Method for establishing hidden crack detection model, hidden crack detection method and defect detection method

Abstract

The invention discloses an establishment method of a hidden crack detection model, a hidden crack detection method and a defect detection method, and relates to the technical field of optical detection.The establishment method of the hidden crack detection model comprises the steps that an initial hidden crack detection database is established; a training sample set is acquired for multiple times, and after the training sample set is acquired each time, the subfissure chip image in the initial subfissure detection database is supplemented according to the currently acquired training sample set, so that the initial subfissure detection database is optimized step by step, and a target subfissure detection database is acquired; and establishing a chip hidden crack detection model according to the target hidden crack detection database. According to the technical scheme provided by the invention, the detection accuracy of the chip subfissure detection model is ensured, and meanwhile, the efficiency of establishing the subfissure detection model is improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of optical detection technology, and in particular to a method for establishing a hidden crack detection model, a hidden crack detection method, and a defect detection method. Background Art

[0002] As electronic products become more and more functional, customers have higher and higher requirements for chip quality, so there is a need to detect defects on chips, especially hidden cracks (tiny cracks on the chip body).

[0003] The traditional chip crack detection method usually relies on manual sampling and manual operation of infrared electron microscope for manual identification; however, manual detection is inevitably affected by subjective factors and experience, resulting in poor detection accuracy and efficiency of chip crack defects, which can no longer meet the requirements for accuracy and speed of detection; in related technologies, intelligent detection models can provide a more accurate and efficient defect detection method, but a large number of samples need to be obtained during the current detection model establishment process, resulting in a more complicated model establishment process. Therefore, how to improve the efficiency of establishing the detection model while ensuring the detection accuracy of the detection model is a technical problem that people in this field need to solve urgently. Summary of the invention

[0004] The embodiments of the present invention provide a method for establishing a hidden crack detection model, a hidden crack detection method and a defect detection method, so as to ensure the detection accuracy of the chip hidden crack detection model and improve the efficiency of establishing the hidden crack detection model.

[0005] According to one aspect of the present invention, a method for establishing a hidden crack detection model is provided, comprising:

[0006] Establish an initial hidden crack detection database;

[0007] Acquire training sample sets multiple times, and after each acquisition of the training sample set, supplement the hidden crack chip images in the initial hidden crack detection database according to the currently acquired training sample set, so as to gradually optimize the initial hidden crack detection database and obtain a target hidden crack detection database;

[0008] A chip hidden crack detection model is established according to the target hidden crack detection database.

[0009] Optionally, the establishing of an initial hidden crack detection database includes:

[0010] Acquire a first preset number of hidden crack chip images and a second preset number of standard chip images;

[0011] The initial hidden crack detection database is established according to a first preset number of hidden crack chip images and a second preset number of standard chip images.

[0012] Optionally, the training sample set is obtained multiple times, and after each training sample set is obtained, the hidden crack chip images in the initial hidden crack detection database are supplemented according to the currently obtained training sample set, so as to gradually optimize the initial hidden crack detection database and obtain a target hidden crack detection database, including:

[0013] Capturing images of the first batch of hidden crack chips to obtain a first training sample set;

[0014] Compare each first training sample in the first training sample set with the images stored in the initial hidden crack detection database, and determine whether there is an unrecognized first training sample; if so, add the unrecognized first training sample as a newly added hidden crack chip image to the initial hidden crack detection database to obtain a first optimized hidden crack detection database;

[0015] Capturing images of the second batch of hidden crack chips to obtain a second training sample set;

[0016] Compare and detect each second training sample in the second training sample set with the images stored in the first optimized hidden crack detection database, and determine whether there is an unidentified second training sample; if so, add the unidentified second training sample as a newly added hidden crack chip image to the first optimized hidden crack detection database to obtain a second optimized hidden crack detection database;

[0017] Capturing images of the third batch of hidden crack chips to obtain a third training sample set;

[0018] Compare each third training sample in the third training sample set with the images stored in the second optimized hidden crack detection database, and determine whether there is an unidentified third training sample; if so, add the unidentified third training sample as a newly added hidden crack chip image to the second optimized hidden crack detection database to obtain a third optimized hidden crack detection database;

[0019] And so on, until it is determined that the target hidden crack detection database is obtained;

[0020] Preferably, the image acquisition is performed by an infrared image acquisition device.

[0021] Optionally, determining to obtain a target hidden crack detection database includes:

[0022] Determine whether the number of times the training sample set is obtained meets the target number, and / or determine whether the hidden crack chip images stored in the currently optimized initial hidden crack detection database meet the target number, and / or determine whether the recognition success rate of the currently optimized initial hidden crack detection database is greater than or equal to the preset recognition success rate; if so, determine that the target hidden crack detection database is obtained.

[0023] Optionally, the unidentified training samples are added to the initial hidden crack detection database after the last optimization, and further include:

[0024] Determine whether the unrecognized training samples have clear images;

[0025] If it is clear, the unidentified training samples are added to the initial hidden crack detection database after the last optimization.

[0026] Optionally, if determining whether the unrecognized training sample is an unclear image, the method further includes:

[0027] Re-acquire images of the hidden crack chips corresponding to the unidentified and unclear training samples, and compare the re-acquired training samples with the images in the last optimized initial hidden crack detection database;

[0028] If there are training samples that are not recognized again, the training samples that are not recognized again are added to the initial hidden crack detection database after the last optimization.

[0029] Optionally, after determining the chip hidden crack detection numerical model according to the target hidden crack detection database, the method further includes:

[0030] Performing hidden crack detection on a preset number of chips to be tested using the chip hidden crack detection model;

[0031] The detection results of the hidden crack detection numerical model are manually sampled and re-inspected to conduct a qualified inspection of the chip hidden crack detection numerical model.

[0032] According to another aspect of the present invention, a method for detecting hidden cracks in a chip is provided, comprising:

[0033] Acquire an image of the chip to be tested;

[0034] The image of the chip to be tested is input into a chip hidden crack detection model, and the image of the chip to be tested is analyzed by the chip hidden crack detection model to obtain a detection result; wherein the chip hidden crack detection model is constructed by the chip hidden crack detection model establishment method described in any embodiment of the present invention.

[0035] Optionally, acquiring the image of the chip to be tested includes:

[0036] The image of the chip to be tested is detected by an infrared image acquisition device.

[0037] According to another aspect of the present invention, a chip defect detection method is provided, including the chip hidden crack detection method according to any embodiment of the present invention.

[0038] Beneficial effects: The embodiments of the present invention provide a method for establishing a hidden crack detection model, a hidden crack detection method and a defect detection method, wherein the method for establishing a hidden crack detection model includes: establishing an initial hidden crack detection database; acquiring training sample sets multiple times, and after each acquisition of the training sample set, supplementing the hidden crack chip images in the initial hidden crack detection database according to the currently acquired training sample set, so as to gradually optimize the initial hidden crack detection database and obtain a target hidden crack detection database; and establishing a chip hidden crack detection model according to the target hidden crack detection database. The technical solution provided by the embodiments of the present invention can gradually optimize the initial hidden crack detection database until the target hidden crack detection database is obtained by acquiring training sample sets multiple times, and after each acquisition of the training sample set, supplementing the hidden crack chip images in the initial hidden crack detection database according to the currently acquired training sample set, without having to acquire a large number of samples at one time to establish the hidden crack detection database, so that the repetition rate of the acquired samples can be reduced, and the total amount of acquired samples can be reduced, so that the efficiency of establishing the hidden crack detection model can be improved while ensuring the detection accuracy of the chip hidden crack detection model.

[0039] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0041] Figure 1 is a flow chart of a method for establishing a hidden crack detection model provided by an embodiment of the present invention;

[0042] Figure 2 is a flow chart of another method for establishing a hidden crack detection model provided by an embodiment of the present invention;

[0043] Figure 3 is a flow chart of another method for establishing a hidden crack detection model provided by an embodiment of the present invention;

[0044] Figure 4is a flow chart of another method for establishing a hidden crack detection model provided by an embodiment of the present invention;

[0045] Figure 5 is a flow chart of another method for establishing a hidden crack detection model provided by an embodiment of the present invention;

[0046] Figure 6 It is a structural schematic diagram of a defect detection production line provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0048] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0049] The embodiment of the present invention provides a method for establishing a hidden crack detection model. Figure 1 is a flow chart of a method for establishing a hidden crack detection model provided by an embodiment of the present invention, with reference to Figure 1 , the method for establishing the hidden crack detection model includes:

[0050] S110: Establish an initial hidden crack detection database.

[0051] Specifically, a first preset number of hidden crack chip images are stored in the initial hidden crack detection database, and the hidden crack chip images stored in the initial hidden crack detection database can be obtained by collecting a corresponding number of sample chips with hidden cracks through an infrared image acquisition device. For example, if 50 hidden crack chip images are stored in the initial hidden crack detection database, the infrared image acquisition device is used to collect images of 50 sample chips with hidden cracks.

[0052] The chips corresponding to the hidden crack chip images stored in the initial hidden crack detection database are chips of the same model, and the chips corresponding to different hidden crack chip images have at least one different hidden crack, and different hidden cracks can be understood as different at least one of the positions, types and sizes of the hidden cracks. If there is at least one hidden crack in the chip corresponding to each hidden crack chip image, then each hidden crack chip image carries feature information of at least one hidden crack.

[0053] S120, acquiring training sample sets multiple times, and after each acquisition of the training sample set, supplementing the hidden crack chip images in the initial hidden crack detection database according to the currently acquired training sample set, so as to gradually optimize the initial hidden crack detection database and obtain a target hidden crack detection database.

[0054] Specifically, each training sample set includes multiple training samples, and each training sample is a hidden crack chip image. Different training sample sets have different numbers of hidden crack chip images, or different training sample sets have the same number of hidden crack chip images, or at least some training sample sets have different numbers of hidden crack chip images. In the same training sample set, different hidden crack chip images correspond to different hidden crack chips, and the chip models corresponding to the hidden crack chip images included in each training sample set are the same as the models of hidden crack chips corresponding to the hidden crack chip images stored in the initial hidden crack detection database. In other words, the hidden crack chip images stored in the target hidden crack detection database obtained through multiple optimizations are all images of the same model of hidden crack chips.

[0055] After each training sample set is obtained, each training sample in the currently obtained training sample set is compared with the hidden crack chip image in the current hidden crack detection database. If there is a training sample in the currently obtained training sample set that is different from the hidden crack chip image stored in the current hidden crack detection database, the training sample is stored in the current hidden crack detection database, so that the current hidden crack detection database is optimized. The training sample set is obtained multiple times, and after each training sample set is obtained, the hidden crack chip image in the initial hidden crack detection database is supplemented according to the currently obtained training sample set. The initial hidden crack detection database can be optimized multiple times, and the hidden crack information stored in the initial hidden crack detection database is gradually improved. There is no need to obtain a large number of samples (hidden crack chip images) at one time to establish a target hidden crack detection database, so the repetition rate of the obtained samples can be reduced, and the total amount of samples obtained can be reduced, so that the detection accuracy of the chip hidden crack detection model can be guaranteed while improving the efficiency of establishing the chip hidden crack detection model.

[0056] If the hidden crack detection database has not been optimized, the current hidden crack detection database is the initial hidden crack detection database; if the hidden crack detection database has been optimized, the current hidden crack detection database is the hidden crack detection database after the last optimization.

[0057] S130, establishing a chip hidden crack detection model according to a target hidden crack detection database.

[0058] Specifically, the chip hidden crack detection model is used to detect and identify hidden crack defects in the chip, and a hidden crack detection database is embedded in the chip hidden crack detection model. The chip hidden crack detection model established according to the target hidden crack detection database is a chip hidden crack detection model that has been optimized multiple times.

[0059] The method for establishing a hidden crack detection model provided by an embodiment of the present invention includes: establishing an initial hidden crack detection database; acquiring training sample sets multiple times, and after each acquisition of the training sample sets, supplementing the hidden crack chip images in the initial hidden crack detection database according to the currently acquired training sample sets, so as to gradually optimize the initial hidden crack detection database and obtain a target hidden crack detection database; and establishing a chip hidden crack detection model according to the target hidden crack detection database. The technical solution provided by an embodiment of the present invention can gradually optimize the initial hidden crack detection database until the target hidden crack detection database is obtained by acquiring training sample sets multiple times, and after each acquisition of the training sample sets, supplementing the hidden crack chip images in the initial hidden crack detection database according to the currently acquired training sample sets, without having to acquire a large number of samples at one time to establish the target hidden crack detection database, so that the repetition rate of the acquired samples can be reduced, and the amount of sample acquisition can be reduced, so that the detection accuracy of the chip hidden crack detection model can be ensured while improving the efficiency of establishing the hidden crack detection model.

[0060] Figure 2 is a flowchart of another method for establishing a hidden crack detection model provided by an embodiment of the present invention, with reference to Figure 2 , the method for establishing the hidden crack detection model includes:

[0061] S210, acquiring a first preset number of hidden crack chip images and a second preset number of standard chip images.

[0062] S220: Establish an initial hidden crack detection database according to a first preset number of hidden crack chip images and a second preset number of standard chip images.

[0063] Specifically, establishing an initial hidden crack detection database includes: acquiring a first preset number of hidden crack chip images and a second preset number of standard chip images; and establishing an initial hidden crack detection database based on the first preset number of hidden crack chip images and the second preset number of standard chip images.

[0064] It can be understood that, while the initial hidden crack detection database stores a first preset number of hidden crack chip images, it also stores a second preset number of standard chip images. The standard chip images stored in the initial hidden crack detection database can be obtained by collecting a corresponding number of sample chips without hidden cracks through an infrared image acquisition device. For example, if the initial hidden crack detection database stores 1000 standard chip images, the infrared image acquisition device is used to collect images of 1000 standard chips (qualified chips). Among them, the chip corresponding to the standard chip image stored in the initial hidden crack detection database is the same model as the chip corresponding to the stored hidden crack chip image. It should be noted that although these 1000 standard chips are of the same model, due to the differences in the surface texture of the chips after processing, the images will show differences, which is equivalent to 1000 different images.

[0065] S230, acquiring training sample sets multiple times, and after each acquisition of the training sample set, supplementing the hidden crack chip images in the initial hidden crack detection database according to the currently acquired training sample set, so as to gradually optimize the initial hidden crack detection database and obtain a target hidden crack detection database.

[0066] S240: Establish a chip hidden crack detection model according to a target hidden crack detection database.

[0067] The method for establishing a hidden crack detection model according to an embodiment of the present invention can store a first preset number of hidden crack chip images in an initial hidden crack detection database while also storing a second preset number of standard chip images. After each training sample in the currently acquired training sample set is compared with the hidden crack chip image in the current hidden crack detection database, it can continue to be compared with the standard chip image, thereby eliminating the possibility that the training sample screened out by the comparison and ready to be stored in the hidden crack detection database is a standard chip image obtained by collecting a standard chip. If the training sample screened out by the comparison is a standard chip image, the training sample can be eliminated.

[0068] Moreover, in the process of using the established chip hidden crack detection model to detect the chip to be tested, if the detection result of the chip hidden crack detection model is that the chip to be tested does not have hidden cracks, it can also be confirmed whether the chip to be tested is a standard chip, thereby reducing the probability of misjudgment due to incomplete hidden crack information stored in the chip hidden crack detection model, thereby further improving the accuracy of the chip hidden crack detection model detection.

[0069] Furthermore, if the detection result of the hidden crack detection model is that the chip to be tested is neither a hidden crack chip nor a standard chip, the detection system can generate an abnormal prompt message to notify manual reconfirmation. If the manual confirms that the chip to be tested is a hidden crack chip, the image of the chip to be tested can also be fed back to the hidden crack detection model and stored in the hidden crack detection database, thereby further improving the information stored in the hidden crack detection model and reducing the missed detection rate of the hidden crack detection model.

[0070] Figure 3 is a flowchart of another method for establishing a hidden crack detection model provided by an embodiment of the present invention, with reference to Figure 3 , the method for establishing the hidden crack detection model includes:

[0071] S310: Establish an initial hidden crack detection database.

[0072] S320, collecting images of the first batch of hidden crack chips to obtain a first training sample set.

[0073] S330, compare each first training sample in the first training sample set with the images stored in the initial hidden crack detection database, and determine whether there is an unidentified first training sample; if so, add the unidentified first training sample as a newly added hidden crack chip image to the initial hidden crack detection database to obtain the first optimized hidden crack detection database.

[0074] S340, collecting images of the second batch of hidden crack chips to obtain a second training sample set.

[0075] S350, compare and detect each second training sample in the second training sample set with the images stored in the first optimized hidden crack detection database, and determine whether there are unidentified second training samples; if so, add the unidentified second training samples as newly added hidden crack chip images to the first optimized hidden crack detection database to obtain a second optimized hidden crack detection database.

[0076] S360: Capture images of the third batch of hidden crack chips to obtain a third training sample set.

[0077] S370, compare each third training sample in the third training sample set with the images stored in the second optimized hidden crack detection database, and determine whether there is an unidentified third training sample; if so, add the unidentified third training sample as a newly added hidden crack chip image to the second optimized hidden crack detection database to obtain a third optimized hidden crack detection database; and so on, until it is determined that the target hidden crack detection database is obtained.

[0078] S380, establishing a chip hidden crack detection model according to a target hidden crack detection database.

[0079] The method for establishing a hidden crack detection model of the embodiment of the present invention refines and supplements the steps of obtaining training sample sets multiple times, and after each training sample set is obtained, supplementing the hidden crack chip images in the initial hidden crack detection database according to the currently obtained training sample set, so as to gradually optimize the initial hidden crack detection database and obtain the target hidden crack detection database. After each training sample set is obtained, each training sample in the training sample set is input into the current hidden crack detection database, and the hidden crack chip images stored in the current hidden crack detection database are traversed. If a training sample contains hidden crack information that is not stored in the current hidden crack detection database, the training sample will be missed, that is, the recognition is not successful. The unrecognized training samples are supplemented as newly added hidden crack chip images in the current hidden crack detection database, so that the hidden crack information stored in the current hidden crack detection database can be improved, thereby being optimized. Among them, image acquisition can be performed by an infrared image acquisition device, and the image acquisition device is used to acquire images of multiple batches of hidden crack chips to acquire training sample sets, so that the internal images of the hidden crack chips can be acquired, thereby facilitating the recognition and sampling of hidden cracks inside the hidden crack chips. The number of hidden crack chips collected in different batches may be the same or different.

[0080] Optionally, when the initial hidden crack detection database stores a second preset number of standard chip images, the training sample set acquired each time may include only hidden crack chip images, or may include both hidden crack chip images and standard chip images. Preferably, the training sample set acquired each time includes only hidden crack chip images, which can improve the efficiency of improving the initial hidden crack detection database.

[0081] Based on the above embodiments, optionally, determining to obtain a target hidden crack detection database includes:

[0082] Determine whether the number of times the training sample set is obtained meets the target number, and / or determine whether the hidden crack chip images stored in the currently optimized hidden crack detection database meet the target number, and / or determine whether the recognition success rate of the currently optimized hidden crack detection database is greater than or equal to the preset recognition success rate; if so, determine that the target hidden crack detection database is obtained.

[0083] Specifically, the judgment condition for determining whether the target hidden crack detection database is obtained can be: one of the number of times the training sample set is obtained, the number of hidden crack chip images stored in the optimized hidden crack detection database, and the recognition success rate of the currently optimized hidden crack detection database, or a combination of any two of them, or all three conditions are met. Among them, the target number of times the training sample set is obtained, the target number of hidden crack chip images stored in the target hidden crack detection database, and the preset recognition success rate of the target hidden crack detection database can all be set according to actual conditions.

[0084] Based on the above embodiments, optionally, before adding the unidentified training samples to the current hidden crack detection database, the method further includes:

[0085] Determine whether the unrecognized training samples have clear images;

[0086] If it is clear, the unidentified training samples are added to the current hidden crack detection database.

[0087] Specifically, since there is a situation where the training sample is not successfully recognized because the training sample image is not clear, before the unrecognized training sample is added to the last optimized hidden crack detection database, it can be determined whether the unrecognized training sample has a clear image. If it is clear, the unrecognized training sample is added to the last optimized hidden crack detection database to prevent abnormal hidden crack chip images from being stored in the hidden crack detection database. If it is determined that the image of the unrecognized training sample is not clear, the training sample is removed.

[0088] Furthermore, if it is determined that the image of the unrecognized training sample is unclear, the method for establishing the hidden crack detection model also includes: re-capturing the image of the hidden crack chip corresponding to the unrecognized and unclear training sample, and comparing the re-acquired training sample with the image in the hidden crack detection database after the last optimization; if there are training samples that are not recognized again, the training samples that are not recognized again are added to the hidden crack detection database after the last optimization, thereby optimizing the current hidden crack detection database.

[0089] Figure 4 is a flowchart of another method for establishing a hidden crack detection model provided by an embodiment of the present invention, with reference to Figure 4 , the method for establishing the hidden crack detection model includes:

[0090] S410: Determine the model of the chip to be tested.

[0091] S420, acquiring a first preset number of hidden crack chip images and a second preset number of standard chip images through an infrared detection lens, wherein the chips corresponding to the standard chip images and the chips corresponding to the hidden crack chip images are both models of the chips to be tested.

[0092] S430, feeding back a first preset number of hidden crack chip images and a second preset number of standard chip images to a hidden crack identification terminal to establish an initial hidden crack detection database.

[0093] S440, placing a hidden crack identification system consisting of an infrared detection lens and a hidden crack identification terminal in a defect detection production line.

[0094] S450, collecting images of a batch of hidden crack chips through an infrared detection lens to obtain a training sample set, and comparing each training sample in the training sample set with an image stored in an initial hidden crack detection database.

[0095] S460, determining whether there are unrecognized training samples during the comparison process; if so, executing step S470; if not, executing step S490;

[0096] S470, determining whether the image of the unrecognized training sample is clear; if it is clear, executing step S490; if it is not clear, executing step S480.

[0097] S480, re-capture images of the hidden crack chips corresponding to the unrecognized and unclear training samples, and compare and judge whether there are unrecognized training samples again. If yes, execute step S490; if no, execute step S4100.

[0098] S490: Supplement the unidentified training samples into the current hidden crack detection database to optimize the current hidden crack detection database.

[0099] S4100, determine whether the target hidden crack detection database is obtained; if so, execute step S4110, if not, return to execute step S450.

[0100] S4110. Establish a chip hidden crack detection model according to a target hidden crack detection database.

[0101] S4120. Perform qualification inspection on the chip hidden crack detection model.

[0102] Specifically, the infrared image acquisition device includes an infrared detection lens. The infrared detection lens and the hidden crack identification terminal are used to form a hidden crack identification system. The hidden crack identification terminal can be a computer, a mobile phone, a tablet or other terminals. The qualified inspection of the chip hidden crack detection model can include: performing hidden crack detection on a small batch of chips to be tested through the chip hidden crack detection model, and manually sampling and re-inspecting the detection results of the hidden crack detection model to determine whether the detection results of the chip hidden crack detection model are accurate. If accurate, it is confirmed that the chip hidden crack detection model is qualified.

[0103] Optionally, after each acquisition of a training sample set, the hidden crack chip images in the initial hidden crack detection database are supplemented according to the currently acquired training sample set to gradually optimize the initial hidden crack detection database. In the step, all the hidden crack chip images in the training sample set that are different from those stored in the current hidden crack detection database can be first screened out, and then the screened out (unidentified) hidden crack chip images can be uniformly supplemented into the current hidden crack detection database.

[0104] Alternatively, the training samples in the training sample set may be compared with the hidden crack chip images stored in the current hidden crack detection database in sequence, and added to the current hidden crack detection database in sequence. Figure 5 is a flowchart of another method for establishing a hidden crack detection model provided by an embodiment of the present invention, with reference to Figure 5 , the method for establishing the hidden crack detection model includes:

[0105] S510: Determine the model of the chip to be tested.

[0106] S520, obtaining a first preset number of hidden crack chip images and a second preset number of standard chip images, based on the first preset number of hidden crack chip images and the second preset number of standard chip images; wherein the chips corresponding to the standard chip images and the chips corresponding to the hidden crack chip images are both models of the chips to be tested.

[0107] S530: Feedback a first preset number of hidden crack chip images and a second preset number of standard chip images to a hidden crack identification terminal to establish an initial hidden crack detection database.

[0108] S540, placing a hidden crack identification system consisting of an infrared detection lens and a hidden crack identification terminal in a defect detection production line.

[0109] S550: Capture images of a third preset number of hidden crack chips through an infrared detection lens to obtain a training sample set.

[0110] S560: Compare the training samples in the training sample set with the images stored in the initial crack detection database in sequence.

[0111] S570, determine whether the currently compared training sample is an unrecognized training sample; if so, execute step S580, if not, execute step S590.

[0112] S580: Supplement the unidentified training samples into the current hidden crack detection database to optimize the current hidden crack detection database.

[0113] S590, determine whether the training samples in the training sample set have been compared; if the comparison is completed, execute step S5100; if the comparison is not completed, return to execute step S560.

[0114] S5100, collect a fourth preset number of images through an infrared detection lens to obtain a training test sample set; and compare each test sample in the test sample set with the image stored in the initial hidden crack detection database to determine the current recognition success rate.

[0115] S5110. Determine whether the current recognition success rate meets the target recognition success rate. If so, execute step S5120. If not, return to execute step S550.

[0116] S5120, determine to obtain a target hidden crack detection database, and establish a chip hidden crack detection model according to the target hidden crack detection database.

[0117] S5130. Perform qualification inspection on the chip hidden crack detection model.

[0118] Specifically, the training samples in the training sample set are compared in sequence, and added to the current hidden crack detection database in sequence, which can further reduce the repetition rate of samples in the hidden crack detection database. A small batch of chips to be tested are tested for hidden cracks using a chip hidden crack detection model, and the detection results of the hidden crack detection model are manually sampled and re-inspected to confirm whether the detection results of the chip hidden crack detection model are all accurate, and after confirming that they are all accurate, the chip hidden crack detection model is qualified. Optionally, after step S570 determines whether the currently compared training sample is an unrecognized training sample and step S5110 determines whether the current recognition success rate meets the target recognition success rate, it can include: determining whether the image of the unrecognized hidden crack chip is clear, and if it is clear, feeding back the unrecognized hidden crack chip image to the hidden crack detection database; if it is not clear, reacquiring the image for detection and judgment.

[0119] The embodiment of the present invention further provides a chip hidden crack detection method, comprising:

[0120] Acquire an image of the chip to be tested;

[0121] The image of the chip to be tested is input into the chip hidden crack detection model, and the image of the chip to be tested is analyzed by the chip hidden crack detection model to obtain the detection result.

[0122] The chip hidden crack detection model is constructed by the hidden crack detection model establishment method described in any embodiment of the present invention.

[0123] Specifically, obtaining the image of the chip to be tested includes: detecting the image of the chip to be tested by an infrared image acquisition device. Inputting the image of the chip to be tested into the chip hidden crack detection model, and analyzing the image of the chip to be tested by the chip hidden crack detection model, including: inputting the image of the chip to be tested into the chip hidden crack detection model, traversing the hidden crack detection database in the chip hidden crack detection model, and obtaining similar images; determining whether the similar image is a hidden crack chip image, and if so, reading the hidden crack information parameters carried by the similar image to obtain the specific hidden crack type, position and size of the chip to be tested; if the similar image is a standard chip image, it is determined that there is no hidden crack defect inside the chip to be tested. The chip hidden crack detection method provided by the embodiment of the present invention improves the accuracy and efficiency of chip hidden crack detection.

[0124] The embodiment of the present invention further provides a defect detection production line, which includes a chip conveyor belt 10, a gear 20, a chip bonding head 40, a material tray 90, and a hidden crack recognition system 60. The hidden crack recognition system 60 includes an infrared detection lens 61 and a hidden crack recognition terminal, and the chip hidden crack detection model is integrated in the hidden crack recognition terminal. The infrared detection lens 61 is used as an infrared image acquisition device for acquiring an image of the chip 30 to be tested.

[0125] The upper surface of the chip conveyor belt 10 includes a plurality of slots 12 for placing the chip to be tested 30. The chip bonding head 40 is used to place the chip to be tested 30 in the slots 12 of the chip conveyor belt 10. The edge of the chip conveyor belt 10 includes a plurality of traction holes 11 arranged at intervals. The gear 20 is stuck in the traction hole 11 to drive the chip conveyor belt 10 to move.

[0126] When the chip to be tested 30 is transferred to the position of the infrared detection lens 61, the infrared detection lens 61 collects the image of the chip to be tested 30 and feeds it back to the hidden crack identification terminal. The hidden crack identification terminal stores the chip hidden crack detection model constructed by the method for establishing the hidden crack detection model described in any embodiment of the present invention, and performs hidden crack detection on the chip to be tested 30 through the chip hidden crack detection model. An infrared detection station and a hidden crack identification terminal are added to the conveyor line, and the conveyor line can convey the chip to be tested 30 to the detection station. The infrared detection lens 61 collects images in real time to the hidden crack identification terminal to realize automatic detection of hidden crack defects.

[0127] The embodiment of the present invention further provides a chip defect detection method, including the chip hidden crack detection method described in any embodiment of the present invention, which has the same technical effect and will not be described in detail here.

[0128] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for establishing a hidden crack detection model, characterized in that: include: Establish an initial hidden crack detection database; Acquire training sample sets multiple times, and after each acquisition of the training sample set, supplement the hidden crack chip images in the initial hidden crack detection database according to the currently acquired training sample set, so as to gradually optimize the initial hidden crack detection database and obtain a target hidden crack detection database; A chip hidden crack detection model is established according to the target hidden crack detection database.

2. The method for establishing a hidden crack detection model according to claim 1, characterized in that: The step of establishing an initial hidden crack detection database comprises: Acquire a first preset number of hidden crack chip images and a second preset number of standard chip images; The initial hidden crack detection database is established according to a first preset number of hidden crack chip images and a second preset number of standard chip images.

3. The method for establishing a hidden crack detection model according to claim 1, characterized in that: Acquire training sample sets multiple times, and after each acquisition of the training sample sets, supplement the hidden crack chip images in the initial hidden crack detection database according to the currently acquired training sample sets, so as to gradually optimize the initial hidden crack detection database and obtain a target hidden crack detection database, including: Capturing images of the first batch of hidden crack chips to obtain a first training sample set; Compare each first training sample in the first training sample set with the images stored in the initial hidden crack detection database, and determine whether there is an unrecognized first training sample; if so, add the unrecognized first training sample as a newly added hidden crack chip image to the initial hidden crack detection database to obtain a first optimized hidden crack detection database; Capturing images of the second batch of hidden crack chips to obtain a second training sample set; Compare and detect each second training sample in the second training sample set with the images stored in the first optimized hidden crack detection database, and determine whether there is an unidentified second training sample; if so, add the unidentified second training sample as a newly added hidden crack chip image to the first optimized hidden crack detection database to obtain a second optimized hidden crack detection database; Capturing images of the third batch of hidden crack chips to obtain a third training sample set; Compare each third training sample in the third training sample set with the images stored in the second optimized hidden crack detection database, and determine whether there is an unidentified third training sample; if so, add the unidentified third training sample as a newly added hidden crack chip image to the second optimized hidden crack detection database to obtain a third optimized hidden crack detection database; And so on, until it is determined that the target hidden crack detection database is obtained; Preferably, the image acquisition is performed by an infrared image acquisition device.

4. The method for establishing a hidden crack detection model according to claim 3, characterized in that: Confirm that the target hidden crack detection database is obtained, including: Determine whether the number of times the training sample set is obtained meets the target number, and / or determine whether the hidden crack chip images stored in the currently optimized hidden crack detection database meet the target number, and / or determine whether the recognition success rate of the currently optimized hidden crack detection database is greater than or equal to the preset recognition success rate; if so, determine that the target hidden crack detection database is obtained.

5. The method for establishing a hidden crack detection model according to claim 3, characterized in that: Before adding unidentified training samples to the current hidden crack detection database, it also includes: Determine whether the unrecognized training samples have clear images; If it is clear, the unidentified training samples are added to the current hidden crack detection database.

6. The method for establishing a hidden crack detection model according to claim 5, characterized in that: If it is determined that the image of the unrecognized training sample is not clear, it also includes: Re-acquire images of the hidden crack chips corresponding to the unidentified and unclear training samples, and compare the re-acquired training samples with the images in the current hidden crack detection database; If there are training samples that are not recognized again, the training samples that are not recognized again are added to the current hidden crack detection database.

7. The method for establishing a hidden crack detection model according to claim 1, characterized in that: After determining the chip hidden crack detection numerical model according to the target hidden crack detection database, the method further includes: Performing hidden crack detection on a preset number of chips to be tested using the chip hidden crack detection model; The detection results of the hidden crack detection numerical model are manually sampled and re-inspected to conduct a qualified inspection of the chip hidden crack detection numerical model.

8. A chip hidden crack detection method, characterized in that: include: Acquire an image of the chip to be tested; The image of the chip to be tested is input into a chip hidden crack detection model, and the image of the chip to be tested is analyzed by the chip hidden crack detection model to obtain a detection result; wherein the chip hidden crack detection model is constructed by the method for establishing a hidden crack detection model described in any one of claims 1 to 7.

9. The chip hidden crack detection method according to claim 8, characterized in that: The step of acquiring an image of the chip to be tested comprises: The image of the chip to be tested is detected by an infrared image acquisition device.

10. A chip defect detection method, characterized in that: A method for detecting hidden cracks in a chip comprising the method described in claim 8 or 9.

Images