CN116128787A - Lead frame shipment method - Google Patents
Lead frame shipment method Download PDFInfo
- Publication number
- CN116128787A CN116128787A CN202111336780.3A CN202111336780A CN116128787A CN 116128787 A CN116128787 A CN 116128787A CN 202111336780 A CN202111336780 A CN 202111336780A CN 116128787 A CN116128787 A CN 116128787A
- Authority
- CN
- China
- Prior art keywords
- lead frame
- flaw
- shipment
- deviation
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 83
- 238000013136 deep learning model Methods 0.000 claims abstract description 62
- 238000012545 processing Methods 0.000 claims abstract description 23
- 230000007547 defect Effects 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims description 24
- 238000011156 evaluation Methods 0.000 claims description 12
- 238000007747 plating Methods 0.000 claims description 6
- 101100134058 Caenorhabditis elegans nth-1 gene Proteins 0.000 claims description 4
- 238000007689 inspection Methods 0.000 abstract description 19
- 238000004519 manufacturing process Methods 0.000 description 9
- 238000007726 management method Methods 0.000 description 8
- 239000000463 material Substances 0.000 description 7
- 238000001514 detection method Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000002950 deficient Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000002716 delivery method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30148—Semiconductor; IC; Wafer
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
A method of shipping a leadframe, comprising: acquiring a source image of the first lead frame according to the first lead frame; detecting a first flaw according to the source image of the first lead frame, wherein the first flaw is a flaw of the first lead frame; when a plurality of first flaws are detected, a plurality of corresponding first flaw images are obtained; processing a plurality of first flaw images by adopting a first deep learning model to obtain the severity level of each first flaw; and judging the shipment grade of the first lead frame according to the severity grade of the first defects. By the shipment method of the lead frame, management and labor cost can be reduced, and quality inspection efficiency and stability are improved.
Description
Technical Field
The invention relates to the technical field of lead frame detection, in particular to a lead frame shipment method.
Background
The quality of the lead frame as a chip carrier for many integrated circuits is reliable and stable, which determines the function and performance of the subsequent final semiconductor product, so that the quality inspection of the lead frame material is required in the actual production process. The lead frame material sheet has a plurality of particles, the size is small, and the manual inspection needs to be visually judged by means of a microscope one by one, so that time and labor are wasted, and the inspection stability is difficult to ensure. To improve quality inspection efficiency and stability, vision-based automated leadframe inspection devices have been designed and manufactured.
Conventional lead frame inspection equipment automatically inspects input lead frame materials, locates all defects on the materials, outputs relevant parameters such as size, position, contrast and the like of each defect, and then manually classifies and judges the defects of the lead frame materials: the method is characterized in that the method is directly scrapped with higher hazard degree, and the method is judged to be good in function with low hazard degree or without influencing the function.
However, when defects of the lead frame materials are classified and classified manually, quality inspection efficiency is low, management and labor cost are improved, and stability of quality inspection is poor due to manual quality inspection.
Disclosure of Invention
The invention solves the technical problem of providing a delivery method of a lead frame, so as to reduce management and labor cost and improve quality inspection efficiency and stability.
In order to solve the above technical problems, the technical solution of the present invention provides a method for delivering lead frames, including: acquiring a source image of the first lead frame according to the first lead frame; detecting a first flaw according to the source image of the first lead frame, wherein the first flaw is a flaw of the first lead frame; when a plurality of first flaws are detected, a plurality of corresponding first flaw images are obtained; processing a plurality of first flaw images by adopting a first deep learning model to obtain the severity level of each first flaw; and judging the shipment grade of the first lead frame according to the severity grade of the first defects.
Optionally, the method for detecting the first flaw according to the source image of the first lead frame includes: providing a standard template image; providing a preset deviation range corresponding to the standard template image; and detecting a first flaw according to the standard template image, the preset deviation range and the source image of the first lead frame.
Optionally, the standard template graph includes a plurality of standard regions, and the preset deviation range includes a sub-preset deviation range corresponding to each standard region.
Optionally, the method for detecting the first flaw according to the standard template image, the preset deviation range and the source image of the first lead frame includes: acquiring a plurality of deviation images and a plurality of corresponding deviation feature data according to the source image of the first lead frame and the standard template image, wherein the deviation feature data comprises deviation position information and deviation feature parameters, and each deviation position information also corresponds to one of a plurality of standard areas; detecting that the first flaw corresponding to the random deviation image exists when the deviation characteristic parameter corresponding to the random deviation image exceeds a designated sub-preset deviation range corresponding to the random deviation image, wherein the designated sub-preset deviation range corresponding to the random deviation image is a sub-preset deviation range of a standard area corresponding to deviation position information corresponding to the random deviation image; and when a plurality of first flaws are detected, the method for acquiring the corresponding first flaw images comprises the following steps: and acquiring a first flaw image according to the deviation image corresponding to the first flaw.
Optionally, the types of standard regions include at least one of plating and critical regions, half-etched regions, functional regions, nonfunctional regions, rail and pilot hole regions, and low threshold regions.
Optionally, the deviation feature parameter includes at least one of an area parameter, a diagonal length parameter, and a contrast parameter corresponding to the deviation image.
Optionally, the method for forming the first deep learning model includes: collecting a plurality of flaw sample images, and position information and flaw characteristic parameters corresponding to each flaw sample image; marking the severity level of the plurality of flaw sample images according to the corresponding position information and flaw characteristic parameters to form a plurality of flaw processing images; the initial first deep learning model is trained a plurality of iterations based on the plurality of flaw processing images.
Optionally, the nth method in the multiple iterative training includes: performing an nth iterative training on a first deep learning model formed by the historical nth-1 iterative training based on an nth one of the plurality of flaw processing images, wherein n is a natural number, and when n=1, the first deep learning model formed by the historical nth-1 iterative training is the initial first deep learning model; evaluating a first deep learning model formed by the nth iterative training to obtain an nth evaluation result; and stopping the iterative training when the nth evaluation result meets a preset condition.
Optionally, the method further comprises: evaluating whether the severity level of the at least 1 first blemish predicted by the first deep learning model is incorrect; when the error is predicted, marking the severity level of the first flaw image corresponding to the first flaw to form a first flaw processing image; based on the first flaw processed images, a first deep learning model is trained that processes the first flaw images.
Optionally, the method for determining the shipment level of the first lead frame according to the severity level of the plurality of first flaws includes: providing a preset shipment rule, wherein the preset shipment rule comprises a corresponding relation between shipment grade and the severity grade of the flaw; and judging a plurality of first flaws and the severity level of the first flaws according to the preset shipment rule, and obtaining the shipment level of the first lead frame.
Optionally, the method for determining the shipment level of the first lead frame according to the severity level of the plurality of first flaws includes: providing a preset shipment rule, wherein the preset shipment rule comprises shipment grade, severity grade of flaws and corresponding relation among clients; providing a customer corresponding to the first lead frame; and judging a plurality of first flaws, the severity levels of the first flaws and customers corresponding to the first lead frames according to the preset shipment rules, and obtaining the shipment level of the first lead frames.
Optionally, the shipment level includes a good sheet level and a bad sheet level, and the shipment method of the lead frame further includes: receiving the first lead frame when the shipment level of the first lead frame is judged to be a good sheet level; and refusing to receive the first lead frame when the shipment grade of the first lead frame is judged to be the bad chip grade.
Optionally, the shipment level further includes a rework level, and the shipment method of the lead frame further includes: and when the shipment grade of the first lead frame is judged to be the reworking grade, reworking the first lead frame.
Optionally, the method further comprises: and when the first flaw is not detected, judging that the shipment grade of the first lead frame is a good grade.
Compared with the prior art, the technical scheme of the embodiment of the invention has the following beneficial effects:
in the shipment method of the lead frame, when a plurality of first flaws are detected, a plurality of corresponding first flaw images are obtained; processing a plurality of first flaw images by adopting a first deep learning model to obtain the severity level of each first flaw; and judging the shipment grade of the first lead frame according to the severity grade of the first defects. Therefore, the shipment method of the lead frame can judge the shipment grade of the first lead frame with high automation, thereby effectively reducing the management and labor cost and improving the quality inspection efficiency and stability.
Drawings
Fig. 1 is a flow chart of a method for shipping a lead frame according to an embodiment of the invention;
FIG. 2 is a schematic view of a source image of a first leadframe according to an embodiment of the invention;
FIGS. 3-4 are flow charts of a method for detecting a first defect according to an embodiment of the present invention;
FIG. 5 is a schematic illustration of a standard template image according to an embodiment of the present invention;
FIGS. 6 and 7 are flow diagrams of a method of forming a first deep learning model according to an embodiment of the present invention;
fig. 8 is a flowchart illustrating a method for determining a shipment level of a first leadframe according to an embodiment of the present invention;
fig. 9 is a flowchart illustrating a method for determining a shipment level of a first leadframe according to another embodiment of the present invention.
Detailed Description
As described in the background art, when defects of lead frame materials are classified and classified manually, on one hand, quality inspection efficiency is still low, and management and labor cost can be improved, and on the other hand, quality inspection by manpower can result in poor quality inspection stability.
In order to solve the technical problems, the technical scheme of the invention provides a shipment method of a lead frame, which is characterized in that when a plurality of first flaws are detected, a plurality of corresponding first flaw images are obtained, then, the first flaw images are processed by adopting a first deep learning model to obtain the severity level of each first flaw, and the shipment level of the first lead frame is judged according to the severity level of the plurality of first flaws, so that the management and labor cost can be reduced, and the quality inspection efficiency and stability are improved.
In order to make the above objects, features and advantages of the present invention more comprehensible, embodiments accompanied with figures are described in detail below.
Fig. 1 is a flow chart of a method for shipping a lead frame according to an embodiment of the invention.
Referring to fig. 1, the method for shipping the lead frame includes:
step S100, acquiring a source image of a first lead frame according to the first lead frame;
when a plurality of first flaws are detected, executing step S300 to obtain a plurality of corresponding first flaw images;
step S400, processing a plurality of first flaw images by adopting a first deep learning model to obtain the severity level of each first flaw;
step S500, judging the shipment grade of the first lead frame according to the severity grade of a plurality of first flaws.
When a plurality of first flaws are detected, a plurality of corresponding first flaw images are obtained, then the first deep learning model is adopted to process the plurality of first flaw images, the severity level of each first flaw is obtained, and the shipment level of the first lead frame is judged according to the severity level of the plurality of first flaws. Therefore, the shipment method of the lead frame can judge the shipment grade of the first lead frame with high automation, thereby effectively reducing the management and labor cost and improving the quality inspection efficiency and stability.
The following detailed description refers to the accompanying drawings.
Referring to fig. 2, fig. 2 is a schematic view of a source image of a first lead frame according to an embodiment of the invention, and a source image 100 of the first lead frame is obtained according to the first lead frame (not shown).
The first lead frame refers to a sheet-shaped lead frame to be shipped currently.
In this embodiment, the source image 100 of the first leadframe is acquired using an optical stage, which includes a CCD camera or the like.
It should be understood that, according to the practical situation, any apparatus that can acquire the source image 100 of the first lead frame may be used by a person skilled in the art, and the apparatus and method for acquiring the source image 100 of the first lead frame should not be taken as a feature that limits the protection scope of the present invention.
With continued reference to fig. 1, for step S200, the first defect is a defect of the first lead frame.
Fig. 3 to 4 are schematic flow diagrams of a method for detecting a first defect according to an embodiment of the present invention, and fig. 5 is a schematic diagram of a standard template image according to an embodiment of the present invention.
Referring to fig. 3, for step S200, a method for detecting a first defect according to a source image of the first lead frame includes:
step S210, providing a standard template image;
step S220, providing a preset deviation range corresponding to the standard template image;
step S230, detecting a first flaw according to the standard template image, the preset deviation range and the source image of the first lead frame.
The standard template image is acquired based on a flawless sheet-like lead frame.
Reference data can be acquired based on the standard template image, the reference data including: at least one of a number of area parameters, a number of diagonal length parameters, and a number of contrast parameters corresponding to the standard template image.
The preset deviation range is a deviation range taking the reference data as a reference, and the preset deviation range can be a proportion or a numerical value.
In this embodiment, the standard template image 200 (shown in fig. 5) includes a plurality of standard regions 201 (shown in fig. 5).
It should be understood that the images of the standard regions 201 may or may not be identical.
In the present embodiment, the types of the standard regions 201 include at least one of Plating and Key regions (Plating and Key), half-etched regions (Half-etched), functional regions (Functional), non-Functional regions (Non-Functional), track and positioning hole regions (Rail and Pilot Hole), and Low threshold regions (Large Low-value).
In this embodiment, the preset deviation range includes sub-preset deviation ranges corresponding to the respective standard areas 201. Thus, the risk of overdischarge or missed detection in the first flaw detection is better reduced.
Specifically, the deviation range for determining whether the first defect is based on the reference data may be different for different types of standard areas 201. For example, since the sheet-like lead frame portion corresponding to the plating and critical area is an important area, the first defect can be judged with a stricter standard by making the sub-preset deviation range corresponding to the plating and critical area smaller. Accordingly, since the low threshold region is a non-important region that needs to be cut later, the first flaw can be judged with a more relaxed standard by making the sub-preset deviation range corresponding to the low threshold region larger. Thus, by the sub-preset deviation ranges corresponding to the respective standard areas 201, the risk of overdischarging or missing detection at the time of the first flaw detection can be reduced well.
In other embodiments, for step S220, the method for providing the preset deviation range corresponding to the standard template image includes:
step S221, pre-configuring and storing initial preset deviation ranges corresponding to a plurality of clients one by one;
step S222, according to the customer information corresponding to the customer, obtaining and providing the preset deviation range in a plurality of initial preset deviation ranges.
Wherein, the preset deviation range in step S222 is one of several initial preset deviation ranges.
In some practical application scenarios, different customers have different shipment standards. Through step S221 and step S222, the detection of the first flaw can be performed on the source image of the first lead frame based on different preset deviation ranges on the basis of the standard template image. Therefore, in the actual mass production process, the goods can be delivered in a targeted manner according to the requirements of different clients conveniently and rapidly, and further, the management and labor cost is reduced better, and the quality inspection efficiency is improved.
Referring to fig. 4, for step S230, a method for detecting a first defect according to the standard template image, the preset deviation range and the source image of the first lead frame includes:
step S231, acquiring a plurality of deviation images and a plurality of corresponding deviation feature data according to the source image of the first lead frame and the standard template image;
step S232, detecting that the first flaw corresponding to the arbitrary deviation image exists when the deviation characteristic parameter corresponding to the arbitrary deviation image exceeds the specified sub-preset deviation range corresponding to the arbitrary deviation image.
Specifically, in step S231, the source image 100 of the first lead frame and the standard template image 200 are compared, whereby a plurality of deviation images and a corresponding plurality of deviation feature data are acquired based on the difference therebetween.
In this embodiment, the offset image may be an image obtained directly from the source image 100 of the first lead frame, where the source image 100 of the first lead frame is different from the standard template image 200, or may be an image obtained by performing image processing based on the image.
At the same time, a corresponding plurality of deviation feature data are acquired based on the image of the source image 100 of the first lead frame that is different from the standard template image 200.
In this embodiment, the deviation feature data includes: deviation position information and deviation characteristic parameters.
The offset position information is used to locate the position of the image at each of the differences, and each offset position information also corresponds to one of the standard areas 201. Thus, according to the deviation position information, it is possible to determine the sub-preset deviation range required for comparison among the plurality of sub-preset deviation ranges when the first flaw is determined.
Specifically, the deviation position information may be coordinate information or a type of a corresponding standard area, or the like.
The deviation characteristic parameters include: at least one of an area parameter, a diagonal length parameter, and a contrast parameter corresponding to the deviation image.
In step S232, the specified sub-preset deviation range corresponding to the arbitrary deviation image is: and in the plurality of sub-preset deviation ranges, the deviation position information corresponding to the random deviation image corresponds to the sub-preset deviation range of the standard area.
Specifically, according to the correspondence between the deviation position information and the standard area, a specific sub-preset deviation range corresponding to an arbitrary deviation image is determined among a plurality of sub-preset deviation ranges.
Referring to fig. 1, for step S300, when a plurality of first flaws are detected, the method for obtaining a plurality of corresponding first flaw images further includes: and acquiring a first flaw image according to the deviation image corresponding to the first flaw.
In this embodiment, the first defect image may be a deviation image corresponding to the first defect directly, or may be an image obtained by performing image processing on the deviation image corresponding to the first defect. Specifically, the image processing is, for example, a contoured image formed from the contours of the deviation image, or a contour-simplified image formed from the contours of the deviation image, or the like.
With continued reference to fig. 1, the first deep learning model is configured to predict a severity level of a corresponding first flaw according to the first flaw image.
Specifically, the number of severity levels includes at least 2 levels.
In this embodiment, the severity level includes 1 level to 9 levels, for a total of 9 levels.
It is to be understood that the number of levels of severity level may also be 2 levels, or any number of levels greater than 2. The number of levels of severity level can be set by those skilled in the art according to actual needs.
Fig. 6 to 7 are flowcharts illustrating a method for forming a first deep learning model according to an embodiment of the present invention.
Referring to fig. 6, fig. 6 is a flowchart illustrating a method for forming the first deep learning model according to an embodiment of the invention, for the first deep learning model in step S400, the method for forming the first deep learning model includes:
step S410, collecting a plurality of flaw sample images, and position information and flaw characteristic parameters corresponding to each flaw sample image;
step S420, marking the severity level of a plurality of flaw sample images according to the corresponding position information and flaw characteristic parameters to form a plurality of flaw processing images;
in step S430, the initial first deep learning model is trained for a plurality of iterations based on the plurality of flaw processing images.
In this embodiment, a plurality of sample lead frames are provided, and a plurality of flaw sample images, and positional information and flaw characteristic parameters corresponding to each flaw sample image are collected based on the plurality of sample lead frames.
In this embodiment, the method for collecting the plurality of flaw sample images based on the plurality of sample lead frames may refer to the method for obtaining the first flaw image based on the first lead frame in steps S100 to S300, and will not be described herein.
Accordingly, based on the plurality of sample lead frames, the method for collecting the position information and the defect characteristic parameter corresponding to each defective sample image may refer to the method from step S100 to step S300 for obtaining the deviation position information and the deviation characteristic parameter corresponding to the deviation image corresponding to the first defective image, which are not described herein.
In other embodiments, the deviation position information and the deviation feature parameter may be obtained after data processing.
In some practical applications, the sample lead frame is a sheet-like lead frame in a trial-production stage, and the first lead frame is a sheet-like lead frame in a mass-production stage. Therefore, the initial first deep learning model is iteratively trained through the sheet-shaped lead frames in the trial production stage, and the first deep learning model used when the shipment grade of the sheet-shaped lead frames to be shipped currently is judged in the mass production stage is formed, so that the highly-automatic shipment judgment of the sheet-shaped lead frames in the mass production stage is realized.
In this embodiment, for step S420, the method for marking the severity level of the plurality of flaw sample images according to the corresponding position information and flaw feature parameters to form a plurality of flaw processed images includes:
step S421, determining the severity level corresponding to each flaw sample image according to the corresponding position information and flaw characteristic parameters;
step S422, providing a color or a filling pattern corresponding to each severity level;
step S423, marking the severity level of each flaw sample image according to the color or the filling pattern corresponding to the severity level of each flaw sample image.
In the present embodiment, the initial first deep learning model in step S430 includes: CNN neural network convolution model.
In other embodiments, the initial first deep learning model may also be a ViT (Vision Transformer) model.
Referring to fig. 7, for step S430, the method for the nth time of the multiple iterative training includes:
step S431, based on the nth of the plurality of flaw processing images, performing the nth iterative training on the first deep learning model formed by the (n-1) th historical iterative training;
step S432, evaluating a first deep learning model formed by the nth iterative training to obtain an nth evaluation result;
when the nth evaluation result meets a preset condition, executing a step S433, and stopping the iterative training;
and when the n-th evaluation result does not meet the preset condition, executing step S434, and performing the n+1th time in multiple iterative training.
The n is a natural number, and when n=1, the first deep learning model formed by the n-1 th historical iterative training is the initial first deep learning model.
Specifically, when the training is iterated, the first deep learning model formed in the previous iteration training is continuously trained through the unused flaw processing image until the evaluation result obtained when the formed first deep learning model is evaluated meets the preset condition, so as to obtain the first deep learning model used in the step S400.
It should be understood that the number of flaw processed images varies according to the number of iterative training.
In this embodiment, the nth evaluation result is: the prediction accuracy of the first deep learning model; the preset conditions are as follows: the prediction accuracy of the first deep learning model meets the preset prediction accuracy.
Specifically, in this embodiment, the first deep learning model is evaluated by the test set and the verifier, and whether the nth evaluation result satisfies the preset condition is determined.
In other embodiments, the nth evaluation result is: the number of iterative exercises performed to form the first deep learning model. And, the preset condition is: the number of iterative training performed to form the first deep learning model reaches a preset number. Namely: when the number of iterative training performed to form the first deep learning model reaches a preset number, executing step S433; otherwise, step S434 is performed.
In some practical applications, after the first deep learning model is formed, the first deep learning model is encapsulated, and the encapsulated first deep learning model is uploaded to a server. On this basis, the system of the shipment method of the lead frame used acquires the first deep learning model from a server to execute step S400.
With continued reference to fig. 1, in this embodiment, the shipment level includes: good and bad grades. Wherein, the good sheet grade refers to: the current sheet-shaped lead frame meets the shipment standard; the bad chip grade refers to: current sheet-like lead frames do not meet shipment standards.
Referring to fig. 8, fig. 8 is a flowchart of a method for determining a shipment level of a first lead frame according to an embodiment of the present invention, for step S500, the method for determining the shipment level of the first lead frame according to severity levels of a plurality of first flaws includes:
step S510, providing a preset shipment rule, wherein the preset shipment rule comprises a corresponding relation between shipment level and severity level of flaws;
and step S520, judging a plurality of first flaws and the severity level of the first flaws according to the preset shipment rule, and obtaining the shipment level of the first lead frame.
Specifically, step S520 refers to: comparing and judging the judged first flaws and the acquired severity levels of the first flaws with corresponding relations in preset shipment rules to obtain corresponding shipment levels.
In another embodiment, referring to fig. 9, fig. 9 is a flowchart illustrating a method for determining a shipment level of a first lead frame according to another embodiment of the present invention, the method for determining the shipment level of the first lead frame includes:
step S530, providing a preset shipment rule, wherein the preset shipment rule comprises shipment grade, severity grade of flaws and corresponding relation among users;
step S540, providing clients corresponding to the first lead frames;
step S550, determining a plurality of first flaws, severity levels of the plurality of first flaws, and customers corresponding to the first lead frames according to the preset shipment rule, and obtaining the shipment level of the first lead frames.
Specifically, step S550 refers to: and comparing and judging the judged first flaws, the acquired severity levels of the first flaws and clients corresponding to the first lead frames with corresponding relations in preset shipment rules to obtain corresponding shipment levels.
In some practical application scenarios, the shipment standards of different customers are different, that is, the requirements of different customers on the correspondence between the severity level of the flaw and the shipment level are different. For example, of the preset 9 severity levels, customer a requires rejecting (not conforming to shipment standards) a first lead frame having a first flaw of 7 or more (severity level), while customer B requires rejecting (not conforming to shipment standards) a first lead frame having a first flaw of 8 or more (severity level). Therefore, through steps S530 to S550, the corresponding shipment level can be determined based on different requirements of different customers. Therefore, in the actual mass production process, the goods can be delivered in a targeted manner according to the requirements of different clients conveniently and rapidly, and further, the management and labor cost is reduced better, and the quality inspection efficiency is improved.
With continued reference to fig. 1, in this embodiment, the method for shipping a lead frame further includes:
when the first defect is not detected, executing step S600, and determining that the shipment level of the first lead frame is a good chip level; the method comprises the steps of,
when the shipment level of the first lead frame is judged to be a good sheet level, executing step S710 to receive the first lead frame;
when the shipment level of the first lead frame is determined to be the bad chip level, step S720 is executed, where the receiving of the first lead frame is refused.
In still other embodiments, the shipment level comprises: good grade, reworked grade, and bad grade. And, the good slice grade means: current sheet-like lead frames meet shipment standards. The reworking grade refers to: current sheet lead frames do not meet shipment standards, but can be reworked to meet shipment standards. Therefore, the shipment quantity of the sheet-shaped lead frame is further increased, and the manufacturing cost is saved. Correspondingly, the bad chip grade refers to: current sheet-like lead frames do not meet shipment standards and are not reworked.
Meanwhile, the shipment method of the lead frame further comprises the following steps: when the shipment level of the first lead frame is judged to be the rework level, step S730 is executed to rework the first lead frame.
With continued reference to fig. 1, in this embodiment, the method for shipping the lead frame further includes:
step S800, evaluating whether the severity level of at least 1 first flaws predicted by the first deep learning model is wrong;
when the error is predicted, step S810 is executed to label the severity level of the first flaw image corresponding to the first flaw, so as to form a first flaw processing image;
step S820, based on the plurality of first flaw processed images, trains a first deep learning model for processing the plurality of first flaw images.
Therefore, the first deep learning model is continuously trained by collecting the first flaw image corresponding to the incorrect prediction result, so that the first deep learning model is further improved, the prediction accuracy of the first deep learning model is further improved, the sheet-shaped lead frame which is not subjected to shipment level judgment is subjected to more accurate shipment level judgment in the follow-up process, and further, the risk of incorrect shipment of the sheet-shaped lead frame is effectively reduced.
In the present embodiment, in step S800, it is evaluated whether the severity level of each first flaw predicted by the first deep learning model is wrong.
In some other embodiments, it may also be evaluated whether the severity level of 1 or part of the first deep learning model prediction is erroneous. For example, the evaluation may be performed by means of spot check.
In the present embodiment, the method of executing step S810 refers to the method of executing step S420 (shown in fig. 6), and will not be described herein.
In the present embodiment, please refer to the method of executing the step S430 (shown in fig. 6) for executing the step S820, which is not described herein.
Accordingly, in some practical applications, after the step S820 is performed to perfect the first deep learning model, the perfect first deep learning model is encapsulated, and the encapsulated first deep learning model is uploaded to a server to update the existing first deep learning model.
Although the present invention is disclosed above, the present invention is not limited thereto. Various changes and modifications may be made by one skilled in the art without departing from the spirit and scope of the invention, and the scope of the invention should be assessed accordingly to that of the appended claims.
Claims (14)
1. A method of shipping a lead frame, comprising:
acquiring a source image of the first lead frame according to the first lead frame;
detecting a first flaw according to the source image of the first lead frame, wherein the first flaw is a flaw of the first lead frame;
when a plurality of first flaws are detected, a plurality of corresponding first flaw images are obtained;
processing a plurality of first flaw images by adopting a first deep learning model to obtain the severity level of each first flaw;
and judging the shipment grade of the first lead frame according to the severity grade of the first defects.
2. The method of shipping a lead frame of claim 1, wherein the method of detecting a first flaw from a source image of the first lead frame comprises:
providing a standard template image;
providing a preset deviation range corresponding to the standard template image;
and detecting a first flaw according to the standard template image, the preset deviation range and the source image of the first lead frame.
3. The method of shipping a lead frame of claim 2, wherein the standard template pattern includes a plurality of standard areas, and wherein the predetermined deviation range includes sub-predetermined deviation ranges corresponding to the standard areas.
4. The method of shipping a lead frame of claim 3, wherein the method of detecting a first flaw from the standard template image, the preset deviation range, and the source image of the first lead frame comprises:
acquiring a plurality of deviation images and a plurality of corresponding deviation feature data according to the source image of the first lead frame and the standard template image, wherein the deviation feature data comprises deviation position information and deviation feature parameters, and each deviation position information also corresponds to one of a plurality of standard areas; detecting that the first flaw corresponding to the random deviation image exists when the deviation characteristic parameter corresponding to the random deviation image exceeds a designated sub-preset deviation range corresponding to the random deviation image, wherein the designated sub-preset deviation range corresponding to the random deviation image is a sub-preset deviation range of a standard area corresponding to deviation position information corresponding to the random deviation image;
and when a plurality of first flaws are detected, the method for acquiring the corresponding first flaw images comprises the following steps: and acquiring a first flaw image according to the deviation image corresponding to the first flaw.
5. The method of claim 3, wherein the types of standard regions include at least one of plating and critical regions, half-etched regions, functional regions, nonfunctional regions, rail and pilot hole regions, and low threshold regions.
6. The method of shipment of lead frames of claim 3, wherein the deviation feature parameter comprises at least one of an area parameter, a diagonal length parameter, and a contrast parameter corresponding to a deviation image.
7. The method of shipping a lead frame of claim 1, wherein the method of forming the first deep learning model comprises:
collecting a plurality of flaw sample images, and position information and flaw characteristic parameters corresponding to each flaw sample image;
marking the severity level of the plurality of flaw sample images according to the corresponding position information and flaw characteristic parameters to form a plurality of flaw processing images;
the initial first deep learning model is trained a plurality of iterations based on the plurality of flaw processing images.
8. The method of shipping a lead frame of claim 7, wherein the nth of the plurality of iterative training methods comprises:
performing an nth iterative training on a first deep learning model formed by the historical nth-1 iterative training based on an nth one of the plurality of flaw processing images, wherein n is a natural number, and when n=1, the first deep learning model formed by the historical nth-1 iterative training is the initial first deep learning model;
evaluating a first deep learning model formed by the nth iterative training to obtain an nth evaluation result;
and stopping the iterative training when the nth evaluation result meets a preset condition.
9. The method of shipping a lead frame of claim 8, further comprising:
evaluating whether the severity level of the at least 1 first blemish predicted by the first deep learning model is incorrect; when the error is predicted, marking the severity level of the first flaw image corresponding to the first flaw to form a first flaw processing image;
based on the first flaw processed images, a first deep learning model is trained that processes the first flaw images.
10. The method of claim 1, wherein determining the shipment level of the first leadframe according to the severity level of the plurality of first flaws comprises:
providing a preset shipment rule, wherein the preset shipment rule comprises a corresponding relation between shipment grade and the severity grade of the flaw;
and judging a plurality of first flaws and the severity level of the first flaws according to the preset shipment rule, and obtaining the shipment level of the first lead frame.
11. The method of claim 1, wherein determining the shipment level of the first leadframe according to the severity level of the plurality of first flaws comprises:
providing a preset shipment rule, wherein the preset shipment rule comprises shipment grade, severity grade of flaws and corresponding relation among clients;
providing a customer corresponding to the first lead frame;
and judging a plurality of first flaws, the severity levels of the first flaws and customers corresponding to the first lead frames according to the preset shipment rules, and obtaining the shipment level of the first lead frames.
12. The method of shipping a lead frame of claim 1, wherein the shipping levels include a good level and a bad level, and wherein the method of shipping a lead frame further comprises:
receiving the first lead frame when the shipment level of the first lead frame is judged to be a good sheet level;
and refusing to receive the first lead frame when the shipment grade of the first lead frame is judged to be the bad chip grade.
13. The method of shipping a lead frame of claim 12, wherein the shipping level further comprises a rework level, and wherein the method of shipping a lead frame further comprises: and when the shipment grade of the first lead frame is judged to be the reworking grade, reworking the first lead frame.
14. The method of shipping a lead frame of claim 12, further comprising: and when the first flaw is not detected, judging that the shipment grade of the first lead frame is a good grade.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111336780.3A CN116128787A (en) | 2021-11-12 | 2021-11-12 | Lead frame shipment method |
| TW111143268A TWI856420B (en) | 2021-11-12 | 2022-11-11 | Lead frame delivery method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111336780.3A CN116128787A (en) | 2021-11-12 | 2021-11-12 | Lead frame shipment method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116128787A true CN116128787A (en) | 2023-05-16 |
Family
ID=86303192
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111336780.3A Pending CN116128787A (en) | 2021-11-12 | 2021-11-12 | Lead frame shipment method |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN116128787A (en) |
| TW (1) | TWI856420B (en) |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5894659A (en) * | 1996-03-18 | 1999-04-20 | Motorola, Inc. | Method for inspecting lead frames in a tape lead bonding system |
| JP2007205974A (en) * | 2006-02-03 | 2007-08-16 | Toppan Printing Co Ltd | Method of inspecting plating, and method of inspecting lead frame |
| US10776912B2 (en) * | 2016-03-09 | 2020-09-15 | Agency For Science, Technology And Research | Self-determining inspection method for automated optical wire bond inspection |
| CN113177934A (en) * | 2021-05-20 | 2021-07-27 | 聚时科技(上海)有限公司 | Lead frame defect positioning and grade judging method based on deep learning |
| CN113592866A (en) * | 2021-09-29 | 2021-11-02 | 西安邮电大学 | Semiconductor lead frame exposure defect detection method |
-
2021
- 2021-11-12 CN CN202111336780.3A patent/CN116128787A/en active Pending
-
2022
- 2022-11-11 TW TW111143268A patent/TWI856420B/en active
Also Published As
| Publication number | Publication date |
|---|---|
| TW202321682A (en) | 2023-06-01 |
| TWI856420B (en) | 2024-09-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102258942B1 (en) | System and method for the automatic determination of critical parametric electrical test parameters for inline yield monitoring | |
| US6766208B2 (en) | Automatic production quality control method and system | |
| JP5168215B2 (en) | Appearance inspection device | |
| CN115953373B (en) | Glass defect detection method, device, electronic equipment and storage medium | |
| CN115360116B (en) | Wafer defect detection method and system | |
| CN114299040B (en) | Ceramic tile flaw detection method and device and electronic equipment | |
| CN111275700A (en) | Terminal defect detection method and system based on deep learning | |
| CN115690670A (en) | Intelligent identification method and system for wafer defects | |
| CN117589770A (en) | PCB patch board detection method, device, equipment and medium | |
| CN110930350A (en) | Machine learning method and automatic optical detection equipment applying same | |
| CN115830403A (en) | Automatic defect classification system and method based on deep learning | |
| US6828776B2 (en) | Method for analyzing defect inspection parameters | |
| CN116128787A (en) | Lead frame shipment method | |
| CN117455887A (en) | PCB incoming material detection method, device, equipment and medium based on machine vision | |
| CN116128788A (en) | Lead frame shipment method | |
| CN113610801B (en) | Defect classification method, device, equipment and storage medium based on minimum unit | |
| CN116894956A (en) | Feature quantity selection method and device | |
| CN116339262A (en) | Numerical control processing production quality monitoring system based on artificial intelligence | |
| CN115266744A (en) | Detection system and method for product in production line | |
| CN114911085A (en) | Methods of Analyzing Defects | |
| US20240404040A1 (en) | Test system and test method to wafers | |
| CN118096747B (en) | Automatic PCBA (printed circuit board assembly) board detection method and system based on deep learning | |
| de Souza Silva et al. | Smart Machine Vision System to Improve Decision-Maker on the Assembly Line | |
| CN119379701A (en) | A particleboard quality detection method and system based on image recognition | |
| CN115222670A (en) | Method for automatically calculating number and grade of pixels affected by defects |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |