Method, device and computer equipment for detecting semiconductor chips

Abstract

The application relates to the technical field of semiconductor chip detection, and discloses a semiconductor chip detection method, device and computer equipment, wherein the method comprises the following steps: acquiring an image of a semiconductor chip as an image to be analyzed; dividing the image to be analyzed into a multi-level image sequence; calculating the similarity between the multi-level image sequence and a preset template image to obtain a target image; comparing the target image with a standard image at a target detection position; when the similarity between the target image and the standard image is lower than a set value, determining that the target detection position has a defect; extracting a two-dimensional code of the semiconductor chip and updating information in the two-dimensional code of the semiconductor chip. The semiconductor chip detection method, device and computer equipment provided by the application can realize automatic detection, improve the automation level and degree, reduce misjudgment caused by human differences and errors, improve the reliability of semiconductor chip production, improve the utilization rate of machines and factories, and improve enterprise operating costs.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor chip detection, and particularly relates to a semiconductor chip detection method, device and computer equipment. BACKGROUND

[0002] At present, the demand for electronic chips is changing rapidly with the rapid development of the electronic information industry, and the chip packaging quality detection method and detection capacity are insufficient, which is a common problem in the industry. The application of powerful digital image processing technology in the production automation detection of semiconductor packaging detection instead of traditional manual detection is the most widely used and welcomed by enterprises, and is also the demand of actual productivity development. In the highly developed semiconductor packaging test industry, the automation degree of modern electronics and machinery is gradually improved, and the integration level of advanced high-tech semiconductor chips is also increasing. Relying on the limited resources of human eyes to operate not only has great work pressure, but also is prone to errors in the production and detection process due to eye fatigue and human differences, so relying on traditional manual quality inspection and detection cannot meet the urgent needs of high-tech industries. SUMMARY

[0003] The present application provides a semiconductor chip detection method, device and computer equipment, which realizes automatic detection to improve the automation level and degree, reduce the misjudgment caused by human differences and errors, and improve the reliability of semiconductor chip production, the utilization rate of machines and factories and the operating cost of enterprises.

[0004] The present application provides a semiconductor chip detection method, which comprises:

[0005] A CCD image sensor is used to obtain an image of a semiconductor chip as a to-be-analyzed image;

[0006] The to-be-analyzed image is divided into a multi-level image sequence according to a preset template image, and the multi-level image sequence is enlarged according to the preset template image;

[0007] A sequential correlation algorithm is used to calculate the similarity of the multi-level image sequence and the preset template image to obtain an image of a target detection position of the semiconductor chip as a target image;

[0008] The target image is compared with a standard image of the target detection position;

[0009] When the similarity of the target image and the standard image is lower than a set value, the target image is corrected, and when the similarity of the corrected target image and the standard image is lower than the set value, it is determined that the target detection position has a defect;

[0010] The CCD image sensor is used to extract the two-dimensional code of the semiconductor chip, and the target image is transmitted through a network to update information in the two-dimensional code of the semiconductor chip.

[0011] Further, before the step of using the CCD image sensor to acquire an image of the semiconductor chip as the image to be analyzed, the method further comprises:

[0012] A substrate of the semiconductor chip is identified, and a two-dimensional code for carrying data information of the semiconductor chip is printed on the substrate by using a laser;

[0013] The CCD image sensor is used to extract the two-dimensional code, and the data of the semiconductor chip acquired by the CCD image sensor is transmitted through a network;

[0014] An image of the two-dimensional code containing data of the semiconductor chip is acquired, the image of the two-dimensional code is positioned, separated, and decoded to obtain information in the two-dimensional code, and the information in the two-dimensional code is stored in a preset semiconductor chip database.

[0015] Further, the step of dividing the image to be analyzed into a multi-level image sequence according to a preset template image comprises:

[0016] A first-level image is formed by weighting and averaging pixel values of every x×y pixels of the template image as one pixel;

[0017] The first-level image is divided into a plurality of region images according to a preset pixel value;

[0018] The image to be analyzed, the first-level image, and the plurality of region images are combined to form the multi-level image sequence;

[0019] One region image in the multi-level image sequence is extracted in sequence to be enlarged to obtain a plurality of second-level region images;

[0020] A row magnification ratio and a column magnification ratio are obtained according to the preset template image and the second-level region image;

[0021] A row mapping value and a column mapping value are obtained according to the row magnification ratio and the column magnification ratio, and the second-level region image is interpolated according to the row mapping value and the column mapping value to obtain an enlarged image of the region image.

[0022] Further, after the step of dividing the first-level image into a plurality of region images according to a preset pixel value, the method further comprises:

[0023] A rectangular coordinate system is established with a lower left corner of the first-level image as an origin;

[0024] The coordinate point range of the plurality of region images is obtained by taking a pixel value as an interval.

[0025] Further, the step of calculating the similarity of the multi-stage image sequence and the template image by using the sequential correlation algorithm to obtain the image of the target detection position of the semiconductor chip as the target image comprises:

[0026] extracting one region image of the plurality of region images in the multi-stage image sequence as a target region image, and obtaining an enlarged image of the target region image;

[0027] extracting a plurality of secondary template images with the characteristics of the semiconductor chip from the preset template image according to a preset pixel value;

[0028] calculating the similarity of the enlarged image of the target region image and the plurality of secondary template images, and obtaining a plurality of similarity values;

[0029] judging whether the plurality of similarity values have a similarity value greater than a set threshold value;

[0030] if the plurality of similarity values have a similarity value greater than the set threshold value, taking the target region image as the image of the target detection position;

[0031] recording the coordinate point range of the target region image, and associating the secondary template image with the target region image if the similarity of the secondary template image with the target region image exceeds the set threshold value;

[0032] if the plurality of similarity values do not have a similarity value greater than the set threshold value, returning to the step of extracting one region image of the plurality of region images as a target region image.

[0033] Further, in the step of calculating the similarity of the target region image and the plurality of secondary template images to obtain a plurality of similarity values, the calculation formula is:

[0034]

[0035] normalizing Q to obtain:

[0036]

[0037] wherein, is the average value of the pixel gray level of the target region image, is the average value of the pixel gray level of the template image.

[0038] Further, when the similarity between the target image and the standard image is lower than a set value, the step of correcting the target image and determining that the target detection position has defects when the similarity between the corrected target image and the standard image is lower than a set value, comprises:

[0039] When the similarity between the target image and the standard image is lower than a set value, performing semantic segmentation on the target image to obtain a mask image;

[0040] Obtaining pixel point features of the mask image and performing principal feature analysis to obtain a principal feature vector;

[0041] Determining a rotation angle of the target image according to the principal feature vector;

[0042] Extracting features of the target image and corresponding secondary template images, and obtaining coordinates of the same feature end points;

[0043] Overlapping the features of the target image and the features of the corresponding secondary template images to obtain feature overlapping points;

[0044] Calculating a deflection angle according to the feature overlapping points and the coordinates of the same feature end points;

[0045] When the difference between the rotation angle and the deflection angle is within a set range, rotating the target image according to the rotation angle to obtain a target rotated image;

[0046] When the similarity between the target rotated image and the standard image is lower than a set value, determining that the target detection position has defects.

[0047] The application further provides a semiconductor chip detection device, comprising:

[0048] An acquisition module is configured to acquire an image of a semiconductor chip as a to-be-analyzed image by using a CCD image sensor;

[0049] A division module is configured to divide the to-be-analyzed image into a multi-level image sequence according to a preset template image;

[0050] A calculation module is configured to calculate the similarity between the multi-level image sequence and the preset template image by using a sequential correlation algorithm to obtain an image of a target detection position of the semiconductor chip as a target image;

[0051] A comparison module is configured to compare the target image with a standard image of the target detection position;

[0052] A determination module is configured to determine that the target detection position has defects when the similarity between the target image and the standard image is lower than a set value;

[0053] An updating module is configured to extract a two-dimensional code of the semiconductor chip by using the CCD image sensor, and transmit the target image through a network to update information in the two-dimensional code of the semiconductor chip.

[0054] The application further provides a computer device, including a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0055] The application further provides a computer readable storage medium, which stores a computer program, and the computer program implements the steps of the above method when executed by a processor.

[0056] The application has the following beneficial effects:

[0057] The CCD image sensor is used to collect the image of the semiconductor chip, and the collected image of the semiconductor chip is divided into a plurality of image sequences, the divided plurality of image sequences are enlarged, and the similarity between the region image and the template image in the plurality of image sequences is calculated in sequence to find the target detection position image of the semiconductor chip, finally, the target detection position image is compared with the standard image, when the similarity between the target image and the standard image is lower than a set value, the target image is rotated to determine the similarity again, when the similarity is still lower than the set value, it is determined that the target detection position of the semiconductor chip has a defect, and the image of the defect position is recorded in the two-dimensional code of the semiconductor chip, so that the faulty semiconductor chip can be traced in the future, and the statistical analysis of multiple faults is facilitated. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 The figure is a method flowchart of an embodiment of the application.

[0059] Figure 2 The figure is a device structure schematic diagram of an embodiment of the application.

[0060] Figure 3 The figure is a computer device internal structure schematic diagram of an embodiment of the application.

[0061] The implementation, functional features and advantages of the application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0062] It should be understood that the specific embodiments described herein are merely intended to explain the application, and are not intended to limit the application.

[0063] As Figure 1 shown, the present application provides a semiconductor chip detection method, comprising:

[0064] S1, using a CCD image sensor to obtain an image of a semiconductor chip as a to-be-analyzed image;

[0065] S2, dividing the to-be-analyzed image into a multi-level image sequence according to a preset template image, and magnifying the multi-level image sequence according to the preset template image;

[0066] S3, using a sequential correlation algorithm to calculate the similarity of the multi-level image sequence and the template image to obtain an image of a target detection position of the semiconductor chip as a target image;

[0067] S4, comparing the target image with a standard image of the target detection position;

[0068] S5, when the similarity of the target image and the standard image is lower than a set value, correcting the target image, and when the similarity of the corrected target image and the standard image is lower than the set value, determining that the target detection position has a defect;

[0069] S6, using the CCD image sensor to extract a two-dimensional code of the semiconductor chip, and transmitting the target image through a network to update the information in the two-dimensional code of the semiconductor chip.

[0070] As described above in steps S1-S6, the image sensor CCD (Charge Coupled Device) is a charge coupled device, which can be generally referred to as a CCD image sensor. It is a kind of semiconductor device, which can convert an optical image into a digital signal after capturing the optical image, has the functions of storing signal charge, transmission and reading. The image sensor can directly convert the optical signal into an analog current signal, and the current signal after amplification is subjected to analog-digital conversion to realize image acquisition, storage, transmission, processing and playback. The process of CCD image acquisition mainly converts the captured image from an analog signal into a digital signal that is easy to recognize and distinguish by a computer, and finally converts it into a digital image, and saves the image in the local computer and the server end.

[0071] The CCD image sensor is used to collect the image of the semiconductor chip, and then the collected image of the semiconductor chip is divided into a plurality of image sequences. Since the semiconductor chip has high refinement degree, the plurality of image sequences are enlarged, and then the similarity between the enlarged image of the region image in the plurality of image sequences and the template image is calculated in sequence, so as to increase the detection precision, and finally the target detection position image of the semiconductor chip is found. Finally, the target detection position image is compared with the standard image. When the similarity between the target image and the standard image is lower than the set value, the target image is rotated to determine the similarity again, and when the similarity is still lower than the set value, it is determined that the target detection position of the semiconductor chip has defects, and the image of the defect position is recorded in the two-dimensional code of the semiconductor chip, so as to trace the faulty semiconductor chip subsequently, and facilitate the statistical analysis of multiple faults. The automatic detection, automatic tracing, automatic analysis, automatic feedback and automatic alarm of the semiconductor chip packaging are realized, no human interference is needed, the automation level and degree are greatly improved, the misjudgment caused by the human differences and errors of the enterprise is greatly reduced, the reliability of the semiconductor chip production is improved, the utilization rate of the machine and the factory is improved, and the operation cost of the enterprise is improved.

[0072] In one embodiment, before the step of acquiring the image of the semiconductor chip by the CCD image sensor as the image to be analyzed, the method further comprises:

[0073] S01, recognizing the substrate of the semiconductor chip, and printing a two-dimensional code for carrying data information of the semiconductor chip on the substrate by using a laser;

[0074] S02, extracting the two-dimensional code by using the CCD image sensor, and transmitting the data of the semiconductor chip collected by the CCD image sensor through a network;

[0075] S03, acquiring the two-dimensional code image containing the data of the semiconductor chip, and obtaining the information in the two-dimensional code after code positioning, separation and decoding of the two-dimensional code image, and storing the information in the two-dimensional code into a preset semiconductor chip database.

[0076] As described in the above steps S01-S03, the green paint on the surface of the semiconductor chip substrate is a thin layer of insulating properties, and the circuit of the substrate under the green paint is used to communicate the chip with the tin ball. The laser marking for semiconductor chip manufacturing can select one of the carbon dioxide laser, semiconductor laser and fiber laser; when printing the two-dimensional code, it is also necessary to ensure that the copper circuit layer is not exposed, so that the laser printing on the surface of the chip will not affect the performance of the semiconductor chip. Then the CCD image sensor extracts the two-dimensional code, and the data of the semiconductor chip collected by the CCD image sensor is transmitted through the network, so as to add the data information (product information, production time, serial number, logistics information, etc.) of the semiconductor chip to the two-dimensional code.

[0077] Finally, the two-dimensional code image containing the semiconductor chip data is obtained, and the information in the two-dimensional code is obtained after code positioning, separation and decoding, and stored in the preset semiconductor chip database for subsequent tracing and analysis. The code positioning and separation are mainly achieved through the following methods and steps: the collected image containing the two-dimensional code is converted into a binary image by using the threshold theory of point operation, that is, the image of the target is subjected to binary mathematical processing. The gray value of the pixel point at the point is set as the threshold value. After obtaining the binary image, an inflation operation is performed thereon, mainly using mathematical inflation transformation, edge detection is performed on the inflated image to obtain the contour of the two-dimensional code region, boundary correction is performed, and finally a relatively complete two-dimensional code recognition region is separated. The decoding process is as follows: after obtaining the standard two-dimensional code image, the symbol is sampled as soon as possible, then the graphic points on the intersection are sampled, and it is determined whether it is a dark block or a light block according to the current threshold value. A reasonable bitmap is constructed, 1 is used to represent relatively deep pixels, and 0 is used to represent relatively shallow pixels, so that the most original binary value of the two-dimensional code is obtained, then the data is subjected to data correction, the original data bit stream is converted into data code word according to the coding rule of the two-dimensional code logic, and decoding is realized.

[0078] A database corresponding to each semiconductor chip is established to record the production history of each semiconductor chip in real time for tracing, which can help enterprises collect engineering data and help production quickly find the chip fault, and then feedback the production for improvement.

[0079] In one embodiment, the step of dividing the image to be analyzed into a plurality of image sequences according to a preset template image and magnifying the plurality of image sequences according to the preset template image comprises:

[0080] S21, the image to be analyzed is divided into a first-level image composed of one pixel according to the weighted average of the pixel value of every x×y pixel of the template image;

[0081] S22, dividing the primary image into several regional images according to preset pixel values;

[0082] S23, combining the image to be analyzed, the primary image, and the several regional images to form the multi-level image sequence;

[0083] S24, extracting one regional image from the multi-level image sequence in sequence to obtain a plurality of secondary regional images;

[0084] S25, obtaining a row magnification ratio and a column magnification ratio according to a preset template image and a regional image;

[0085] S26, obtaining a row mapping value and a column mapping value according to the row magnification ratio and the column magnification ratio, and performing interpolation on the secondary regional image according to the row mapping value and the column mapping value to obtain a magnified image of the regional image.

[0086] As described in steps S21-S26 above, the image to be analyzed is weighted and averaged according to the pixel value of every x×y pixels of the template image to form a primary image, that is, the pixel of the image to be analyzed is the same as that of the template image, and then the primary image is divided into several regional images according to preset pixel values. When dividing the regional images, the regions have continuity. When the side of an image row or column is less than the preset pixel value, it is identified as a regional image. In this way, several regional images are obtained. Finally, the image to be analyzed, the primary image, and the several regional images are combined to form the multi-level image sequence, so that the subsequent extracted images can be calculated for similarity. An expansion matrix is constructed. The pixel values of the first row of the regional image are copied to the middle position of the first row of the expansion matrix. The pixel value of the first column of the first row of the regional image is copied to the position of the first column of the first row of the expansion matrix. The pixel value of the last column of the first row of the regional image is copied to the position of the last column of the first row of the expansion matrix, to form the pixel values of the first row of the expansion matrix. The pixel values of the last row, the first column, and the last column of the expansion matrix are obtained in the same way as described above. The pixel values of the regional image are copied to the middle position of the expansion matrix, and finally the secondary regional image is obtained. The row magnification ratio and the column magnification ratio are calculated according to the preset template image and the secondary regional image. The regional image is a*b, and the template image is M*N (M>a, N>b). The magnification ratio in the row direction is x=N / b, and the magnification ratio in the column direction is y=M / a. According to the magnification ratio x in the row direction, the mapping value in the row direction is calculated as 1 / x. According to the magnification ratio y in the column direction, the mapping value in the column direction is calculated as 1 / y. Using the mapping values, the secondary regional image can be interpolated in the column direction and the row direction using the existing technology to obtain the magnified image of the regional image.

[0087] In one embodiment, after the step of dividing the primary image into a plurality of region images according to a preset pixel value, the method further comprises:

[0088] S221, establishing a rectangular coordinate system with the lower left corner of the primary image as the origin;

[0089] S222, obtaining the coordinate point range of the plurality of region images by taking one pixel value as one interval.

[0090] As described in steps S221-S222, a rectangular coordinate system is established with the lower left corner of the primary image as the origin, and then one pixel value is taken as one interval, i.e. one pixel on the coordinate axis in the rectangular coordinate system is 1, two pixels are 2, and so on. The numerical value on the coordinate axis is obtained, and the coordinate point range of the plurality of region images is obtained according to the numerical value on the coordinate axis, so as to record the position of the required region image in the primary image, i.e. to obtain the position of the required detection position in the semiconductor chip.

[0091] In one embodiment, the step of calculating the similarity of the multi-level image sequence and the preset template image by using the sequential correlation algorithm to obtain the image of the target detection position of the semiconductor chip as the target image comprises:

[0092] S31, extracting one region image of the plurality of region images in the multi-level image sequence as a target region image, and obtaining an enlarged image of the target region image;

[0093] S32, extracting a plurality of secondary template images having the characteristics of the semiconductor chip from the preset template image according to a preset pixel value;

[0094] S33, calculating the similarity of the enlarged image of the target region image and the plurality of secondary template images, and obtaining a plurality of similarity values;

[0095] S34, determining whether there is a similarity value greater than a set threshold value in the plurality of similarity values;

[0096] S35, if there is a similarity value greater than the set threshold value in the plurality of similarity values, taking the target region image as the image of the target detection position;

[0097] S36, recording the coordinate point range of the target region image, and extracting the secondary template image similar to the target region image and associated with the target region image whose similarity exceeds the set threshold value;

[0098] S37, if there is no similarity value greater than the set threshold value in the plurality of similarity values, returning to the step of extracting one region image of the plurality of region images as a target region image.

[0099] As described in the steps S31-S35, a region image of the several region images in the multi-level image sequence is extracted as a target region image in sequence, and an enlarged image of the target region image is obtained, so as to analyze the several regions of the one-level image one by one; a plurality of secondary template images with semiconductor chip characteristics in the preset template image are extracted according to a preset pixel value, that is, each secondary template image is an image with semiconductor chip characteristics, and is also a template image corresponding to a fault point of a subsequent semiconductor chip; the similarity between the enlarged image of the target region image and the plurality of secondary template images is calculated, and a plurality of similarity values are obtained. Since the corresponding relationship between the target region image and the plurality of secondary template images is unknown, the similarity between each target region image and each secondary template image is calculated. For example, there are 100 target region images and 50 secondary template images, and the calculated similarity has 100*50. It is judged whether there is a similarity value greater than a set threshold value in the plurality of similarity values. There are 50 similarity values obtained by the target region and the 50 secondary template images, and it is judged whether there is a similarity value greater than a set threshold value in the 50 similarity values. If there is, it indicates that the target region image has a corresponding relationship with the corresponding secondary template image, indicating that the target region is an image of a target detection position. At this time, the coordinate point range of the target region image is recorded, and the secondary template image with a similarity to the target region image exceeding the set threshold value is associated with the target region image, so as to trace back the target detection position subsequently. If there is no similarity value greater than the set threshold value in the 50 similarity values, it indicates that the target region image is not an image of the target detection position, and the step of extracting one region image in the multi-region image as a target region image is returned, and the similarity calculation of the next target region image is performed.

[0100] In one embodiment, in the step of calculating the similarity between the target region image and each template image in the plurality of template images and obtaining a plurality of similarity values, the calculation formula is:

[0101]

[0102] Normalization is performed on Q to obtain:

[0103]

[0104] wherein, is the average value of the pixel gray level of the target region image, is the average value of the pixel gray level of the template image.

[0105] In one embodiment, the step of correcting the target image when the similarity between the target image and the standard image is lower than a set value, and determining that the target detection position has a defect when the similarity between the corrected target image and the standard image is lower than a set value, comprises:

[0106] S51, when the similarity of the target image and the standard image is lower than a set value, performing semantic segmentation on the target image to obtain a mask image;

[0107] S52, obtaining pixel point features of the mask image and performing principal feature analysis to obtain a principal feature vector;

[0108] S53, determining a rotation angle of the target image according to the principal feature vector;

[0109] S54, extracting features of the target image and corresponding secondary template images and obtaining coordinates of the same feature end points;

[0110] S55, superimposing the features of the target image and the corresponding secondary template images to obtain feature superimposition points;

[0111] S56, calculating a deflection angle according to the feature superimposition points and the coordinates of the same feature end points;

[0112] S57, when the difference between the rotation angle and the deflection angle is within a set range, rotating the target image according to the rotation angle to obtain a target rotated image;

[0113] S58, when the similarity of the target rotated image and the standard image is lower than a set value, determining that the target detection position has defects.

[0114] As described in steps S51-S58, since the angle of the image also affects the similarity of the image, when the similarity of the target image and the standard image is lower than the set value, it is also necessary to determine whether the angle of the image has affected the similarity of the image. The target image is subjected to a semantic segmentation algorithm such as FCNN, SegNet, DeepLab v1 / v2 / v3, watershed, and grabcut to obtain a mask image, obtain pixel features of the mask image and perform principal component analysis to obtain a principal component vector. The principal component analysis converts a group of variables that may be correlated into a group of linearly uncorrelated variables through orthogonal transformation, and the converted group of variables is the principal component vector. Finally, the reference direction is obtained according to the principal component vector, the included angle between the principal component vector and the reference direction is calculated, and then the rotation angle of the target image is determined according to the included angle. The feature coordinates of the target image and the corresponding secondary template image are extracted, and the coordinates of the same feature end points are obtained, i.e. the coordinates of the same positions of the target image and the secondary template image. Then the features of the target image and the corresponding features of the secondary template image are overlapped to obtain feature overlapping coordinates. According to the overlapping coordinates and the coordinates of the two feature end points, the deflection angle of the feature end points can be calculated. Multi-directional calculation of the angle of the target image can avoid errors. Finally, when the difference between the rotation angle and the deflection angle is within a set range, it indicates that the target image can be rotated according to the rotation angle, so the target image is rotated according to the rotation angle to obtain a target rotated image. The similarity of the target rotated image and the standard image is calculated again, and when the similarity of the target rotated image and the standard image is lower than the set value, it is determined that the target detection position has a defect.

[0115] In one embodiment, before the step of extracting the two-dimensional code of the semiconductor chip using the CCD image sensor and transmitting the target image through the network to update the information in the two-dimensional code of the semiconductor chip, the method further comprises:

[0116] S061, adding a fault mark to the target image for subsequent traceability, and starting an alarm device to issue an alarm to remind the management personnel that the current semiconductor chip detection has failed.

[0117] As described in the above step S061, the system automatically compares the target image with the template image according to the target image, judges according to the standard and marks the fault, and the system records the two-dimensional code information and engineering data of the semiconductor chip fault and the corresponding image information, so as to facilitate the traceability of the semiconductor chip. The traceability of the semiconductor chip gives each chip product a unique identity, that is, a two-dimensional code. All products circulating in the market can understand the material composition, production process and sales channel of the product, and the core test performance index of the semiconductor chip through the two-dimensional code. By using the application of the two-dimensional code and combining the collection of image processing, a perfect semiconductor chip database system is established. The quality index monitoring of the whole production process is analyzed in real time, some engineering problems are quickly and automatically responded, product tracking and tracing are realized, the influence range of the problem product and the recovery of the sold product are realized, the economic loss and spiritual loss of the consumers and enterprises are reduced, and the limited manual labor is greatly liberated.

[0118] As Figure 2 shown, the present application also provides a semiconductor chip detection device, comprising:

[0119] An acquisition module 1 is configured to acquire an image of a semiconductor chip as an image to be analyzed by using a CCD image sensor.

[0120] A division module 2 is configured to divide the image to be analyzed into a multi-level image sequence according to a preset template image, and to magnify the multi-level image sequence according to the preset template image.

[0121] A calculation module 3 is configured to calculate the similarity between the multi-level image sequence and the preset template image by using a sequential correlation algorithm, so as to obtain an image of a target detection position of the semiconductor chip as a target image.

[0122] A comparison module 4 is configured to compare the target image with a standard image of the target detection position.

[0123] A determination module 5 is configured to correct the target image when the similarity between the target image and the standard image is lower than a set value, and to determine that there is a defect in the target detection position when the similarity between the corrected target image and the standard image is lower than the set value.

[0124] An updating module 6 is configured to extract a two-dimensional code of the semiconductor chip by using the CCD image sensor, and to transmit the target image through a network to update the information in the two-dimensional code of the semiconductor chip.

[0125] In one embodiment, the semiconductor chip detection device further comprises:

[0126] The recognition module is configured to recognize a substrate of the semiconductor chip and print a two-dimensional code for carrying data information of the semiconductor chip on the substrate by using a laser.

[0127] The transmission module is configured to extract the two-dimensional code by using the CCD image sensor and transmit the data of the semiconductor chip collected by the CCD image sensor through a network.

[0128] The storage module is configured to acquire a two-dimensional code image containing data of the semiconductor chip, locate, separate and decode the two-dimensional code image to obtain information in the two-dimensional code, and store the information in the two-dimensional code into a preset semiconductor chip database.

[0129] In one embodiment, the division module 2 comprises:

[0130] The primary image unit is configured to average pixel values of every x×y pixels of a template image to form a primary image.

[0131] The region image unit is configured to divide the primary image into a plurality of region images according to a preset pixel value.

[0132] The combination unit is configured to combine the to-be-analyzed image, the primary image and the plurality of region images to form the multi-level image sequence.

[0133] The secondary region image unit is configured to sequentially extract one region image in the multi-level image sequence to obtain a plurality of secondary region images.

[0134] The magnification ratio unit is configured to obtain a row magnification ratio and a column magnification ratio according to the template image and the secondary region image.

[0135] The magnified image unit is configured to obtain a row mapping value and a column mapping value according to the row magnification ratio and the column magnification ratio, and perform interpolation on the secondary region image according to the row mapping value and the column mapping value to obtain a magnified image of the region image.

[0136] In one embodiment, the calculation module 3 comprises:

[0137] The coordinate system module is configured to establish a rectangular coordinate system with a lower left corner of the primary image as an origin.

[0138] The coordinate point range module is configured to obtain a coordinate point range of the plurality of region images with one pixel value as an interval.

[0139] In one embodiment, the calculation module 3 comprises:

[0140] a target region image extraction unit configured to extract one region image of a plurality of region images in the multi-stage image sequence as a target region image, and obtain an enlarged image of the target region image;

[0141] a secondary template image unit configured to extract a plurality of secondary template images having the semiconductor chip features from the preset template image according to a preset pixel value;

[0142] a similarity value calculation unit configured to calculate similarities between the enlarged image of the target region image and the plurality of secondary template images, and obtain a plurality of similarity values;

[0143] a judgment unit configured to judge whether the plurality of similarity values has a similarity value greater than a set threshold value;

[0144] a target detection position unit configured to take the target region image as an image of the target detection position when the plurality of similarity values has a similarity value greater than the set threshold value;

[0145] a correlation unit configured to record a coordinate point range of the target region image, and extract a secondary template image having a similarity greater than a set threshold value with the target region image and correlate the secondary template image with the target region image;

[0146] a return unit configured to return to the step of extracting one region image of a plurality of region images as a target region image when the plurality of similarity values does not have a similarity value greater than the set threshold value.

[0147] In one embodiment, the similarity value calculation unit, the calculation formula is:

[0148]

[0149] Normalization of Q gives:

[0150]

[0151] wherein, is an average value of pixel gray levels of the target region image, is an average value of pixel gray levels of the template image.

[0152] In one embodiment, the determination module 6 comprises:

[0153] a mask image unit configured to perform semantic segmentation on the target image to obtain a mask image when a similarity between the target image and the standard image is lower than a set value;

[0154] a main feature analysis unit configured to obtain pixel point features of the mask image and perform main feature analysis to obtain a main feature vector;

[0155] a rotation angle unit configured to determine a rotation angle of the target image according to the main feature vector;

[0156] a feature endpoint unit configured to extract features of the target image and the corresponding secondary template image and obtain coordinates of the same feature endpoint;

[0157] a feature coincidence point unit configured to coincide the features of the target image and the corresponding secondary template image to obtain feature coincidence points;

[0158] a deflection angle unit configured to calculate a deflection angle according to the coordinates of the feature coincidence points and the same feature endpoint;

[0159] a target rotated image unit configured to rotate the target image according to the rotation angle to obtain a target rotated image when a difference between the rotation angle and the deflection angle is within a set range;

[0160] a defect unit configured to determine that the target detection position has a defect when a similarity between the target rotated image and the standard image is lower than a set value.

[0161] In one embodiment, the semiconductor chip detection method further comprises:

[0162] a marking module configured to add a fault mark to the target image for subsequent tracing and start an alarm device to issue an alarm to remind the management personnel that the current semiconductor chip detection has a fault.

[0163] The above modules and units are used to execute the steps of the semiconductor chip detection method, and the specific implementation manners are described in the semiconductor chip detection method, which will not be described here.

[0164] As shown in Figure 3 The present application also provides a computer device, which can be a server, and the internal structure thereof can be as shown in Figure 3 The computer device comprises a processor, a memory, a network interface and a database connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store all data required by the process of the semiconductor chip detection method. The network interface of the computer device is configured to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement the semiconductor chip detection method.

[0165] Those skilled in the art can understand that, Figure 3The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied.

[0166] An embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the semiconductor chip detection method.

[0167] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware, and the computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments can be included. Any reference to memory, storage, databases, or other media in this application and the embodiments used herein can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0168] It should be noted that in this document, the terms "comprising", "including", or any other variant thereof are intended to cover non-exclusive inclusions, so that processes, devices, articles or methods including a series of elements not only include those elements, but also include other elements not explicitly listed, or include elements inherent to such processes, devices, articles or methods. Without more limitations, the element defined by the statement "including a" does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0169] The above description is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method of detecting a semiconductor chip, characterized by, The application relates to a semiconductor chip defect detection method and device. The CCD image sensor is used to acquire an image of a semiconductor chip as an image to be analyzed; The image to be analyzed is divided into a plurality of image sequences according to a preset template image, and the plurality of image sequences are enlarged according to the preset template image; specifically, the image to be analyzed is divided into a first-level image according to the weighted average of pixel values of every x*y pixels of the template image; the first-level image is divided into a plurality of region images according to a preset pixel value, a rectangular coordinate system is established with the lower left corner of the first-level image as the origin, and a pixel value is taken as an interval to obtain the coordinate point range of the plurality of region images; the image to be analyzed, the first-level image and the plurality of region images are combined to form the plurality of image sequences; A region image in the plurality of image sequences is extracted in sequence to be enlarged to obtain a plurality of second-level region images; According to the preset template image and the second-level region image, a row enlargement ratio and a column enlargement ratio are obtained; According to the row enlargement ratio and the column enlargement ratio, a row mapping value and a column mapping value are obtained, and the second-level region image is interpolated according to the row mapping value and the column mapping value to obtain an enlarged image of the region image; The sequential correlation algorithm is used to calculate the similarity between the plurality of image sequences and the preset template image to obtain an image of a target detection position of the semiconductor chip as a target image; The target image is compared with a standard image of the target detection position; When the similarity between the target image and the standard image is lower than a set value, the target image is corrected, and when the similarity between the corrected target image and the standard image is lower than the set value, it is determined that the target detection position has a defect; The CCD image sensor is used to extract a two-dimensional code of the semiconductor chip, and the target image is transmitted through a network to update information in the two-dimensional code of the semiconductor chip. Before the step of acquiring the image of the semiconductor chip as the image to be analyzed by the CCD image sensor, the method further comprises the following steps: A substrate of the semiconductor chip is identified, and a two-dimensional code for carrying data information of the semiconductor chip is printed on the substrate by using a laser; The two-dimensional code is extracted by the CCD image sensor, and the data of the semiconductor chip collected by the CCD image sensor is transmitted through a network; 2. The method of claim 1, wherein the step of detecting the semiconductor chip is performed by a method comprising: A two-dimensional code image containing semiconductor chip data is acquired, information in the two-dimensional code is obtained after code positioning, separation and decoding of the two-dimensional code image, and the information in the two-dimensional code is stored in a preset semiconductor chip database. The step of calculating the similarity between the plurality of image sequences and the template image by using the sequential correlation algorithm to obtain the image of the target detection position of the semiconductor chip as the target image comprises the following steps: A region image in the plurality of image sequences is extracted as a target region image, and an enlarged image of the target region image is acquired; According to the preset pixel value, a plurality of second-level template images with the characteristics of the semiconductor chip in the preset template image are extracted; 3. The method of claim 1, wherein the step of detecting the semiconductor chip is performed by a method comprising: ​ ​ ​ Calculate the similarity of the enlarged image of the target region image and the plurality of secondary template images, and obtain a plurality of similarity values; Determine whether there is a similarity value greater than a set threshold value in the plurality of similarity values; If there is a similarity value greater than a set threshold value in the plurality of similarity values, the target region image is taken as the image of the target detection position; Record the coordinate point range of the target region image, and associate the secondary template image with the target region image whose similarity exceeds the set threshold value with the target region image; If there is no similarity value greater than a set threshold value in the plurality of similarity values, return to the step of extracting one of the plurality of region images as the target region image.

4. The method of claim 3, wherein the step of detecting the semiconductor chip is performed by a method comprising: In the step of calculating the similarity of the target region image and the plurality of secondary template images to obtain a plurality of similarity values, the calculation formula is: Q is normalized to obtain: wherein, is the average value of the pixel gray levels of the target region image, is the average value of the pixel gray levels of the template image.

5. The method for detecting semiconductor chips according to claim 1, characterized in that, When the similarity of the target image and the standard image is lower than a set value, the target image is corrected, and when the similarity of the corrected target image and the standard image is lower than a set value, it is determined that the target detection position has defects, and the step comprises: When the similarity of the target image and the standard image is lower than a set value, the target image is subjected to semantic segmentation to obtain a mask image; Obtain the pixel point features of the mask image and perform principal feature analysis to obtain a principal feature vector; Determine the rotation angle of the target image according to the principal feature vector; Extract the features of the target image and the corresponding secondary template image, and obtain the coordinates of the same feature end point; The features of the target image and the corresponding secondary template image are superimposed to obtain a feature superposition point; According to the feature superposition point and the coordinates of the same feature end point, the deflection angle is calculated; When the difference between the rotation angle and the deflection angle is within a set range, the target image is rotated according to the rotation angle to obtain a target rotated image; When the similarity of the target rotated image and the standard image is lower than a set value, it is determined that the target detection position has defects.

6. An apparatus for detecting a semiconductor chip, characterized by comprising: Comprise: An acquisition module is configured to acquire an image of a semiconductor chip as a to-be-analyzed image using a CCD image sensor; A division module is configured to divide the to-be-analyzed image into a multi-level image sequence according to a preset template image, and to enlarge the multi-level image sequence according to the preset template image; specifically comprising: A first image unit is configured to average the pixel values of every x×y pixels of the to-be-analyzed image according to the template image to obtain a first image composed of one pixel; A region image unit is configured to divide the first image into a plurality of region images according to a preset pixel value, to establish a rectangular coordinate system with the lower left corner of the first image as the origin, and to obtain the coordinate point range of the plurality of region images with one pixel value as an interval; A combination unit is configured to combine the to-be-analyzed image, the first image, and the plurality of region images to form the multi-level image sequence; A secondary region image unit is configured to sequentially extract one region image in the multi-level image sequence for enlargement to obtain a plurality of secondary region images; An amplification ratio unit is configured to obtain a row amplification ratio and a column amplification ratio according to a preset template image and a secondary region image; An amplification image unit is configured to obtain a row mapping value and a column mapping value according to the row amplification ratio and the column amplification ratio, and perform interpolation on the secondary region image according to the row mapping value and the column mapping value to obtain an amplified image of the region image; A calculation module is configured to calculate the similarity between a multi-level image sequence and a preset template image by using a sequential correlation algorithm to obtain an image of a target detection position of the semiconductor chip as a target image; A comparison module is configured to compare the target image with a standard image of the target detection position; A determination module is configured to correct the target image when the similarity between the target image and the standard image is lower than a set value, and determine that the target detection position has a defect when the similarity between the corrected target image and the standard image is lower than the set value; An update module is configured to extract a two-dimensional code of the semiconductor chip by using the CCD image sensor, and transmit the target image through a network to update information in the two-dimensional code of the semiconductor chip. 7.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-6 when the computer program is executed by the processor. The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 5.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 5.

Images