A method, system and device for detecting packaging defects of microelectronic glass insulating terminals
By sector division and feature extraction of microelectronic glass insulated terminal images, the problem of low detection accuracy in the prior art is solved, and higher detection accuracy and better characterization of local differences in glass packaging are achieved.
Patent Information
- Application Number
- CN202210086914.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-01-25
- Publication Date
- 2025-07-01
- Estimated Expiration
- 2042-01-25
AI Technical Summary
The existing glass packaging detection methods have poor adaptability to complex image backgrounds and local differences, resulting in low detection accuracy, especially in accurate positioning of glass shedding areas.
By preprocessing the image of the insulated terminal of the microelectronic glass to be measured, it is divided into several sectors, and the internal and neighborhood features of the sector are extracted, and then spliced and input into the classifier to output the defect position detection result.
The detection accuracy of microelectronic glass insulated terminal packaging defects is improved, the characterization ability of local differences in glass packaging is enhanced, and the subjectivity and missed detection rate are reduced.
Smart Images
Figure CN114581373B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of electronic packaging detection, and more specifically, to a method, system and device for detecting packaging defects of microelectronic glass insulating terminals. Background Art
[0002] Packaging is an important part of electronic components, which can protect electronic components from external intrusion and ensure the safe operation of electronic devices. Electronic component packaging mainly uses materials such as metals, glasses, polymers and ceramics to achieve different protection functions. In the actual production process, non-uniform thermal stress may cause the glass packaging to fall off and crack, allowing humid air to enter the component interior, reducing the dielectric performance of the microelectronic glass insulating terminal, resulting in problems such as leakage, breakdown and short circuit, causing immeasurable damage to electronic devices. Therefore, detecting packaging defects of microelectronic glass insulating terminals is an essential step in actual production. Currently, in the industry, the quality of the glass packaging of insulating terminals is mainly detected by visual observation with the aid of a microscope. This detection is time-consuming and laborious, highly subjective and extremely prone to missed detections. Therefore, the automatic detection of packaging defects of microelectronic glass insulating terminals is an urgent problem to be solved in the field of electronic packaging manufacturing.
[0003] There is a method for detecting glass surface defects, which acquires a first glass surface image and a second glass surface image of the glass to be detected, performs mean filtering on the first glass surface image and the second glass surface image respectively using a first filtering template to obtain a first filtered image and a second filtered image; then extracts the difference regions between the first filtered image and the second filtered image, and the difference regions correspond one-to-one to the surface defects of the glass to be detected, thereby obtaining the surface defects of the glass to be detected.
[0004] However, when the above method detects an image, it uses traditional global mean filtering to process the image, which has poor adaptability to complex image backgrounds and local differences. In particular, the characteristic of global smoothing will increase the difficulty of accurately locating the glass detachment area with blurred boundaries, reducing the detection accuracy of glass surface defects. Summary of the Invention
[0005] In order to overcome the defects of the existing glass packaging detection method, such as poor adaptability to complex image backgrounds and local differences and low detection accuracy, the present invention provides a method, system and device for detecting packaging defects of microelectronic glass insulating terminals.
[0006] To solve the above technical problems, the technical solution of the present invention is as follows:
[0007] In a first aspect, the present invention proposes a method for detecting packaging defects of microelectronic glass insulating terminals, including the following steps:
[0008] S1: Preprocess the image of the microelectronic glass insulation terminal to be measured, obtain the area to be detected, and divide the area to be detected into several sectors;
[0009] S2: Extract the internal features of the sectors and the neighborhood features of the sectors;
[0010] S3: Concatenate the internal features of the sectors and the neighborhood features of the sectors, input the concatenation result into the classifier, and the classifier outputs the detection result of the defect position of the microelectronic glass insulation terminal package.
[0011] Preferably, in S1, the specific steps for obtaining the area to be detected include:
[0012] Select the blue threshold for the image of the microelectronic glass insulation terminal to obtain the initial glass package area; perform opening operation and closing operation on the initial glass package area in sequence; in the glass package area after the opening operation and closing operation, select the largest connected domain as the area to be detected.
[0013] Preferably, in S1, with the center of the area of the pin of the microelectronic glass insulation terminal as the center of the circle, divide the area to be detected into several sectors according to the longitude and latitude directions, specifically including the following steps:
[0014] In the longitude direction, divide the area to be detected into 2 m+2 sectors, where m represents the longitude parameter.
[0015] In the latitude direction, divide the area to be detected into c(n + 1) concentric circles; the expression of the latitude division coefficient c is:
[0016]
[0017] where n represents the longitude parameter, R represents the outer diameter of the area to be detected of the microelectronic glass insulation terminal, r represents the radius of the area occupied by the pin, is the floor operation.
[0018] After the division in the longitude and latitude directions, divide the area to be detected into c(n + 1)×2 m+2 sectors.
[0019] Preferably, in S2, the internal features of the sectors include the basic statistical features of the sectors, the gray-scale change rate of the sectors, the reflection features of the sectors, and the direction statistical features of the sectors.
[0020] Preferably, in S2, the specific steps for extracting the neighborhood features of the sectors include:
[0021] Obtain the centroid of the sector SEC j,k with longitude j and latitude k and several of its neighborhood sectors, and with the centroid as the center, generate a side length of A number of square blocks;
[0022] Input the square blocks into a classification model for classification to obtain the classification results of the sectors, and form a classification table of the current sector with the classification results of the sectors.
[0023] According to the classification table of the current sector, starting from the neighborhood sector at the upper left corner of sector SEC j,k Arrange the classification results of the neighborhood sectors in clockwise order to form the neighborhood feature SN of sector SEC j,k j,k .
[0024] Preferably, input the square blocks into a Resnet50 network classification model for classification to obtain the classification results of the sectors, and form a classification table of the current sector with the classification results of the sectors.
[0025] Preferably, in S3, splice the internal feature of the sector and the neighborhood feature SN of the sector j,k to obtain the fine classification feature F of the sector j,k , input the fine classification feature F of the sector j,k into a GBDT classifier, and the GBDT classifier outputs the defect position of the microelectronic glass insulation terminal package; wherein, the expression of the fine classification feature F of the sector j，k is as follows:
[0026]
[0027] Wherein, represents the basic statistical feature of the sector, represents the gray change rate of the sector, represents the reflection feature of the sector, represents the direction statistical feature of the sector.
[0028] In a second aspect, the present invention proposes a microelectronic glass insulation terminal package defect detection system, including:
[0029] A preprocessing module for preprocessing the image of the microelectronic glass insulation terminal to obtain the area to be detected.
[0030] A feature extraction module for dividing the area to be detected into a number of sectors and extracting the internal feature of the sector and the neighborhood feature of the sector.
[0031] A detection module, including a classification unit, for splicing the internal feature of the sector and the neighborhood feature of the sector, inputting the splicing result into the classification unit, and the classification unit outputs the detection result of the defect position of the microelectronic glass insulation terminal package.
[0032] Preferably, the feature extraction module includes a sector basic statistical feature extraction unit, a sector gray-scale change rate extraction unit, a sector reflection feature extraction unit, a sector direction statistical feature extraction unit, and a sector neighborhood feature extraction unit.
[0033] In a third aspect, the present invention provides a microelectronic glass insulation terminal packaging defect detection device, including a memory and a processor. A computer program is stored on the memory, and when the processor executes the computer program in the memory, the steps of any one of the microelectronic glass insulation terminal packaging defect detection methods are implemented.
[0034] Compared with the prior art, the beneficial effects of the technical solution of the present invention are as follows: The present invention divides the area to be detected in the microelectronic glass insulation terminal packaging into several sectors. The partition processing can better characterize the local differences of the glass packaging compared with the method of processing the image using global mean filtering. In addition, by extracting features from the divided sectors and their neighborhood information, multi-dimensional features of the sectors are obtained, which can better characterize the normal areas and defect areas of the sectors, increase the contrast between the two, and further improve the detection accuracy of microelectronic glass insulation terminal packaging defects. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 is a flowchart of the microelectronic glass insulation terminal packaging defect detection method.
[0036] Figure 2 is an imaging diagram of the microelectronic glass insulation terminal.
[0037] Figure 3 is a flowchart for extracting the SN feature.
[0038] Figure 4 is an imaging diagram of the true area and the detected area of the microelectronic glass insulation terminal packaging defect area.
[0039] Figure 5 is an architecture diagram of the microelectronic glass insulation terminal packaging defect detection system.
[0040] Figure 6 is a detection flowchart of the microelectronic glass insulation terminal packaging defect detection system. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0041] The drawings are only for illustrative purposes and should not be construed as limitations on this patent;
[0042] The technical solution of the present invention will be further described below with reference to the drawings and embodiments.
[0043] Embodiment 1
[0044] Please refer to Figure 1, this embodiment proposes a method for detecting packaging defects of microelectronic glass insulating terminals, including the following steps:
[0045] S1: Preprocess the image of the microelectronic glass insulating terminal to be tested, obtain the area to be detected, and divide the area to be detected into several sectors;
[0046] S2: Extract the internal features of the sector and the neighborhood features of the sector;
[0047] S3: Concatenate the internal features of the sector and the neighborhood features of the sector, input the concatenation result into the classifier, and the classifier outputs the detection result of the defect location of the microelectronic glass insulating terminal packaging.
[0048] In this embodiment, the internal features of the sector include the basic statistical features of the sector, the gray change rate of the sector, the reflection feature of the sector, and the direction statistical feature of the sector.
[0049] In the packaging defects of microelectronic glass insulating terminals, there is a situation where the pixel contrast between the peeling area and the non-peeling area is low. In response to this situation, the present invention performs multi-dimensional feature extraction according to the characteristics of each sector, including the basic statistical features of the sector, the gray change rate of the sector, the reflection feature of the sector, and the direction statistical feature of the sector, in order to better characterize the normal area and the defect area of the sector, increase the contrast between the two, and improve the defect detection accuracy. Among them, the gray change rate of the sector characterizes the information of the sector in its neighborhood and the information in its area to be detected; the reflection feature of the sector reduces the interference of reflection on the surface of the insulating terminal; the direction statistical feature of the sector reveals the potential defect information inside the sector.
[0050] The present invention divides the area to be detected of the microelectronic glass insulating terminal packaging into several sectors. The partition processing can better characterize the local differences of the glass packaging compared with the method of processing the image using global mean filtering. In addition, extracting features from the divided sectors and their neighborhood information to obtain multi-dimensional features of the sectors can better characterize the normal area and the defect area of the sectors, increase the contrast between the two, and further improve the detection accuracy of the packaging defects of microelectronic glass insulating terminals. In addition, in response to the insufficient neighborhood information of the sector, neighborhood feature extraction is proposed. The neighborhood feature characterizes the neighborhood information of the sector, effectively improves the accuracy of the classifier, and accurately locates the defects of the microelectronic glass insulating terminal.
[0051] Embodiment 2
[0052] This embodiment proposes a method for detecting packaging defects of microelectronic glass insulating terminals, including the following steps:
[0053] S1: Preprocess the image of the microelectronic glass insulating terminal to be tested, obtain the area to be detected, and divide the area to be detected into several sectors.
[0054] The microelectronic glass insulating terminal is composed of a pin, a glass package, and a peripheral metal package. Inside the glass package is a blue insulating material. In this embodiment, a blue threshold is selected for the image of the microelectronic glass insulating terminal to obtain an initial glass package area; then an opening operation and a closing operation are sequentially performed on the initial glass package area. In the glass package area after the opening operation and the closing operation, the largest connected region is selected as the region to be detected, and the pin region and the peripheral metal package region in the image of the microelectronic glass insulating terminal are removed to obtain the region to be detected. As Figure 2 shown Figure 2 is an imaging diagram of the microelectronic glass insulating terminal.
[0055] According to the fact that the region to be detected presents a ring shape, a sector partitioning strategy is proposed. In this embodiment, with the center of the region of the pin of the microelectronic glass insulating terminal as the center of the circle, the region to be detected is divided into several sectors in the direction of the longitude and latitude lines. The specific steps are as follows:
[0056] In the longitude direction, the region to be detected is divided into 2 m+2 sectors, and m = 1, 2, 3 represents the longitude parameter.
[0057] In the latitude direction, the region to be detected is divided into c(n + 1) concentric circles, and n = 1, 2, 3, 4 represents the latitude parameter; in order to adapt to microelectronic glass insulating terminals of different specifications and models, the expression of the latitude division coefficient c is defined as:
[0058]
[0059] where R represents the outer diameter of the region to be detected of the microelectronic glass insulating terminal, r represents the radius of the region occupied by the pin, is the floor operation.
[0060] Obviously, the circular regions corresponding to the first n + 1 concentric circles are the regions occupied by the pins. Therefore, after the division in the longitude and latitude directions, the region to be detected is divided into c(n + 1)×2 m+2 sectors.
[0061] S2: Extract the internal features of the sectors and the neighborhood features of the sectors.
[0062] In this embodiment, the internal features of the sectors include the basic statistical features of the sectors, the gray change rate of the sectors, the reflection features of the sectors, and the direction statistical features of the sectors.
[0063] The basic statistical features of the sectors are expressed as follows:
[0064]
[0065] where Indicates the sector SEC with longitude j and latitude k j,k The average Red channel value of, Indicates the sector SEC j，k The average Green channel value of, Indicates the sector SEC j,k The average Blue channel value of, Indicates the sector SEC j,k The average Hue channel value of, Indicates the sector SEC j,k The average Saturation channel value of, Indicates the sector SEC j,k The average Value channel value of; Std j,k Indicates the sector SEC j,k The standard deviation of grayscale within.
[0066] The glass encapsulation of the microelectronic glass insulating terminal falling off or breaking will cause a large difference in the grayscale of the sector. Therefore, in this embodiment, the sector is used as the reference unit to define the local grayscale change rate and the global grayscale change rate of the sector, thereby constituting a class of features of the sector grayscale change rate.
[0067] Let the average grayscale of the sector SEC j,k be The average grayscales of the sectors on the left, right, and inner sides of the nearest neighborhood of the sector SEC j,k are respectively Then the local grayscale change rate of the left sector in the neighborhood The local grayscale change rate of the right sector in the neighborhood The local grayscale change rate of the inner sector in the neighborhood and the global grayscale change rate of the sector The expressions are as follows:
[0068]
[0069]
[0070]
[0071] From the local grayscale change rate of the sector, it can be seen that if the local grayscale change rate of the sector in a certain direction is large, there may be potential defects in the sector in that direction. At this time, the global grayscale change rate of the sector is defined as:
[0072]
[0073] where G ROI is the average grayscale of the area to be detected. From the global grayscale change rate of the sector it can be seen that when is larger, the sector SEC j,k may be a reflective area, that is, there is a glass package in this area; when is smaller, there may be potential defects in this sector.
[0074] Using the local gray change rate of the left sector of the neighborhood and the local gray change rate of the right sector of the neighborhood and the local gray change rate of the inner sector of the neighborhood and the global gray change rate of the sector to construct the sector gray change rate Its expression is as follows:
[0075]
[0076] Sector gray change rate characterizes the gray relationship between the sector and its neighborhood and the area to be detected.
[0077] Because the glass package of the microelectronic glass insulating terminal is very likely to produce a reflective phenomenon, so if the glass package in the area to be detected falls off, the reflective phenomenon will be weakened. This characteristic can be used to judge whether the reflective glass package area has fallen off. Therefore, define the white occupancy rate WOR j,k and flatness CER j,k and characterize this phenomenon from two dimensions of color and texture.
[0078] Define the white occupancy rate WOR j,k as:
[0079]
[0080] represents the blue sector.
[0081] The above formula shows that when the white occupancy rate WOR j,k is larger, the sector SEC j,k has the possibility of being a reflective area.
[0082] Define the sector flatness CER j,k as:
[0083]
[0084]
[0085] where Canny(x, y) is the Canny operator operation. When the sector flatness CER j,k is larger, the sector SEC j,k has the possibility of being a reflective area.
[0086] Using the sector white occupation ratio WOR j,k and the sector white occupation ratio WOR j,k Construct the sector reflection feature Its expression is as follows:
[0087]
[0088] When the glass encapsulation of the microelectronic glass insulating terminal falls off, there may be an obvious edge in the falling-off area, that is, the gray-scale gradient in a certain direction may be large. Therefore, let G(x, y) be the gray scale of the pixel point (x, y) in the sector, then the vertical direction gradient D y (x, y) and the horizontal direction gradient D x (x, y) are respectively:
[0089] D y (x, y) = G(x, y + 1) - G(x, y - 1)
[0090] D x (x, y) = G(x + 1, y) - G(x - 1, y)
[0091] The gradient direction θ(x, y) is:
[0092]
[0093] Calculate the 9-direction statistical features of the sector SEC j,k equally spaced (z = 1, 2,..., 9) are:
[0094]
[0095]
[0096] Indicates sectors in different directions; Dz represents the direction, (z = 1, 2,..., 9), indicating 9 different directions.
[0097] Using the 9-direction statistical features Construct a 9-dimensional sector direction statistical feature Its expression is as follows:
[0098]
[0099] When the differences between all the direction statistical features in
[0100] Although the internal features of the above sectors have utilized some information of the sectors and their neighboring pixels for feature extraction, the information of the sector neighborhood (SN), i.e., the correlation between sectors, has not been fully utilized. Therefore, in this embodiment, the SN features of each sector are constructed based on Resnet to comprehensively represent the sector neighborhood information.
[0101] As Figure 3 shown, Figure 3 the flowchart for extracting SN features is as follows, specifically including the following steps:
[0102] Obtain the sector SEC with longitude j and latitude k j,k and the centroids of its 8 neighboring sectors, and generate 8 neighboring square blocks with side length centered at the centroids. In this embodiment, a total of 9 square blocks are generated, including the sector SEC j,k square block and the 8 neighboring square blocks of the sector SEC j,k .
[0103] Input the sector SEC j,k square block and the 8 neighboring square blocks of the sector SEC j,k into the Resnet50 network classification model for classification, obtain the classification results of the corresponding 9 sectors, and form the classification table of the current sector with the classification results of the sectors;
[0104] According to the classification table of the current sector, starting from the upper left neighborhood of the sector SEC j,k , arrange the classification results of the neighboring sectors in a clockwise order to form the neighborhood feature SN j,k of the sector SEC j,k .
[0105] S3: Concatenate the internal features of the sector and the neighborhood features of the sector, input the concatenated result into the classifier, and the classifier outputs the defect location detection result of the microelectronic glass insulation terminal package.
[0106] In this embodiment, the internal features of the sector including the basic statistical features gray change rate of the sector reflective feature of the sector and the directional statistical features of the sector are concatenated with the neighborhood feature SN j,k of the sector to obtain the fine classification feature F j,k of the sector. Input the fine classification feature F j,k of the sector into the GBDT classifier, and the GBDT classifier outputs the defect location of the microelectronic glass insulation terminal package; among them, the fine classification feature F j,kThe expression is as follows:
[0107]
[0108] The present invention divides the area to be detected in the microelectronic glass insulated terminal package into several sectors. Compared with the method of processing the image using global mean filtering, the sector processing can better characterize the local differences in the glass package, especially for the images of microelectronic glass insulated terminals with low differences between peeling defects and non-peeling defects. In addition, by extracting features from the divided sectors and their neighborhood information to obtain multi-dimensional features of the sectors, the normal areas and defect areas of the sectors can be better characterized, the contrast between the two can be increased, and the detection accuracy of the defects in the microelectronic glass insulated terminal package can be further improved. Dividing the detection area into sectors can also increase the amount of data. For the defect annotation of a microelectronic glass insulated terminal image sample, several annotations can be obtained under the sector division method, improving the annotation efficiency.
[0109] In the defects of microelectronic glass insulated terminal packages, there is a situation where the pixel contrast between the peeling area and the non-peeling area is low. In response to this situation, the present invention performs multi-dimensional feature extraction according to the characteristics of each sector, including sector basic statistical features, sector gray-scale change rate, sector reflection features, and sector direction statistical features, in order to better characterize the normal area and defect area of the sector, increase the contrast between the two, and improve the defect detection accuracy. Among them, the sector gray-scale change rate characterizes the information of the sector in its neighborhood and the information in its area to be detected; the sector reflection features reduce the interference of reflection on the surface of the insulated terminal; the sector direction statistical features reveal the potential defect information inside the sector.
[0110] In addition, in view of the insufficient neighborhood information of the sectors, neighborhood feature extraction is proposed. The neighborhood features characterize the neighborhood information of the sectors, effectively improving the accuracy of the classifier and precisely locating the defects in the microelectronic glass insulated terminal package, such as Figure 4 shown Figure 4 (a) is an imaging diagram of the true area of the defect in the microelectronic glass insulated terminal package, Figure 4 (b) is an imaging diagram of the detection area of the defect in the microelectronic glass insulated terminal package. As can be seen from Figure 4 , the defect area detected by the defect detection method for microelectronic glass insulated terminal packages proposed by the present invention is basically consistent with the true area of the defect, indicating that the present invention has a high accuracy in the detection of defects in microelectronic glass insulated terminal packages.
[0111] Embodiment 3
[0112] Please refer to Figures 5 - 6, this embodiment also proposes a microelectronic glass insulation terminal packaging defect detection system, which is applied to the microelectronic glass insulation terminal packaging defect detection method proposed in the above embodiment, and includes: a preprocessing module, a feature extraction module, and a detection module. Among them, the feature extraction module includes a sector basic statistical feature extraction unit, a sector gray change rate extraction unit, a sector reflection feature extraction unit, a sector direction statistical feature extraction unit, and a sector neighborhood feature extraction unit.
[0113] As Figure 6 shown, Figure 6 FIG. is a detection flow chart of the microelectronic glass insulation terminal packaging defect detection system. In the specific implementation process, the preprocessing module first preprocesses the image of the microelectronic glass insulation terminal, selects the blue threshold for the image of the microelectronic glass insulation terminal to obtain the initial glass packaging area; then performs opening and closing operations on the initial glass packaging area in sequence. In the glass packaging area after the opening and closing operations, select the largest connected domain as the area to be detected, remove the pin area and the peripheral metal packaging area in the microelectronic glass insulation terminal image, obtain the area to be detected, and then take the center of the area of the pins of the microelectronic glass insulation terminal as the center of the circle, and divide the area to be detected into several sectors in the direction of the longitude and latitude lines, and input the sectors into the feature extraction module.
[0114] The sector basic statistical feature extraction unit, the sector gray change rate extraction unit, the sector reflection feature extraction unit, and the sector direction statistical feature extraction unit in the feature extraction module respectively extract the internal features of the sector including the sector basic statistical features, the sector gray change rate, the sector reflection features, and the sector direction statistical features; the sector neighborhood feature extraction unit classifies according to the feature information of the sector and its neighboring sectors by using the Resnet50 network classification model to obtain the classification result of the corresponding sector, and forms the classification table of the current sector with the classification results of the sectors, and then arranges the classification results of the neighboring sectors according to the classification table of the current sector to form the neighborhood features of the sector, and inputs the internal features and the neighborhood features of the sector into the detection module.
[0115] The detection module splices the internal features and the neighborhood features of the sector to obtain the sector fine classification features. The classification unit in the detection module uses the GBDT classifier to output the detection result of the defect position of the microelectronic glass insulation terminal packaging according to the sector fine classification features.
[0116] The terms describing the positional relationship in the drawings are for illustrative purposes only and should not be construed as a limitation of this patent;
[0117] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, rather than limitations on the implementation manners of the present invention. For those of ordinary skill in the relevant art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to enumerate all implementation manners here. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the claims of the present invention.
Claims
1. A method for detecting packaging defects of a microelectronic glass insulating terminal, characterized in that, It includes the following steps: S1: Preprocess the image of the microelectronic glass insulating terminal to be measured, obtain the area to be detected, and divide the area to be detected into several sectors; S2: Extract the internal features of the sector and the neighborhood features of the sector; S3: Concatenate the internal features of the sector and the neighborhood features of the sector, input the concatenation result into the classifier, and the classifier outputs the detection result of the defect location of the microelectronic glass insulating terminal package.
2. The method for detecting packaging defects of the microelectronic glass insulating terminal according to claim 1, characterized in that, In S1, the specific steps for obtaining the area to be detected include: Select the blue threshold for the image of the microelectronic glass insulating terminal to obtain the initial glass package area; Perform opening operation and closing operation on the initial glass package area in sequence; In the glass package area after the opening operation and the closing operation, select the largest connected domain as the area to be detected.
3. The method for detecting packaging defects of the microelectronic glass insulating terminal according to claim 1, wherein In S1, with the center of the area of the pin of the microelectronic glass insulating terminal as the center of the circle, divide the area to be detected into several sectors in the direction of the longitude and latitude lines, and the specific steps include: In the warp direction, the area to be detected is divided into 2 m+2 sectors, where m represents the longitude parameter; In the latitude direction, divide the area to be detected into c(n + 1) concentric circles; the expression of the latitude division coefficient c is: Wherein, n represents the latitude parameter, R represents the outer diameter of the area to be detected of the microelectronic glass insulating terminal, and r represents the radius of the area occupied by the pin, is the floor operation; After the division in the longitude and latitude directions, the area to be detected is divided into c(n + 1)×2 m+2 sectors.
4. The method for detecting packaging defects of the microelectronic glass insulating terminal according to claim 1, characterized in that, In S2, the internal features of the sector include the basic statistical features of the sector, the gray change rate of the sector, the reflection feature of the sector, and the direction statistical feature of the sector.
5. The method for detecting packaging defects of the microelectronic glass insulating terminal according to claim 4, wherein, In S2, the specific steps for extracting the neighborhood features of the sector include: Obtain the centroid of the sector SEC with longitude j and latitude k j,k and the centroids of several adjacent sector areas, and generate several square blocks with side length centered at the centroid; Input the square block into the classification model for classification to obtain the classification result of the sector, and form the classification table of the current sector with the classification results of the sectors; According to the classification table of the current sector, starting from the neighborhood sector at the upper left corner of sector SEC j,k arrange the classification results of the neighborhood sectors in clockwise order to form the neighborhood feature SN j,k of sector SEC j,k .
6. The method for detecting packaging defects of a microelectronic glass insulating terminal according to claim 5, characterized in that, Input the square block into the Resnet50 network classification model for classification to obtain the classification result of the sector, and form the classification table of the current sector with the classification results of the sectors.
7. The method for detecting packaging defects of the microelectronic glass insulating terminal according to claim 6, wherein, In S3, the internal features of the sector and the neighborhood features SN of the sector j,k are spliced to obtain the fine classification feature F of the sector j,k , and the fine classification feature F of the sector j,k is input into the GBDT classifier, and the GBDT classifier outputs the defect location of the microelectronic glass insulation terminal package; among them, the fine classification feature F of the sector j,k is expressed as follows: Among them, represents the basic statistical features of the sector, represents the gray-scale change rate of the sector, represents the reflection feature of the sector, represents the directional statistical features of the sector.
8. A microelectronic glass insulation terminal packaging defect detection system, characterized in that It includes: A preprocessing module for preprocessing the image of the microelectronic glass insulating terminal, obtaining the area to be detected, and dividing the area to be detected into several sectors; A feature extraction module for extracting the internal features of the sector and the neighborhood features of the sector; A detection module, including a classification unit, for concatenating the internal features of the sector and the neighborhood features of the sector, inputting the concatenation result into the classification unit, and the classification unit outputs the detection result of the defect location of the microelectronic glass insulating terminal package.
9. The microelectronic glass insulation terminal package defect detection system according to claim 8, wherein The feature extraction module includes a basic statistical feature extraction unit of the sector, a gray change rate extraction unit of the sector, a reflection feature extraction unit of the sector, a direction statistical feature extraction unit of the sector, and a neighborhood feature extraction unit of the sector.
10. A microelectronic glass insulation terminal packaging defect detection device, characterized in that, It includes a memory and a processor. A computer program is stored on the memory. When the processor executes the computer program in the memory, the steps of the microelectronic glass insulating terminal package defect detection method according to any one of claims 1 to 7 are implemented.