Wafer edge searching positioning and ID number identification control system and control method thereof
Through an integrated control system, laser and image sensor technology are used to achieve accurate positioning and ID number identification of wafers, solving the problems of low equipment stability and efficiency in the existing technology, reducing costs and improving wafer transmission management accuracy.
Patent Information
- Application Number
- CN202411105448.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-13
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-08-13
AI Technical Summary
In the existing wafer edge positioning and ID number identification technology, the equipment has poor stability, high production costs, low work efficiency, large equipment space and easy to cause damage to the wafer.
Adopting an integrated control system, including FPGA controller, laser emitter, photoelectric sensor, ARM processor, image sensor and OCR identification components, parallel line laser signals are sent through laser emitters, photoelectric sensors collect edge position signals, and ARM processors The wafer offset data is calculated and the edge search motion module is controlled for positioning, and the wafer ID number is identified using image sensors and OCR identification components.
Improve the integration of wafer positioning equipment, reduce equipment development costs, reduce space, improve work efficiency, and avoid damage to the wafer by fixtures, ensuring the edge tangent position identification between different devices. just.
Smart Images

Figure CN119028892A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of semiconductor production equipment, and in particular to a control system for wafer edge finding and positioning and ID number recognition and a control method thereof. Background Art
[0002] In the semiconductor production process, in order to ensure the identification and alignment of the cutting edge position of the wafer when it flows between different process equipment, a dedicated wafer edge finding and positioning device is required. At the same time, the identification code (i.e. ID number) in the wafer needs to be read to manage the wafer transmission or to process the wafer in the next step according to the preset requirements.
[0003] At present, the industry installs the wafer clamping component PA and OCR (Optical Character Recognition) component separately. After adjusting the appropriate position, the wafer clamping component is used to calibrate the wafer position. After calibration, the wafer ID number is read through the OCR component. This method requires the wafer clamping component and the OCR component to be controlled separately, and then assembled according to actual process requirements to achieve the integrated application of wafer edge positioning and ID number recognition. This method has poor equipment stability, increased production costs, low work efficiency, and the assembled entire mechanism itself occupies a large space. In addition, the clamping action of the fixture can easily cause damage to the wafer during operation. Summary of the invention
[0004] In view of this, the present invention provides a control system and a control method for wafer edge finding and ID number recognition to solve the problems of poor equipment stability, high production cost, low work efficiency, large equipment space occupation and easy damage to the wafer in the existing wafer edge finding and ID number recognition technologies.
[0005] The present invention provides a control system for wafer edge-finding positioning and ID number recognition, the control system comprising an FPGA controller, a laser transmitter, a photoelectric sensor, an ARM processor, an image sensor, an OCR recognition component and an edge-finding motion module;
[0006] The laser emitter and the photoelectric sensor are both electrically connected to the FPGA controller, the FPGA controller is communicatively connected to the ARM processor, and the image sensor, the OCR recognition component and the edge-finding motion module are all electrically connected to the ARM processor;
[0007] The FPGA controller is used to send a laser emission instruction to the laser emitter;
[0008] The laser transmitter is used to send a parallel line laser signal to the wafer to be inspected according to the received laser emission instruction;
[0009] The photoelectric sensor is used to collect the edge position electrical signal generated after the parallel line laser signal passes through the edge of the wafer to be inspected, and transmit the edge position electrical signal to the FPGA controller;
[0010] The FPGA controller is used to transmit the received edge position electrical signal to the ARM processor;
[0011] The ARM processor is used to calculate the wafer offset data according to the received edge position electrical signal, and control the edge finding motion module to move according to the wafer offset data, so as to perform edge finding and positioning on the wafer to be inspected;
[0012] The ARM processor is also used to send a wafer image acquisition instruction to the image sensor;
[0013] The image sensor is used to obtain a wafer surface image of the wafer to be inspected according to the received image acquisition instruction, and transmit the wafer surface image to the OCR recognition component through the ARM processor;
[0014] The OCR recognition component is used to identify and output the ID number of the wafer to be inspected based on the received wafer surface image.
[0015] Optionally, the control system further includes a first PWM control unit;
[0016] The FPGA controller is electrically connected to the laser transmitter via the first PWM control unit;
[0017] The FPGA controller is specifically used for:
[0018] Sending the laser emission instruction to the first PWM control unit;
[0019] The first PWM control unit is used to generate a first PWM control signal according to the received laser emission instruction, and send the first PWM control signal to the laser emitter;
[0020] The laser transmitter is specifically used for:
[0021] Under the control of the first PWM control signal, the parallel line laser signal is sent to the wafer to be inspected.
[0022] Optionally, the control system further includes an ADC converter;
[0023] The photoelectric sensor is electrically connected to the FPGA controller through the ADC converter;
[0024] The ADC converter is used to convert the analog edge position electrical signal collected by the photoelectric sensor into an edge position digital signal, and transmit the edge position digital signal to the FPGA controller, and through the FPGA controller, transmit the edge position digital signal to the ARM processor.
[0025] Optionally, the wafer offset data includes a center offset position and a cutting edge offset angle of the wafer to be inspected;
[0026] The ARM processor calculates the wafer offset data according to the received edge position electrical signal, specifically including:
[0027] Calculate the edge position coordinates of the wafer to be inspected according to the edge position digital signal transmitted by the FPGA controller;
[0028] Using the least square method, curve fitting is performed according to the edge position coordinates to obtain the edge curve of the wafer to be inspected;
[0029] The center offset position and the trimming offset angle of the wafer to be inspected are calculated based on the edge curve.
[0030] Optionally, the OCR recognition component identifies and outputs the ID number of the wafer to be inspected based on the received wafer surface image, specifically including:
[0031] Based on the DB optimization method, text detection is performed on the wafer surface image to obtain polygonal information of the text area;
[0032] Performing clipping and perspective transformation processing on the polygonal information of the text area to obtain a rectangular frame of the text area;
[0033] Based on the SVTR text recognition method, text recognition is performed on the text area rectangular frame to obtain the ID number of the wafer to be inspected and output it.
[0034] Optionally, the OCR recognition component performs text detection on the wafer surface image based on a DB optimization method to obtain text area polygon information, specifically including:
[0035] Using a backbone network to extract features from the wafer surface image to obtain multi-stage features;
[0036] The DBFPN neck network structure is used to fuse the multi-stage features to obtain fused features;
[0037] Decoding the fused features to obtain a probability feature map and a threshold feature map, and fusing the probability feature map and the threshold feature map to obtain a binary feature map;
[0038] Calculating the loss functions of the probability feature map, the threshold feature map, and the binary feature map respectively to obtain a probability map loss, a threshold map loss, and a binary map loss;
[0039] The probability map loss, the threshold map loss and the binary map loss are used to perform model training, and the text area polygon information corresponding to the wafer surface image is obtained according to the trained text detection model.
[0040] Optionally, the OCR recognition component performs text recognition on the text area rectangular frame based on the SVTR text recognition method, obtains the ID number of the wafer to be inspected and outputs it, specifically including:
[0041] Input the text area rectangular frame into a preset text recognition model to obtain a target text area;
[0042] Processing the target text region using a progressive overlapping patch embedding method to obtain character components in multiple formats;
[0043] All character components are gradually downsampled in height, and a parallel linear prediction method is used to decode all character components after the gradual downsampling in height, so as to obtain and output the ID number of the wafer to be inspected.
[0044] Optionally, the control system further includes a second PWM control unit and a fill light;
[0045] The ARM processor is electrically connected to the fill light through the second PWM control unit;
[0046] The ARM processor is further configured to send a fill light instruction to the second PWM control unit while sending the wafer acquisition instruction to the image sensor;
[0047] The second PWM control unit is used for generating a second PWM control signal according to the received fill light instruction, and sending the second PWM control signal to the fill light;
[0048] The fill light is used to emit a fill light source to the wafer to be inspected under the control of the second PWM control signal, so as to provide fill light to the surface of the wafer to be inspected.
[0049] Optionally, the control system further comprises a stage control component;
[0050] The stage control component is electrically connected to the ARM processor, and the stage control component is also connected to a stage driver for placing the wafer to be inspected;
[0051] The ARM processor is also used to send a stage motion instruction to the stage control component before sending a wafer image acquisition instruction to the image sensor and before the FPGA controller sends a laser emission instruction to the laser emitter;
[0052] The stage control component is used to control the rotation of the stage on which the wafer to be inspected is placed according to the received stage movement instruction.
[0053] In addition, the present invention also provides a control method for wafer edge finding and positioning and ID number recognition, which uses the aforementioned wafer edge finding and positioning and ID number recognition control system to perform wafer edge finding and positioning and ID number recognition, and the control method includes:
[0054] Using the FPGA controller, send laser emission instructions to the laser transmitter;
[0055] Using a laser transmitter, according to the received laser emission instruction, a parallel line laser signal is sent to the wafer to be inspected;
[0056] Using a photoelectric sensor, collecting an edge position electrical signal generated after the parallel line laser signal passes through the edge of the wafer to be inspected, and transmitting the edge position electrical signal to the FPGA controller;
[0057] Utilizing the FPGA controller, transmitting the received edge position electrical signal to the ARM processor;
[0058] Using the ARM processor, according to the received edge position electrical signal, the wafer offset data is calculated, and according to the wafer offset data, the edge finding motion module is controlled to move, so as to perform edge finding and positioning on the wafer to be inspected;
[0059] Using the ARM processor, sending a wafer image acquisition instruction to an image sensor;
[0060] Using the image sensor, according to the received image acquisition instruction, a wafer surface image of the wafer to be inspected is acquired, and the wafer surface image is transmitted to an OCR recognition component through the ARM processor;
[0061] The OCR recognition component is used to identify and output the ID number of the wafer to be inspected based on the received wafer surface image.
[0062] The beneficial effects of the present invention are as follows: in the entire control system, the integration between the laser transmitter, the photoelectric sensor and the FPGA controller is utilized to obtain the edge position electrical signal generated after the parallel line laser signal passes through the edge of the wafer to be inspected, and then through the data interaction between the FPGA controller and the ARM processor, the ARM processor is utilized to obtain the wafer offset data according to the edge position electrical signal, and then the edge-finding motion module is controlled to move according to the wafer offset data, so as to realize the edge-finding positioning of the wafer; then the integration between the image sensor, the OCR recognition component and the ARM processor is utilized to obtain the wafer surface image, and the ID number of the wafer to be inspected is obtained and output according to the obtained wafer surface image recognition, so as to realize the identification of the wafer ID number, and then ensure the identification and alignment requirements of the trimming position when the wafer flows between different process equipment and the management requirements of the wafer transmission;
[0063] The control system and control method for wafer edge finding and ID number recognition of the present invention realize the integrated design of wafer edge finding and ID number recognition function, greatly improve the integration of wafer positioning equipment, reduce equipment development costs, the equipment occupies a small space, has high work efficiency, will not cause damage to the wafer by the clamping action of the fixture, and the equipment has high stability. BRIEF DESCRIPTION OF THE DRAWINGS
[0064] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings:
[0065] Figure 1 A structural diagram of a control system for wafer edge finding, positioning and ID number recognition in Embodiment 1 of the present invention is shown;
[0066] Figure 2 A structural diagram of another control system for wafer edge positioning and ID number recognition in the first embodiment of the present invention is shown;
[0067] Figure 3 A flow chart of a control method for wafer edge finding and ID number recognition in the second embodiment of the present invention is shown. DETAILED DESCRIPTION
[0068] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 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 those skilled in the art without creative work are within the scope of protection of the present invention.
[0069] Embodiment 1
[0070] A control system for wafer edge positioning and ID number recognition, such as Figure 1 As shown, the control system includes an FPGA controller, a laser transmitter, a photoelectric sensor, an ARM processor, an image sensor, an OCR recognition component and an edge-finding motion module;
[0071] The laser emitter and the photoelectric sensor are both electrically connected to the FPGA controller, the FPGA controller is communicatively connected to the ARM processor, and the image sensor, the OCR recognition component and the edge-finding motion module are all electrically connected to the ARM processor;
[0072] The FPGA controller is used to send a laser emission instruction to the laser emitter;
[0073] The laser transmitter is used to send a parallel line laser signal to the wafer to be inspected according to the received laser emission instruction;
[0074] The photoelectric sensor is used to collect the edge position electrical signal generated after the parallel line laser signal passes through the edge of the wafer to be inspected, and transmit the edge position electrical signal to the FPGA controller;
[0075] The FPGA controller is used to transmit the received edge position electrical signal to the ARM processor;
[0076] The ARM processor is used to calculate the wafer offset data according to the received edge position electrical signal, and control the edge finding motion module to move according to the wafer offset data, so as to perform edge finding and positioning on the wafer to be inspected;
[0077] The ARM processor is also used to send a wafer image acquisition instruction to the image sensor;
[0078] The image sensor is used to obtain a wafer surface image of the wafer to be inspected according to the received image acquisition instruction, and transmit the wafer surface image to the OCR recognition component through the ARM processor;
[0079] The OCR recognition component is used to identify and output the ID number of the wafer to be inspected based on the received wafer surface image.
[0080] In the entire control system of this embodiment, the integration between the laser transmitter, the photoelectric sensor and the FPGA controller is utilized to obtain the edge position electrical signal generated after the parallel line laser signal passes through the edge of the wafer to be inspected, and then through the data interaction between the FPGA controller and the ARM processor, the ARM processor is used to obtain the wafer offset data according to the edge position electrical signal, and then the edge-finding motion module is controlled to move according to the wafer offset data to achieve edge-finding positioning of the wafer; then the integration between the image sensor, the OCR recognition component and the ARM processor is utilized to obtain the wafer surface image, and the ID number of the wafer to be inspected is obtained and output according to the acquired wafer surface image recognition, so as to realize the identification of the wafer ID number, and then ensure the identification and alignment requirements of the cutting edge position and the management requirements of the wafer transmission when the wafer flows between different process equipment.
[0081] The control system for wafer edge finding, positioning and ID number recognition in this embodiment realizes an integrated design of wafer edge finding, positioning and ID number recognition, which greatly improves the integration of wafer positioning equipment and reduces equipment development costs. The equipment occupies a small space and has high working efficiency. There will be no phenomenon of damage to the wafer caused by the clamping action of the fixture, and the equipment has high stability.
[0082] Specifically, in this embodiment, the FPGA (Field-Programmable Gate Array) controller is a field programmable gate array, which contains a large number of programmable logic units, memories and input and output interfaces. A hardware description language (such as VHDL or Verilog) can be used to write programs to implement specific hardware logic functions. The ARM (Advanced RISC Machine) processor is a microprocessor architecture that uses a reduced instruction set (RISC) architecture and has the characteristics of low power consumption, high performance and low cost. Various hardware resources can be provided through software programming to implement various complex functions.
[0083] Preferably, if Figure 2 As shown, the control system further includes a first PWM control unit;
[0084] The FPGA controller is electrically connected to the laser transmitter via the first PWM control unit;
[0085] The FPGA controller is specifically used for:
[0086] Sending the laser emission instruction to the first PWM control unit;
[0087] The first PWM control unit is used to generate a first PWM control signal according to the received laser emission instruction, and send the first PWM control signal to the laser emitter;
[0088] The laser transmitter is specifically used for:
[0089] Under the control of the first PWM control signal, the parallel line laser signal is sent to the wafer to be inspected.
[0090] When triggering the parallel line laser signal, this embodiment uses the first PWM control signal generated by the first PWM control unit for control. The output power of the laser can be controlled by adjusting the pulse width based on the PWM technology. Energy is output only during the pulse period, thereby reducing energy waste during non-working time and improving overall energy efficiency. Compared with the continuous control mode, energy consumption is reduced. The pulse width of the first PWM control signal can be accurately controlled by the PWM technology, thereby achieving fine adjustment of the laser output power and high-precision control of the triggering of the parallel line laser signal, thereby achieving fine control of the edge finding and positioning of the entire wafer.
[0091] Preferably, if Figure 2 As shown, the control system also includes an ADC converter;
[0092] The photoelectric sensor is electrically connected to the FPGA controller through the ADC converter;
[0093] The ADC converter is used to convert the analog edge position electrical signal collected by the photoelectric sensor into an edge position digital signal, and transmit the edge position digital signal to the FPGA controller, and through the FPGA controller, transmit the edge position digital signal to the ARM processor.
[0094] When the parallel line laser signal emitted by the laser transmitter irradiates the wafer surface, the laser signal after passing through the edge of the wafer to be inspected will be different due to the edges at different positions of the wafer. The laser signal after passing through the edge of the wafer is collected and photoelectrically sensed by the photoelectric sensor to obtain different edge position electrical signals, which are analog electrical signals. This embodiment uses the ADC converter between the photoelectric sensor and the FPGA controller to conveniently convert the analog electrical signal into a digital electrical signal (i.e., edge position digital signal), thereby facilitating signal transmission and reception in the signal format supported by the FPAG controller and facilitating subsequent signal analysis by the ARM processor to obtain wafer offset data, thereby realizing edge finding and positioning of the wafer.
[0095] In the actual production and manufacturing process of wafers, wafers are usually placed on the stage manually or by a robot. When the wafer is placed on the stage, the negative pressure adsorption component built into the stage fixes the wafer on the stage. If the wafer is offset, the center of the wafer and the center of the stage will be offset, and the cut edges or corners on the wafer (cut edges or corners are usually used for orientation marking and auxiliary processing in wafer manufacturing) will have an angular offset.
[0096] In this embodiment, the wafer offset data includes the center offset position and the cutting edge offset angle of the wafer to be inspected.
[0097] Based on the above wafer offset data, the subsequent edge-finding motion module can facilitate accurate wafer edge-finding and positioning, ensuring that the wafer to be inspected can be subsequently manufactured according to the set position and direction.
[0098] Preferably, the ARM processor calculates the wafer offset data according to the received edge position electrical signal, specifically including:
[0099] Calculate the edge position coordinates of the wafer to be inspected according to the edge position digital signal transmitted by the FPGA controller;
[0100] Using the least square method, curve fitting is performed according to the edge position coordinates to obtain the edge curve of the wafer to be inspected;
[0101] The center offset position and the trimming offset angle of the wafer to be inspected are calculated based on the edge curve.
[0102] The above method can accurately fit the actual position of the wafer to be inspected, and then determine the wafer offset data (including the center offset position and the cutting edge offset angle) between the wafer to be inspected and the ideal set position (that is, the ideal position without center offset and angle offset), so as to facilitate the subsequent ARM processor to control the movement of the edge-finding motion module, so as to accurately find the edge and position the actual wafer to be inspected according to the center offset position and the cutting edge offset angle, and ensure the identification and alignment of the cutting edge position when the wafer flows between different process equipment.
[0103] Among them, the ARM processor has an integrated computing unit, which calculates the edge position coordinates according to the edge position digital signal, and uses the fitted edge curve to calculate the center offset position and the cutting edge offset angle, both of which can be calculated using mathematical geometry knowledge and mathematical trigonometric function knowledge, which is a conventional technology; the specific operation method of least squares method for curve fitting is also a conventional operation method, and the specific details will not be repeated here.
[0104] This embodiment utilizes the integration of the above-mentioned laser emitter, photoelectric sensor, FPGA controller, ARM processor and edge-finding motion module to achieve precise edge-finding positioning of the wafer, thereby effectively ensuring the identification and alignment of the cutting edge position when the wafer flows between different process equipment.
[0105] Preferably, the OCR recognition component identifies and outputs the ID number of the wafer to be inspected based on the received wafer surface image, which specifically includes:
[0106] Based on the DB optimization method, text detection is performed on the wafer surface image to obtain polygonal information of the text area;
[0107] Performing clipping and perspective transformation processing on the polygonal information of the text area to obtain a rectangular frame of the text area;
[0108] Based on the SVTR text recognition method, text recognition is performed on the text area rectangular frame to obtain the ID number of the wafer to be inspected and output it.
[0109] In the above-mentioned OCR recognition component of this embodiment, the DB optimization method is used to realize text detection, and the SVTR text recognition method is used to realize text recognition, forming a PP-OCRv3 model framework to realize OCR recognition of wafer ID number, and the recognition accuracy and speed are effectively improved, and the recognition error rate is also reduced, which effectively ensures the needs for wafer transmission management in the wafer manufacturing process, and contributes to the automation and intelligence of wafer manufacturing.
[0110] Specifically, the OCR recognition component performs text detection on the wafer surface image based on the DB optimization method to obtain text area polygon information, which specifically includes:
[0111] Using a backbone network to extract features from the wafer surface image to obtain multi-stage features;
[0112] The DBFPN neck network structure is used to fuse the multi-stage features to obtain fused features;
[0113] Decoding the fused features to obtain a probability feature map and a threshold feature map, and fusing the probability feature map and the threshold feature map to obtain a binary feature map;
[0114] Calculating the loss functions of the probability feature map, the threshold feature map, and the binary feature map respectively to obtain a probability map loss, a threshold map loss, and a binary map loss;
[0115] The probability map loss, the threshold map loss and the binary map loss are used to perform model training, and the text area polygon information corresponding to the wafer surface image is obtained according to the trained text detection model.
[0116] Using the backbone network to extract features can obtain multi-stage features containing rich image information, which is convenient for subsequent feature analysis. Using the DBFPN neck network structure for feature fusion can realize features at different stages, enhance the robustness and diversity of feature representation, and enable the model to capture text information of different scales at the same time. Through decoding, a probability feature map and a threshold feature map can be obtained. The probability feature map is used to represent the probability that each pixel in the image belongs to the text area; the threshold feature map is a map of the same size as the probability feature map, which is the threshold that should be used when binarizing each pixel position in the probability feature map. Based on the probability feature map and threshold feature map obtained by decoding, an approximate binary feature map can be calculated through a differentiable binarization operation, so that the entire network can be trained end-to-end. Finally, the probability map loss, threshold map loss and binary map loss are used to train the model, which can ensure that the boundaries of the text area in the wafer surface image are accurately predicted, and the text area polygon information containing character position information is obtained.
[0117] In this embodiment, when the backbone network is used to extract features from the wafer surface image, the scale parameter of the backbone network (the parameter used to adjust the network width, also referred to as the parameter used to adjust the number of channels) is adjusted from 1x to 0.5x, while ensuring accuracy and efficiency, the number of feature map channels is reduced and the network structure is simplified. At the same time, due to the SE module (Squeeze-and-Excitation) module of the network, which is an attention mechanism for enhancing the performance of convolutional neural networks (CNNs)), when extracting features from large-resolution images, the accuracy improvement is very limited, but the time cost is very high. Therefore, in the actual process, the SE module of the network is also omitted, which can reduce the overall running time of the model.
[0118] In this embodiment, when the DBFPN (Deep Bilateral Feature Pyramid Network) neck network structure is used to fuse multi-stage features, inner_channels (it refers to the use of 1×1 convolution to reduce the feature map to the same number of channels in order to facilitate the merging of feature maps of different channels during the feature fusion process) is reduced from 256 to 96, thereby reducing the size of the model while ensuring accuracy.
[0119] Specifically, the OCR recognition component performs text recognition on the text area rectangular frame based on the SVTR text recognition method, obtains the ID number of the wafer to be inspected and outputs it, specifically including:
[0120] Input the text area rectangular frame into a preset text recognition model to obtain a target text area;
[0121] Processing the target text region using a progressive overlapping patch embedding method to obtain character components in multiple formats;
[0122] All character components are gradually downsampled in height, and a parallel linear prediction method is used to decode all character components after the gradual downsampling in height, so as to obtain and output the ID number of the wafer to be inspected.
[0123] The preset text recognition model can select a suitable model according to the actual situation, such as a model based on a convolutional neural network (CNN), which can identify the target text area, which can be a word, line or paragraph; then the progressive overlapping patch embedding method is used for processing, and the identified target text area can be divided into smaller blocks (patches), which may overlap to capture more context information, and each patch may contain part of the character or character components (such as strokes, line segments, etc.), providing a basis for subsequent recognition steps. Then, before converting the representation of the character components into the final character sequence, these representations are highly gradually downsampled. This process helps to convert character components or patches of different sizes into a uniform size for subsequent parallel processing; finally, a parallel linear predictor (such as CTC-Loss optimized RNN, Transformer, etc.) is used to decode the unified representation and realize parallel linear prediction. Based on the feature representation of the character components, the most likely character sequence can be predicted, and then the ID number of the final wafer to be inspected can be obtained and output. The entire text recognition process is accurate and reliable, and the obtained ID number has a high accuracy rate.
[0124] Specifically, in this embodiment, the text area rectangular frame is input into the preset text recognition model, and an image in the H×W×3 format can be output. Multiple images are obtained by progressive overlapping patch embedding. Character components (CharacterComponents).
[0125] Before all character components are gradually downsampled, the hybrid blocks of each output in the network model are used to extract, fuse, merge or combine features at different scales to form a unified representation of the character components. These uniformly represented character components are then gradually downsampled, and character components or patches of different sizes are converted into of uniform size.
[0126] Specifically, in the text recognition process based on the SVTR text recognition method in this embodiment, the SVTR_LCNet network is used, and all ReLU (Rectified Linear Unit) activation functions in the entire SVTR_LCNet network are replaced with h-swish activation functions. The h-swish activation function combines the advantages of Sigmoid and linear functions, and can activate negative inputs more smoothly, which may bring better gradient flow and model learning ability. After replacing it with h-swish, the model may be able to capture more nonlinear features, thereby improving the overall accuracy by 1% to 2%. In the fifth stage of the SVTR_LCNet network, the kernel size of the depthwise separable convolution (DW) is changed from the default size (i.e., 3x3) to 5x5, which can capture more complex spatial features, help the model better understand global features, and thus improve the overall accuracy by 0.5% to 1%. At the same time, adding SE (Squeeze-and-Excitation) modules to the last two Squeeze-and-Excitation (SEP) blocks in the fifth stage of the SVTR_LCNet network can enhance the model's ability to judge the importance of features, enhance important features and suppress unimportant features, thereby improving the overall accuracy by 0.5% to 1%. In the SVTR_LCNet network, a 1280-dimensional fully connected (FC) layer is added after the global average pooling (GAP) layer. The global average pooling layer is usually used to reduce the spatial dimension of the feature map to 1x1, thereby retaining the most important features; adding a high-dimensional fully connected layer can further increase the model's nonlinear expression ability and feature fusion ability. This additional FC layer can help the model learn more complex feature combinations, thereby improving the accuracy of classification or regression tasks, thereby improving the overall accuracy by 2% to 3%.
[0127] Preferably, if Figure 2 As shown, the control system also includes a second PWM control unit and a fill light;
[0128] The ARM processor is electrically connected to the fill light through the second PWM control unit;
[0129] The ARM processor is further configured to send a fill light instruction to the second PWM control unit while sending the wafer acquisition instruction to the image sensor;
[0130] The second PWM control unit is used for generating a second PWM control signal according to the received fill light instruction, and sending the second PWM control signal to the fill light;
[0131] The fill light is used to emit a fill light source to the wafer to be inspected under the control of the second PWM control signal, so as to provide fill light to the surface of the wafer to be inspected.
[0132] The control system of this embodiment, while collecting the wafer surface image, also triggers the fill light to emit a fill light source through the second PWM control unit to fill the surface of the wafer to be inspected, which can ensure the quality of the collected wafer surface image, thereby enabling the entire control system to adapt to different usage environments on the one hand, and on the other hand, based on the high-quality wafer surface image, effectively improving the subsequent recognition accuracy of the wafer ID number. At the same time, similar to the first PWM control unit, the second PWM control unit is used to control the fill light, which can achieve high-precision control of the triggering of the fill light source.
[0133] In this embodiment, the second PWM control unit may be the same as or different from the first PWM control unit, the pulse width of the first PWM control signal used to control the laser emitter and the pulse width of the second PWM control signal used to control the fill light may be the same as or different, and the two do not interfere with each other, and independent control of the laser emitter and the fill light can be achieved respectively.
[0134] Preferably, if Figure 2 As shown, the control system also includes a stage control component;
[0135] The stage control component is electrically connected to the ARM processor, and the stage control component is also connected to a stage driver for placing the wafer to be inspected;
[0136] The ARM processor is also used to send a stage motion instruction to the stage control component before sending a wafer image acquisition instruction to the image sensor and before the FPGA controller sends a laser emission instruction to the laser emitter;
[0137] The stage control component is used to control the rotation of the stage on which the wafer to be inspected is placed according to the received stage movement instruction.
[0138] Before triggering the emission of parallel line laser signals and before triggering the wafer surface image acquisition, the ARM processor is used to control the operation of the stage control component to drive the rotation of the stage on which the wafer to be inspected is placed. This ensures that the subsequently acquired wafer surface images are real-time, and at the same time ensures that the subsequently obtained edge position electrical signals are at different positions of the edge of the wafer to be inspected, thereby ensuring that the final ID number and wafer offset data are accurate, and ensuring the stability and reliability of the control system that integrates the wafer edge positioning function with the ID number recognition function.
[0139] It should be understood that the stage of this embodiment is a common device in wafer production, and the specific structural composition is not listed here. The edge-finding motion module is specifically a motion module configured for the stage that can be configured in three dimensions: X (translation on the x-axis), Y (translation on the y-axis), and R (rotation in the xy plane), and the stage control component is specifically a motor module that can control the rotation of the stage in the xy plane.
[0140] The photoelectric sensor is specifically a linear array CCD sensor; the image sensor is specifically a CCD area array camera, which is connected to the ARM chip through the MIPI-CSI interface, and the edge-finding motion module is connected to the ARM chip through the Ethernet communication interface; the fill light is specifically an LED light source; the above components as well as the FPGA controller and the ARM processor can all choose appropriate specifications or product models according to actual conditions, and the specific details are not listed here.
[0141] Embodiment 2
[0142] A control method for wafer edge finding and ID number recognition is provided, wherein the control system for wafer edge finding and ID number recognition in the first embodiment is used to perform wafer edge finding and ID number recognition, such as Figure 3 As shown, the control method includes:
[0143] S1: Use the FPGA controller to send a laser emission instruction to the laser emitter;
[0144] S2: using a laser transmitter to send a parallel line laser signal to the wafer to be inspected according to the received laser emission instruction;
[0145] S3: using a photoelectric sensor to collect an edge position electrical signal generated after the parallel line laser signal passes through the edge of the wafer to be inspected, and transmitting the edge position electrical signal to the FPGA controller;
[0146] S4: using the FPGA controller to transmit the received edge position electrical signal to the ARM processor;
[0147] S5: using the ARM processor to calculate wafer offset data according to the received edge position electrical signal, and controlling the edge finding motion module to move according to the wafer offset data, so as to perform edge finding and positioning on the wafer to be inspected;
[0148] S6: Using the ARM processor, sending a wafer image acquisition instruction to the image sensor;
[0149] S7: using the image sensor to obtain the wafer surface image of the wafer to be inspected according to the received image acquisition instruction, and transmitting the wafer surface image to the OCR recognition component through the ARM processor;
[0150] S8: using the OCR recognition component to identify and output the ID number of the wafer to be inspected based on the received wafer surface image.
[0151] In this embodiment, the integration between the laser emitter, the photoelectric sensor and the FPGA controller is utilized to obtain the edge position electrical signal generated after the parallel line laser signal passes through the edge of the wafer to be inspected, and then through the data interaction between the FPGA controller and the ARM processor, the ARM processor is used to obtain the wafer offset data according to the edge position electrical signal, and then the edge-finding motion module is controlled to move according to the wafer offset data to achieve edge-finding positioning of the wafer; then the integration between the image sensor, the OCR recognition component and the ARM processor is utilized to obtain the wafer surface image, and the ID number of the wafer to be inspected is obtained and output according to the acquired wafer surface image recognition, so as to realize the identification of the wafer ID number, and then ensure the identification and alignment requirements of the cutting edge position and the management requirements of the wafer transmission when the wafer flows between different process equipment.
[0152] The control method for wafer edge finding and positioning and ID number recognition in this embodiment realizes the integrated design of wafer edge finding and positioning function and ID number recognition function, greatly improves the integration of wafer positioning equipment, reduces equipment development cost, occupies a small space, has high working efficiency, will not cause damage to the wafer by the clamping action of the fixture, and has high stability.
[0153] It should be understood that the wafer edge finding and positioning described in steps S1 to S5 of this embodiment and the wafer ID number identification described in steps S6 to S8 may be performed simultaneously or sequentially, and this embodiment does not limit this.
[0154] The control system for wafer edge finding and ID number recognition used in the control method described in this embodiment has the same structure as the control system for wafer edge finding and ID number recognition described in Example 1. For details not included in this embodiment, please refer to Example 1 and Figure 1 to Figure 2 The detailed description will not be repeated here.
[0155] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.
Claims
1. A control system for wafer edge positioning and ID number recognition, It is characterized in that The control system includes an FPGA controller, a laser transmitter, a photoelectric sensor, an ARM processor, an image sensor, an OCR recognition component and an edge-finding motion module; The laser emitter and the photoelectric sensor are both electrically connected to the FPGA controller, the FPGA controller is communicatively connected to the ARM processor, and the image sensor, the OCR recognition component and the edge-finding motion module are all electrically connected to the ARM processor; The FPGA controller is used to send a laser emission instruction to the laser emitter; The laser transmitter is used to send a parallel line laser signal to the wafer to be inspected according to the received laser emission instruction; The photoelectric sensor is used to collect the edge position electrical signal generated after the parallel line laser signal passes through the edge of the wafer to be inspected, and transmit the edge position electrical signal to the FPGA controller; The FPGA controller is used to transmit the received edge position electrical signal to the ARM processor; The ARM processor is used to calculate the wafer offset data according to the received edge position electrical signal, and control the edge finding motion module to move according to the wafer offset data, so as to perform edge finding and positioning on the wafer to be inspected; The ARM processor is also used to send a wafer image acquisition instruction to the image sensor; The image sensor is used to obtain a wafer surface image of the wafer to be inspected according to the received image acquisition instruction, and transmit the wafer surface image to the OCR recognition component through the ARM processor; The OCR recognition component is used to identify and output the ID number of the wafer to be inspected based on the received wafer surface image.
2. The control system for wafer edge positioning and ID number recognition according to claim 1, It is characterized in that The control system further comprises a first PWM control unit; The FPGA controller is electrically connected to the laser transmitter via the first PWM control unit; The FPGA controller is specifically used for: Sending the laser emission instruction to the first PWM control unit; The first PWM control unit is used to generate a first PWM control signal according to the received laser emission instruction, and send the first PWM control signal to the laser emitter; The laser transmitter is specifically used for: Under the control of the first PWM control signal, the parallel line laser signal is sent to the wafer to be inspected.
3. The control system for wafer edge positioning and ID number recognition according to claim 1, It is characterized in that The control system also includes an ADC converter; The photoelectric sensor is electrically connected to the FPGA controller through the ADC converter; The ADC converter is used to convert the analog edge position electrical signal collected by the photoelectric sensor into an edge position digital signal, and transmit the edge position digital signal to the FPGA controller, and through the FPGA controller, transmit the edge position digital signal to the ARM processor.
4. The control system for wafer edge positioning and ID number recognition according to claim 3, It is characterized in that The wafer offset data includes the center offset position and the cutting edge offset angle of the wafer to be inspected; The ARM processor calculates the wafer offset data according to the received edge position electrical signal, specifically including: Calculate the edge position coordinates of the wafer to be inspected according to the edge position digital signal transmitted by the FPGA controller; Using the least square method, curve fitting is performed according to the edge position coordinates to obtain the edge curve of the wafer to be inspected; The center offset position and the trimming offset angle of the wafer to be inspected are calculated based on the edge curve.
5. The control system for wafer edge positioning and ID number recognition according to claim 1, It is characterized in that The OCR recognition component identifies and outputs the ID number of the wafer to be inspected based on the received wafer surface image, specifically including: Based on the DB optimization method, text detection is performed on the wafer surface image to obtain polygonal information of the text area; Performing clipping and perspective transformation processing on the polygonal information of the text area to obtain a rectangular frame of the text area; Based on the SVTR text recognition method, text recognition is performed on the text area rectangular frame to obtain the ID number of the wafer to be inspected and output it.
6. The control system for wafer edge positioning and ID number recognition according to claim 5, It is characterized in that The OCR recognition component performs text detection on the wafer surface image based on the DB optimization method to obtain text area polygon information, specifically including: Using a backbone network to extract features from the wafer surface image to obtain multi-stage features; The DBFPN neck network structure is used to fuse the multi-stage features to obtain fused features; Decoding the fused features to obtain a probability feature map and a threshold feature map, and fusing the probability feature map and the threshold feature map to obtain a binary feature map; Calculating the loss functions of the probability feature map, the threshold feature map, and the binary feature map respectively to obtain a probability map loss, a threshold map loss, and a binary map loss; The probability map loss, the threshold map loss and the binary map loss are used to perform model training, and the text area polygon information corresponding to the wafer surface image is obtained according to the trained text detection model.
7. The control system for wafer edge positioning and ID number recognition according to claim 5, It is characterized in that The OCR recognition component performs text recognition on the text area rectangular frame based on the SVTR text recognition method, obtains the ID number of the wafer to be inspected and outputs it, specifically including: Input the text area rectangular frame into a preset text recognition model to obtain a target text area; Processing the target text region using a progressive overlapping patch embedding method to obtain character components in multiple formats; All character components are gradually downsampled in height, and a parallel linear prediction method is used to decode all character components after the gradual downsampling in height, so as to obtain and output the ID number of the wafer to be inspected.
8. The control system for wafer edge positioning and ID number recognition according to claim 1, It is characterized in that The control system also includes a second PWM control unit and a fill light; The ARM processor is electrically connected to the fill light through the second PWM control unit; The ARM processor is further configured to send a fill light instruction to the second PWM control unit while sending the wafer acquisition instruction to the image sensor; The second PWM control unit is used for generating a second PWM control signal according to the received fill light instruction, and sending the second PWM control signal to the fill light; The fill light is used to emit a fill light source to the wafer to be inspected under the control of the second PWM control signal, so as to provide fill light to the surface of the wafer to be inspected.
9. The control system for wafer edge positioning and ID number recognition according to any one of claims 1 to 8, It is characterized in that The control system also includes a stage control component; The stage control component is electrically connected to the ARM processor, and the stage control component is also connected to a stage driver for placing the wafer to be inspected; The ARM processor is also used to send a stage motion instruction to the stage control component before sending a wafer image acquisition instruction to the image sensor and before the FPGA controller sends a laser emission instruction to the laser emitter; The stage control component is used to control the rotation of the stage on which the wafer to be inspected is placed according to the received stage movement instruction.
10. A control method for wafer edge positioning and ID number recognition, It is characterized in that The wafer edge finding and ID number identification control system according to any one of claims 1 to 9 is used to perform wafer edge finding and ID number identification, and the control method includes: Using the FPGA controller, send laser emission instructions to the laser transmitter; Using a laser transmitter, according to the received laser emission instruction, a parallel line laser signal is sent to the wafer to be inspected; Using a photoelectric sensor, collecting an edge position electrical signal generated after the parallel line laser signal passes through the edge of the wafer to be inspected, and transmitting the edge position electrical signal to the FPGA controller; Utilizing the FPGA controller, transmitting the received edge position electrical signal to the ARM processor; Using the ARM processor, according to the received edge position electrical signal, the wafer offset data is calculated, and according to the wafer offset data, the edge finding motion module is controlled to move, so as to perform edge finding and positioning on the wafer to be inspected; Using the ARM processor, sending a wafer image acquisition instruction to an image sensor; Using the image sensor, according to the received image acquisition instruction, a wafer surface image of the wafer to be inspected is acquired, and the wafer surface image is transmitted to an OCR recognition component through the ARM processor; The OCR recognition component is used to identify and output the ID number of the wafer to be inspected based on the received wafer surface image.
Citation Information
Patent Citations
Positioning method, positioning apparatus and positioning circuit system of wafer pre-alignment system
CN107768294A
Semiconductor wafer, method of identifying semiconductor wafer and semiconductor wafer identifying apparatus using the method
JP2002367871A
Front surface and back surface orientation detection of transparent substrate
TW202209521A
Substrate edge exposure apparatus
US20070267582A1
