the detection images into a defect recognition model to obtain corresponding recognition types and confidence levels; Determine whether the maximum confidence level in the second recognition group is greater than a set confidence threshold; if so, use the recognition type corresponding to the maximum confidence level as the defect type corresponding to the second image group; if not, mark the defect type as unrecognized; Sequentially acquiring the defect type corresponding to each second image group and the corresponding plurality of inspection images; Each defect type and the corresponding inspection image are integrated into the recognition result.

[0043] See also Figure 2 , another aspect of the present application provides a semiconductor sample defect detection system, comprising: an image acquisition module, a data processing module and a data storage module; The image acquisition module is configured to acquire a first image group and a second image group of the semiconductor sample to be inspected; the first image group and the second image group both include inspection images of the semiconductor sample to be inspected captured by a plurality of cameras at a plurality of set positions; The data processing module sequentially inputs each inspection image in the first image group into a defect recognition model to obtain its corresponding recognition type and confidence level; the defect recognition model is trained by an artificial intelligence model; and generates a secondary acquisition plan based on the recognition type and confidence level of each inspection image; the secondary acquisition plan includes a plurality of second camera coordinates and a second shooting direction. Acquire a second image group of the semiconductor sample to be inspected, the second image group including inspection images of the semiconductor to be inspected in a plurality of set directions in the secondary acquisition scheme; generate a recognition result based on the inspection images in the first image group and the second image group; The data storage module is used to store data such as the first image group, the second image group and the recognition result.

[0044] Some of the data in the above formula are calculated by removing the dimensions and taking their numerical values. The formula is a formula that is closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0045] How this application works: The present application obtains a first image group of the semiconductor sample to be inspected, inputs each inspection image in the first image group into the defect recognition model in sequence to obtain its corresponding recognition type and confidence; generates a secondary acquisition plan based on the recognition type and confidence of each inspection image; obtains a second image group of the semiconductor sample to be inspected, and generates a recognition result based on the inspection images in the first image group and the second image group; performs a secondary inspection by analyzing the first recognition result, so that some defects with low defect reliability detected at a set position will have a targeted secondary inspection, thereby improving the accuracy of defect recognition; and further increasing the accuracy of semiconductor surface defect detection.

[0046] The above embodiments are only used to illustrate the technical method of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present application.

Claims

1. A method for detecting defects in semiconductor samples, characterized in that: include: Acquire a first image group of the semiconductor sample to be inspected, wherein the first image group includes inspection images of the semiconductor sample to be inspected captured by a camera at a plurality of set positions; Inputting each detection image in the first image group into the defect recognition model in sequence to obtain its corresponding recognition type and confidence level; The defect recognition model is obtained through artificial intelligence model training; Generate a secondary acquisition plan based on the recognition type and confidence of each detection image; The secondary acquisition scheme includes a number of second camera coordinates and a second shooting direction; Acquire a second image group of the semiconductor sample to be inspected, wherein the second image group includes inspection images of the semiconductor to be inspected in a plurality of set directions in the secondary acquisition scheme; A recognition result is generated based on the detection images in the first image group and the second image group.

2. The method for defect detection of a semiconductor sample according to claim 1, wherein: The step of acquiring a first image group of the semiconductor sample to be inspected comprises: Constructing a three-dimensional coordinate system centered on the center of the semiconductor sample to be inspected; obtaining a number of set first camera coordinates and a first shooting direction; capturing an image of the semiconductor sample to be inspected in the camera coordinates and in the acquisition direction, and obtaining a detection image corresponding to the first camera coordinates and the first shooting direction; The detection images at the set first camera coordinates and the first shooting direction are integrated into a first image group.

3. The method for defect detection of a semiconductor sample according to claim 1, wherein: The defect recognition model is obtained through artificial intelligence model training, including: Acquire a plurality of semiconductor samples corresponding to various defect types, perform image acquisition on each of the semiconductor samples to obtain a plurality of inspection images, and generate a plurality of training data and inspection data based on the inspection images; The artificial intelligence model is trained using training images; the trained artificial intelligence model is tested using test images to obtain an artificial intelligence model whose input is the test image and whose output is the corresponding defect type; the defect type is output as the recognition type, and the confidence level corresponding to the defect type is output together, and finally a defect recognition model is obtained whose input is the test image and whose output is the recognition type and its corresponding confidence level.

4. The method for defect detection of a semiconductor sample according to claim 3, wherein: The generating of a plurality of training data and test data based on the detection image includes: The corresponding recognition types are calibrated for several defective parts in the detection image, and each detection image and its corresponding recognition types are integrated into training data and test data.

5. The method for defect detection of a semiconductor sample according to claim 1, wherein: The generating of a secondary acquisition scheme based on the recognition type and confidence of each detection image includes: Obtaining the locations of defects corresponding to each identification type on the semiconductor to be inspected; integrating identification types and confidence levels with the same location into a defect identification group; Determine whether the maximum value of the confidence in the defect identification group is greater than a set confidence threshold. If yes, mark the identification type as a defect type; if not, generate a secondary acquisition plan based on the confidence in the defect identification group. Obtain the secondary acquisition plan corresponding to each defect identification group in turn.

6. The method for detecting defects in a semiconductor sample according to claim 5, wherein: Generating the secondary acquisition plan according to the confidence level in the defect identification group includes: When there is only one confidence level in the defect recognition group, the first camera coordinate corresponding to the confidence level is obtained and the first camera coordinate is used as the detection coordinate; a plurality of second camera coordinates and corresponding second shooting directions are generated based on the detection coordinate; When there are multiple confidence levels in the defect recognition group, the first camera coordinates corresponding to each confidence level are obtained, the detection coordinates are calculated based on each first camera coordinate and the confidence level, and a number of second camera coordinates and corresponding second shooting directions are generated based on the detection coordinates; The second camera coordinates and their corresponding second shooting directions are integrated into a secondary acquisition solution.

7. The method for detecting defects in semiconductor samples according to claim 6, wherein: The step of calculating the detection coordinates based on the first camera coordinates and the confidence level includes: Get each first camera coordinate and its corresponding confidence, convert each first camera coordinate into spherical coordinate, marked as , and its corresponding confidence is marked as ZDi; where i is the number of the first camera coordinate; through the formula: Calculate the detection coordinates ; Wherein, i=1, 2,…, I; I is the total number of the first camera coordinates.

8. The method for detecting defects in semiconductor samples according to claim 6, wherein: The generating of a plurality of second camera coordinates and corresponding second shooting directions based on the detected coordinates includes: The detected coordinates are used as the second camera coordinates of the target, and the direction of the detected coordinates pointing to the origin is used as the second shooting direction of the target; Obtain a set reference offset angle; obtain a number of directions whose included angle with the target second shooting direction is the reference offset angle as a reference direction set, and randomly select a set number of directions from the reference direction set as reference second shooting directions; Calculating the distance between the target second camera coordinates and the origin; selecting the coordinates of a point in the reference second shooting direction that is equal to the distance from the origin as the reference second camera coordinates; Sequentially obtain the reference second camera coordinates corresponding to each reference second shooting direction; The second camera coordinates include target second camera coordinates and reference second camera coordinates; the second shooting direction includes target second shooting direction and reference second shooting direction.

9. The method for defect detection of a semiconductor sample according to claim 1, wherein: Generating the recognition result based on the detection images in the first image group and the second image group includes: Acquire a defect type identified based on the first image group, and a detection image corresponding to the defect type; Extracting a plurality of detection images from the second image group; inputting the detection images into a defect recognition model to obtain corresponding recognition types and confidence levels; Determine whether the maximum confidence level in the second recognition group is greater than a set confidence threshold; if so, use the recognition type corresponding to the maximum confidence level as the defect type corresponding to the second image group; if not, mark the defect type as unrecognized; Sequentially acquiring the defect type corresponding to each second image group and the corresponding plurality of inspection images; Each defect type and the corresponding inspection image are integrated into the recognition result.

10. A semiconductor sample defect detection system, based on a semiconductor sample defect detection method according to any one of claims 1 to 9, characterized in that: include: Image acquisition module, data processing module and data storage module; The image acquisition module is configured to acquire a first image group and a second image group of the semiconductor sample to be inspected; The first image group and the second image group each include detection images of the semiconductor sample to be detected captured by a plurality of cameras at a plurality of set positions; The data processing module sequentially inputs each detection image in the first image group into the defect recognition model to obtain its corresponding recognition type and confidence level; The defect recognition model is obtained through artificial intelligence model training; a secondary acquisition scheme is generated based on the recognition type and confidence of each detection image; the secondary acquisition scheme includes a plurality of second camera coordinates and a second shooting direction; Acquire a second image group of the semiconductor sample to be inspected, the second image group including inspection images of the semiconductor to be inspected in a plurality of set directions in the secondary acquisition scheme; generate a recognition result based on the inspection images in the first image group and the second image group; The data storage module is used to store the first image group, the second image group and the recognition result.

Images