Intelligent wafer package identification method, system and equipment

Abstract

The invention belongs to the technical field of packaging, and relates to an intelligent wafer packaging identification method, system and equipment. The method comprises the following steps: acquiring a label image and preprocessing the label image; a label front view is obtained according to the preprocessed label image, so that label information is clearer and easier to read; obtaining a label text according to the label front view, and automatically completing extraction and recognition of text information; the corresponding packaging program is selected according to the label text, and the packaging program is started, so that packaging automation is realized, the production efficiency is improved, and the problem that errors are easily caused when the packaging program is manually selected is solved; the packaging picture after packaging is obtained, whether the packaging mode meets the requirement or not is judged through the packaging picture, the packaging quality is monitored and detected in real time, and it is ensured that the packaging mode meets the requirement. According to the invention, the label image of the wafer can be accurately identified and read, the packaging program can be accurately called, the packaged packaging bag can be accurately identified, and the packaging quality of the wafer is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of packaging, and relates to an intelligent wafer packaging identification method, system and equipment. Background Art

[0002] As market demands become increasingly diversified, the packaging requirements for wafer shipments are becoming more diverse and complex. However, in this link, the existing operating methods still mainly rely on human eye recognition to distinguish different packages.

[0003] At the operational level, workers need to rely on personal experience and visual judgment to distinguish between various packaging methods, which undoubtedly increases the complexity of operations and the possibility of errors. Due to the lack of effective fool-proofing measures, such as standardized operating procedures, visual aids or automated identification systems, there is a risk of mixing packaging, which affects product quality.

[0004] At the inspection level, there are also problems that cannot be ignored. At present, the packaged products are not checked by the system for foolproof card control, which means that packaging errors are difficult to detect and correct later. Once the wrongly packaged products are delivered to customers, it will not only lead to customer dissatisfaction and complaints, but may also lead to a series of subsequent problems such as returns and compensation, causing losses to the company.

[0005] In summary, the following problems exist in the current packaging process for shipments: Operational level: relying on personnel to distinguish different packaging methods, without fool-proof measures, it is easy to mix packaging; Inspection level: There is no systematic anti-fool card control after packaging, and packaging errors are not easy to find. Summary of the invention

[0006] The purpose of the present invention is to provide an intelligent wafer packaging identification method, system and equipment to solve the technical problems of packaging mixing and packaging errors that are difficult to detect. The present invention can accurately identify and read the label image of the wafer, accurately call the packaging program, and accurately identify the packaging bag after packaging to ensure the packaging quality of the wafer.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides an intelligent wafer packaging identification method, comprising the following steps: Acquire a label image and preprocess the label image to obtain a preprocessed label image; Obtain a label front view according to the preprocessed label image; Get the label text according to the label front view; Select the corresponding packaging program according to the label text and start the packaging program; Get the packaging picture after packaging, and judge whether the packaging method meets the requirements through the packaging picture.

[0008] Furthermore, the label image is obtained and preprocessed as follows: Get label image; Convert the color space of the label image to LAB color space and separate the L channel; Perform contrast-limited adaptive histogram equalization on the L channel; Merge the L channel and the AB channel after the contrast-limited adaptive histogram equalization processing; Convert the merged LAB color space to RGB color space to obtain an RGB color space image; The median filtering method is applied to denoise the RGB color space image and eliminate the interference of complex background.

[0009] Furthermore, the contrast-limited adaptive histogram equalization processing is performed on the L channel, specifically as follows: The image of L channel is divided into blocks of nxn pixels using contrast limited adaptive histogram equalization; Use clipLimit=4 contrast to perform histogram equalization on each block individually; Interpolation methods are used to smoothly connect the histogram equalized blocks.

[0010] Furthermore, the label front view is obtained according to the preprocessed label image, as follows: Obtain the coordinates of the four vertices of the label according to the preprocessed label image; The front view of the label is obtained according to the preprocessed label image and the coordinates of the four vertices of the label.

[0011] Furthermore, the coordinates of the four vertices of the label are obtained according to the preprocessed label image, as follows: The Canny algorithm is used to detect all edge information of the preprocessed label image, and the Sobel operator size is set to 7 to calculate the image gradient to obtain the edge image. The edge image is binarized to obtain a binary image. The contour extraction algorithm is used to identify and extract all closed contour areas from the binary image, the number of pixels covered by each closed contour area is calculated, the first six closed contour areas with the largest number of pixels are obtained and sorted in descending order, and the polygon approximation method is used to reduce the number of vertices of the contour for the six sorted closed contour areas to obtain the approximated polygon; Get the coordinates of the four vertices of the label based on the approximated polygon.

[0012] Furthermore, the front view of the label is obtained according to the preprocessed label image and the coordinates of the four vertices of the label, as follows: Calculate the perspective transformation matrix based on the four vertex coordinates of the label; The preprocessed label image is spatially mapped using a perspective transformation matrix to correct the distortion and tilt of the preprocessed label image, thereby obtaining a front view of the label.

[0013] Furthermore, the label text is obtained according to the label front view, as follows: Collect standard label images containing preset fonts, annotate the preset fonts on the standard label images, and construct a label text dataset based on the annotated standard label images; Use DB algorithm to obtain the text area in the front view of the label to be detected; The CRNN algorithm is used in combination with the labeled text data set to perform text recognition on the text area and obtain the recognition result; The tilted text in the recognition result is processed by the direction classifier to obtain the label text.

[0014] Furthermore, the packaging picture after packaging is obtained, and whether the packaging method meets the requirements is determined by the packaging picture, as follows: Obtain the packaged image after packaging, and use a small area sampling method to obtain pixels of the package image; The red light band, green light band and blue light band of the pixels of the packaging image are split according to the image channel dimension to form a grayscale image of a single light band; The grayscale images of the single light bands are combined in a permutation and combination manner to obtain a combined grayscale image; Perform image fusion on the grayscale image of the combined image to form a new packaging image; The pixels in the detection area of ​​the new packaging image are judged in multiple dimensions according to the pixel attributes, and all packaging images are traversed for scoring. Whether the packaging type meets the requirements is determined based on the multi-dimensional judgment results and the scoring results.

[0015] In a second aspect, the present invention provides an intelligent wafer packaging identification system, comprising: Preprocessing module: used to obtain the label image and preprocess the label image to obtain the preprocessed label image; Front view acquisition module: used to acquire the label front view according to the preprocessed label image; Label text acquisition module: used to obtain label text according to the label front view; Packaging program startup module: used to select the corresponding packaging program according to the label text and start the packaging program; Packaging judgment module: used to obtain the packaging picture after packaging, and judge whether the packaging method meets the requirements through the packaging picture.

[0016] In a third aspect, the present invention provides an electronic device, comprising: a processor; a memory for storing computer program instructions; and steps for implementing an intelligent wafer packaging identification method when executing the computer program.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. The method of the present invention obtains a label image and preprocesses the label image to reduce noise and interference in the image, improve the accuracy of subsequent processing, and obtain a preprocessed label image; obtains a label front view based on the preprocessed label image to make the label information clearer and easier to read; obtains the label text based on the label front view, and automatically completes the extraction and recognition of text information; selects the corresponding packaging program based on the label text, and starts the packaging program, which is conducive to realizing packaging automation, improving production efficiency, and solving the problem of easy errors in manually selecting packaging programs; obtains the packaging picture after packaging, and judges whether the packaging method meets the requirements through the packaging picture, and monitors and detects the packaging quality in real time to ensure that the packaging method meets the requirements. The present invention can realize accurate recognition and reading of the label image of the wafer, realize accurate calling of the packaging program, and can accurately identify the packaging bag after packaging to ensure the packaging quality of the wafer.

[0018] 2. The system of the present invention includes: a preprocessing module, a front view acquisition module, a label text acquisition module, a packaging program startup module and a packaging judgment module. The preprocessing module is used to acquire the label image and preprocess the label image to obtain the preprocessed label image; the front view acquisition module is used to acquire the label front view according to the preprocessed label image; the label text acquisition module is used to acquire the label text according to the label front view; the packaging program startup module is used to select the corresponding packaging program according to the label text and start the packaging program; the packaging judgment module is used to acquire the packaging picture after packaging, and judge whether the packaging method meets the requirements through the packaging picture. The various modules cooperate with each other to realize the accurate recognition and reading of the label image of the wafer, realize the accurate call of the packaging program, and can accurately identify the packaging bag after packaging to ensure the packaging quality of the wafer.

[0019] 3. The device of the present invention can also realize accurate recognition and reading of the label image of the wafer, realize accurate calling of the packaging program, and can accurately identify the packaging bag after packaging to ensure the packaging quality of the wafer. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of the method of the present invention; Figure 2 It is a system module diagram of the present invention; Figure 3 is a flow chart of a method according to an embodiment of the present invention; Figure 4 This is an effect diagram of obtaining the label text according to the front view of the label according to an embodiment of the present invention, wherein a is the front view of the label, and b is the label text obtained according to the front view of the label; Figure 5 RGB pixel value distribution diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first", "second", etc. in the specification of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] The present invention is further described in detail below in conjunction with the accompanying drawings: See also Figure 1 The present invention discloses an intelligent wafer packaging identification method, comprising the following steps: S1, obtain the label image and preprocess the label image to obtain the preprocessed label image, as follows: Get label image; Convert the color space of the label image to LAB color space and separate the L channel; The L channel is subjected to contrast-limited adaptive histogram equalization, as follows: The image of the L channel is divided into blocks of nxn pixels using contrast-limited adaptive histogram equalization, where n is preferably 16; Use clipLimit=4 contrast to perform histogram equalization on each block individually; Use interpolation methods to smoothly connect the histogram equalized blocks; Merge the L channel and the AB channel after the contrast-limited adaptive histogram equalization processing; Convert the merged LAB color space to RGB color space to obtain an RGB color space image; The median filter method is used to denoise the RGB color space image and eliminate the interference of complex background to improve the accuracy of internal label text recognition. Figure 5 It is the distribution map of RGB pixel values.

[0024] It should be noted that the LAB color space is a nonlinear color space based on human visual perception.

[0025] L: stands for Lightness, with a value range of 0 to 100, where 0 represents pure black and 100 represents pure white.

[0026] A: Represents the color component from green to red, with a value range of -128 to 127, where negative values ​​tend to be green and positive values ​​tend to be red.

[0027] B: represents the color component from blue to yellow, with a value range of -128 to 127, where negative values ​​tend to be blue and positive values ​​tend to be yellow.

[0028] RGB color space is a color model based on the three basic colors of red, green and blue. RGB color space is based on the three basic colors of red, green and blue. By adjusting the brightness and superposition ratio of these three colors, rich and wide colors can be produced. There are infinite different colors in nature, but the human eye can only distinguish a limited number of different colors. The RGB model can represent more than 16 million different colors. It is very close to the colors of nature in the eyes of the human eye, so it is also called the natural color model.

[0029] clipLimit is the threshold used to limit the degree of contrast enhancement in the contrast-limited adaptive histogram equalization algorithm.

[0030] S2, obtain the label front view according to the preprocessed label image, as follows: Get the coordinates of the four vertices of the label according to the preprocessed label image, as follows: The Canny algorithm is used to detect all edge information of the preprocessed label image, and the Sobel operator size is set to 7 to calculate the image gradient to obtain the edge image. The edge image is binarized to obtain a binary image. The contour extraction algorithm is used to identify and extract all closed contour areas from the binary image, the number of pixels covered by each closed contour area is calculated, the first six closed contour areas with the largest number of pixels are obtained and sorted in descending order, and the polygon approximation method is used to reduce the number of vertices of the contour for the six sorted closed contour areas to obtain the approximated polygon; Get the coordinates of the four vertices of the label based on the approximated polygon; The front view of the label is obtained according to the preprocessed label image and the coordinates of the four vertices of the label, as follows: Calculate the perspective transformation matrix based on the four vertex coordinates of the label; The preprocessed label image is spatially mapped using a perspective transformation matrix to correct the distortion and tilt of the preprocessed label image, thereby obtaining a front view of the label.

[0031] It should be noted that the Canny algorithm is a multi-level edge detection algorithm. The Sobel operator is a discrete differential operator used for edge detection, which combines the first-order gradient calculation of image brightness to identify the edges in the image.

[0032] S3, obtain the label text according to the label front view, as follows: Collect standard label images containing preset fonts, annotate the preset fonts on the standard label images, and construct a label text dataset based on the annotated standard label images; Use DB algorithm to obtain the text area in the front view of the label to be detected; The CRNN algorithm is used in combination with the labeled text data set to perform text recognition on the text area and obtain the recognition result; The tilted text in the recognition result is processed by the direction classifier to obtain the label text. Figure 4 .

[0033] It should be noted that the DB algorithm refers to the Differentiable Binarization algorithm, which is a segmentation-based scene text detection algorithm. The purpose of this algorithm is to convert the probability map generated by the segmentation method into a bounding box and text area, which includes the binarization post-processing process.

[0034] The CRNN algorithm, full name Convolutional Recurrent Neural Network, is a deep learning model that is particularly suitable for processing image-based sequence recognition problems, especially scene text recognition.

[0035] S4, selecting a corresponding packaging program according to the label text, and starting the packaging program; S5, obtaining the packaging picture after packaging, and judging whether the packaging method meets the requirements through the packaging picture, as follows: Obtain the packaged image after packaging, and use a small area sampling method to obtain pixels of the package image; The red light band, green light band and blue light band of the pixels of the packaging image are split according to the image channel dimension to form a grayscale image of a single light band; The grayscale images of the single light bands are combined in a permutation and combination manner to obtain a combined grayscale image; Perform image fusion on the grayscale image of the combined image to form a new packaging image; The pixels in the detection area of ​​the new packaging image are judged in multiple dimensions according to the pixel attributes, and all packaging images are traversed for scoring. Whether the packaging type meets the requirements is determined based on the multi-dimensional judgment results and the scoring results.

[0036] See also Figure 1 In another feasible embodiment of the present invention, the following is adaptively modified according to the situation. It includes the following steps: Acquire the label image and preprocess the label image to reduce noise and interference in the image, improve the accuracy of subsequent processing, and obtain the preprocessed label image; Obtain the label front view based on the preprocessed label image to eliminate image deformation and distortion caused by different shooting angles, making the label information clearer and easier to read; Obtain label text based on the label front view, automatically complete the extraction and recognition of text information, greatly improving processing efficiency and accuracy; Select the corresponding packaging program according to the label text and start the packaging program, which is conducive to realizing packaging automation, improving production efficiency, and solving the problem of easy errors in manually selecting packaging programs; The packaging pictures after packaging are obtained, and whether the packaging method meets the requirements is determined by the packaging pictures, and the packaging quality is monitored and tested in real time to ensure that the packaging method meets the requirements. Once a packaging problem is found, the cause of the problem can be immediately traced and located, and corresponding corrective measures can be taken to avoid the spread and impact of quality problems. The present invention can realize accurate recognition and reading of the label image of the wafer, realize accurate calling of the packaging program, can accurately identify the packaging bag after packaging, and ensure the packaging quality of the wafer.

[0037] Embodiment 1: See also Figure 3 This embodiment discloses an intelligent wafer packaging identification method. By combining with an automatic packaging machine, the present invention can call different packaging methods to perform packaging operations and automatically check the packaging results to ensure the correctness of the packaging method.

[0038] The method of the present invention is specifically as follows: S1. The present invention is used to intelligently identify inner packaging labels and extract relevant information such as customer name, product model, batch number, etc. The method of the present invention can capture images of these inner packaging labels and use image recognition algorithms to accurately read and extract the label texts.

[0039] S2. The label text is fed back to the machine system, which will automatically call the corresponding customer's packaging program to execute packaging; S3. After packaging, the packaging method is identified through image recognition, including checking whether the packaging appearance, size, sealing method, etc. meet expectations, ensuring packaging quality, and promptly discovering and correcting any possible packaging errors; S4. The system compares whether the packaging method meets the requirements and presents the results to the operators for follow-up work so that they can understand the status of the packaging operation in a timely manner. If any packaging that does not meet the requirements is found, the operator can immediately take measures to correct it, thereby avoiding the outflow of defective products.

[0040] Embodiment 2: See also Figure 1 This embodiment discloses an intelligent wafer packaging identification method, comprising the following steps: S11. Image preprocessing: In order to solve the problems of uneven illumination and weak contrast in label image shooting, the image color space is converted to LAB and the L channel is separated. The L brightness image is subjected to contrast-limited adaptive histogram equalization using a block size of 16x16 pixels and a contrast amplification limit of clipLimit=4. Finally, the L, A, and B channels are merged and converted to the RGB color space to improve the local contrast of the image. At the same time, the label image will have noise, and the median filter is applied to eliminate the interference of the complex background to improve the accuracy of the internal label text recognition, see Figure 5 It is the distribution map of RGB pixel values.

[0041] S12, label detection: First, Canny is used to detect all edges of the label grayscale image. Since some label contours cannot be detected due to partial exposure of the label image, the size of the Sobel operator is set to 7 to calculate the image gradient and accurately obtain the label contour. Secondly, a contour extraction algorithm is used to obtain all contours of the binary image, the number of pixels covered by each extracted contour (i.e., contour area) is calculated, and the first six contours with the largest contour area are sorted in descending order to locate the contour of the binary image with the largest area; Based on the first six contours with the largest area, within a maximum distance of 2% of the original contour perimeter, the curve is approximated with a straight line segment to reduce the number of contour vertices, and it is determined whether the approximated polygon has exactly four vertices, and finally the wafer label image edge is obtained.

[0042] S13, label space mapping: S12 has obtained the coordinates of the four corner points of the positioning label and calculated its perspective transformation matrix to ensure that the label maintains the correct proportion and position relationship after transformation. The perspective transformation matrix is ​​used to perform spatial mapping on the preprocessed label image to correct the distortion and tilt of the label image, thereby obtaining a front view of the shipping label.

[0043] S14. Label text detection and recognition: For the preset fonts of factory labels, a label text dataset is constructed and annotated. The label text dataset is used for training in combination with the CRNN algorithm to facilitate accurate text recognition. The DB (Differentiable Binarization) algorithm is used for text detection. By setting a dual threshold card to control label text detection, in the DB output probability map, pixels with a score higher than the preset threshold 1 are considered as text pixels; when the average score of all text pixels within the detection result frame exceeds the preset threshold 2, it is considered as a text area.

[0044] The CRNN algorithm is used to perform text recognition on the above text area. The CRNN algorithm is a convolutional recurrent neural network algorithm. A direction classifier is introduced to process the tilted text in the recognition result. When the predicted result points to a 180-degree rotation and the score exceeds the preset threshold, the image flipping operation is automatically performed to improve the text recognition accuracy. Figure 4 Get the effect of label text based on the label front view.

[0045] S15, label post-processing: If the wafer label cannot be identified or is identified inaccurately, manual re-inspection is performed. The label text identification information is fed back to the database in a timely manner.

[0046] S2, packaging type detection implementation; Refined detection area: To avoid the impact of the background environment of the factory shipping packaging, which may cause packaging type identification errors, the present invention determines the packaging inspection area based on statistical probability and identifies the packaging type. A small area sampling method is used to extract the pixels of the packaging image of the packaging to be inspected, and multiple distributed small areas are used for judgment to avoid the influence of more concentrated sampling on the judgment results; on the one hand, it provides a reliable source of detection samples for the classification and detection tasks of shipping packaging; on the other hand, it can also avoid algorithm redundancy, improve the detection speed in the production process, and effectively help the efficiency of production and shipment in the factory.

[0047] The present invention makes multi-dimensional judgments on the pixels in the detection area, comprehensively considers all pixel attributes, combines the properties of the shipping packaging materials, and realizes accurate judgment of the type of packaging materials based on the multiple correlations of pixel values.

[0048] Since the original images are difficult to distinguish packaging types, the classification accuracy is low, and they do not have the classification characteristics of packaging classification, in order to solve this problem, based on the color gamut of the pixels of the packaging images, the band combination method is used to seek new rules and find a new method for easy classification of packaging type images. This method includes splitting the red light band (R), green light band (G), and blue light band (B) of the pixels of the packaging image according to the image channel (C) dimension, and forming a single light band R, G, B grayscale image after splitting. Band combination is performed to generate single-band combination grayscale images of R+G, G+B and B+R. Finally, image fusion is performed on the combination grayscale images of single light bands to form a new packaging image that fuses R+G, G+B and B+R. The generation of the new packaging image can show obvious color distinction, which is conducive to the accurate classification of packaging images.

[0049] The new packaging image information helps to distinguish between packaging bag 2 and packaging bag 1, and clarifies the classification boundary between the two packaging. Figure 5 , which is the RGB pixel value distribution diagram of package bag 1 and package bag 2.

[0050] The pixels in the detection area of ​​the new packaging image of the band combination are judged in multiple dimensions, all pixel attributes are comprehensively considered, and the nature of the shipping packaging material is combined. The type of packaging material is accurately judged based on the multiple correlations of pixel values. At the same time, all packaging images are traversed for scoring, and the packaging type is judged based on the scoring results, which effectively improves the accuracy of the outer packaging classification results. The specific steps of traversing all packaging images and scoring are as follows: all outer packaging images in each lot are judged, and the final outer packaging type is determined based on the overall judgment to avoid misjudgment caused by the judgment of a single image.

[0051] The present invention can realize intelligent classification of packaging images and accurate positioning of each detection item; can realize accurate detection of inner packaging labels and extract accurate text information; can eliminate FOSB background interference and implement multi-level parallel target detection; can realize wafer packaging type detection by using pixel value multiple correlation judgment; and can combine multi-source heterogeneous databases and images to realize comparison of detection results with customer needs; The packaging process of the present invention does not need to rely on an online system to call a packaging program, and can use this method to distinguish product requirements offline and execute the corresponding packaging program; Increase intelligent packaging operations and inspections to avoid packaging mix-ups and inspection errors, and improve packaging quality.

[0052] The method of the invention can realize intelligent fool-proof packaging operation mode, which is of great significance.

[0053] Based on the above method, the present invention discloses an intelligent wafer packaging identification system, see Figure 2 ,include: Preprocessing module: used to obtain the label image and preprocess the label image to obtain the preprocessed label image; Front view acquisition module: used to acquire the label front view according to the preprocessed label image; Label text acquisition module: used to obtain label text according to the label front view; Packaging program startup module: used to select the corresponding packaging program according to the label text and start the packaging program; Packaging judgment module: used to obtain the packaging picture after packaging, and judge whether the packaging method meets the requirements through the packaging picture.

[0054] The various modules of the system of the present invention cooperate with each other, can realize accurate recognition and reading of the label image of the wafer, realize accurate calling of the packaging program, can accurately identify the packaging bag after packaging, and ensure the packaging quality of the wafer.

[0055] An electronic device comprises: a processor; a memory for storing computer program instructions; and steps for implementing an intelligent wafer packaging identification method when executing the computer program.

[0056] A storage medium stores computer program instructions. When the computer program instructions are loaded and run by a processor, the processor executes an intelligent wafer packaging identification method.

[0057] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0058] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0059] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0060] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0061] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the present invention.

Claims

1. An intelligent wafer packaging identification method, characterized in that: The following steps are involved: Acquire a label image and preprocess the label image to obtain a preprocessed label image; Obtain a label front view according to the preprocessed label image; Get the label text according to the label front view; Select the corresponding packaging program according to the label text and start the packaging program; Get the packaging picture after packaging, and judge whether the packaging method meets the requirements through the packaging picture.

2. The intelligent wafer packaging identification method according to claim 1, characterized in that: The label image is obtained and preprocessed as follows: Get label image; Convert the color space of the label image to LAB color space and separate the L channel; Perform contrast-limited adaptive histogram equalization on the L channel; Merge the L channel and the AB channel after the contrast-limited adaptive histogram equalization processing; Convert the merged LAB color space to RGB color space to obtain an RGB color space image; The median filtering method is applied to denoise the RGB color space image and eliminate the interference of complex background.

3. The intelligent wafer packaging identification method according to claim 2, characterized in that: The contrast-limited adaptive histogram equalization process is performed on the L channel, specifically as follows: The image of L channel is divided into blocks of nxn pixels using contrast limited adaptive histogram equalization; Use clipLimit=4 contrast to perform histogram equalization on each block individually; Interpolation methods are used to smoothly connect the histogram equalized blocks.

4. The intelligent wafer packaging identification method according to claim 1, characterized in that: The label front view is obtained according to the preprocessed label image, specifically as follows: Obtain the coordinates of the four vertices of the label according to the preprocessed label image; The front view of the label is obtained according to the preprocessed label image and the coordinates of the four vertices of the label.

5. The intelligent wafer packaging identification method according to claim 4, characterized in that: The coordinates of the four vertices of the label are obtained according to the preprocessed label image, as follows: The Canny algorithm is used to detect all edge information of the preprocessed label image, and the Sobel operator size is set to 7 to calculate the image gradient to obtain the edge image. The edge image is binarized to obtain a binary image. The contour extraction algorithm is used to identify and extract all closed contour areas from the binary image, the number of pixels covered by each closed contour area is calculated, the first six closed contour areas with the largest number of pixels are obtained and sorted in descending order, and the polygon approximation method is used to reduce the number of vertices of the contour for the six sorted closed contour areas to obtain the approximated polygon; Get the coordinates of the four vertices of the label based on the approximated polygon.

6. The intelligent wafer packaging identification method according to claim 4, characterized in that: The front view of the label is obtained according to the preprocessed label image and the coordinates of the four vertices of the label, as follows: Calculate the perspective transformation matrix based on the four vertex coordinates of the label; The preprocessed label image is spatially mapped using a perspective transformation matrix to correct the distortion and tilt of the preprocessed label image, thereby obtaining a front view of the label.

7. The intelligent wafer packaging identification method according to claim 1, characterized in that: The label text is obtained according to the label front view, as follows: Collect standard label images containing preset fonts, annotate the preset fonts on the standard label images, and construct a label text dataset based on the annotated standard label images; Use DB algorithm to obtain the text area in the front view of the label to be detected; The CRNN algorithm is used in combination with the labeled text data set to perform text recognition on the text area and obtain the recognition result; The tilted text in the recognition result is processed by the direction classifier to obtain the label text.

8. The intelligent wafer packaging identification method according to claim 1, characterized in that: The method of obtaining the packaging picture after packaging and judging whether the packaging method meets the requirements through the packaging picture is as follows: Obtain the packaged image after packaging, and use a small area sampling method to obtain pixels of the package image; The red light band, green light band and blue light band of the pixels of the packaging image are split according to the image channel dimension to form a grayscale image of a single light band; The grayscale images of the single light bands are combined in a permutation and combination manner to obtain a combined grayscale image; Perform image fusion on the grayscale image of the combined image to form a new packaging image; The pixels in the detection area of ​​the new packaging image are judged in multiple dimensions according to the pixel attributes, and all packaging images are traversed for scoring. Whether the packaging type meets the requirements is determined based on the multi-dimensional judgment results and the scoring results.

9. An intelligent wafer packaging identification system, characterized in that: include: Preprocessing module: used to obtain the label image and preprocess the label image to obtain the preprocessed label image; Front view acquisition module: used to acquire the label front view according to the preprocessed label image; Label text acquisition module: used to obtain label text according to the label front view; Packaging program startup module: used to select the corresponding packaging program according to the label text and start the packaging program; Packaging judgment module: used to obtain the packaging picture after packaging, and judge whether the packaging method meets the requirements through the packaging picture.

10. An electronic device comprising: Processor; memory, electronic device used to store computer program instructions; characterized in that it is used to implement the steps of the intelligent wafer packaging identification method as described in any one of claims 1-8 when executing the computer program.

Images