TWI782539B - Intelligent processing method and system - Google Patents

Abstract

The present invention provides an intelligent processing method and system to integrate processing, detection and calibration processing equipment on the production line. The system includes a marking device, a coarse judgment device, a fine judgment device, a processing information acquisition device, and a measurement/parameter calibration device. The above-mentioned marking device is used for marking processing information on the processed products. The coarse judging device is an automatic vision detection device implemented by an optical inspection device, which can preliminarily screen out defective processed products. The above-mentioned fine judgment device uses artificial intelligence to identify the types of the above-mentioned defects. The processing information acquisition device is used to identify the processing parameters of the processing information. The measurement/parameter calibration device can measure the size of the defect, and calculate the best result of adjusting the processing parameter according to the size of the defect and the processing parameters.

Description

Method and system for intelligent processing

The present invention relates to the optimization of an automated processing flow, in particular to a method and system for intelligent processing.

The production process is closely linked from processing, testing, and calibration of processing machines. Any mistake in any step will eventually lead to problems such as insufficient product yield and reduced production capacity. In the past, the above-mentioned production process was often divided into different specialized departments, and it was easy to waste a lot of time and cost in manual communication or cooperation. For example: Although automatic optical inspection of products can replace manual general inspection, when defective products are found, manual re-inspection is still required to measure and identify defects, and then submit defect data to the production unit. Then correct the parameters of the production machine. The above-mentioned process is not only incoherent and time-consuming, but also in most cases, it is impossible to obtain the ideal yield immediately by calibrating the parameters of the machine once, and actually requires multiple calibrations.

With the continuous development of technology, the term "Industry 4.0" was proposed in 2011. The main spirit is to automate the production process intelligently and actively eliminate production problems.

Therefore, in order to solve the above problems, the field should develop processing methods and systems in line with the spirit of "Industry 4.0" to improve the efficiency of the production process and continuously improve the yield of products.

In order to solve the above problems, the embodiment of the present invention develops a method and system for intelligent processing. The above system integrates technologies such as product traceability, optical instrument initial inspection, artificial intelligence-based re-inspection, virtual measurement, and automatic parameter optimization. The applicant of the present invention refers to the patents filed in the past, including: TW109132726, TW109124011, TW110107077, TW109212512 and TW109136592, as a reference for the technology of the present invention.

Specifically, an embodiment of the present invention provides an intelligent processing system, the above-mentioned intelligent processing system is applied to at least one processing device, and the above-mentioned processing device processes the object to be processed according to the processing parameters and generates the object to be processed. The above-mentioned intelligent processing system includes at least one marking device, a rough judgment device, a fine judgment device, a processing information acquisition device, and a measurement/parameter correction device.

According to yet another embodiment, the above-mentioned marking device marks processing information on the above-mentioned workpiece, wherein the above-mentioned processing information includes at least one of the above-mentioned processing parameters. The recording methods of the processing information of the above-mentioned marking device include characters, numbers, symbols and image codes, wherein the above-mentioned image codes include one-dimensional barcodes and two-dimensional barcodes. The processing information of the above-mentioned marking device further includes the manufacturing batch, product model and manufacturing date.

According to yet another embodiment, the rough judging device is for initially checking whether the processed object has at least one defect and generating at least one image of the processed object, wherein the rough judging device is an automatic visual inspection device. The preliminary inspection method of the automatic visual inspection device includes contour comparison, position coordinate comparison, 3D contour comparison, color luster comparison and brightness comparison. The above-mentioned automatic visual detection device comprises at least one imaging unit, at least one distance measuring unit, a computing unit and a database. The above distance measuring unit is used to generate image capturing distance information for capturing the processed object. The above-mentioned database is for storing multiple standard size information, and each of the above-mentioned standard size information includes the analysis of the unit area degree, imaging distance, corresponding pixel matrix and corresponding actual size and film thickness. The calculation unit compares the standard size information in the database according to the resolution information of the processed object image and the imaging distance information, and calculates and generates the processed object image with actual size information. The above-mentioned standard size information in the above-mentioned database can be, for example, a two-dimensional size or a three-dimensional size, and the above-mentioned actual size with a two-dimensional size or a three-dimensional size is further calculated by the above-mentioned calculation unit. The image of the workpiece can be 2D or 3D.

According to yet another embodiment, the above-mentioned fine judging device is connected to the above-mentioned rough judging device for re-checking whether the image of the processed object has at least one defect, and identifying the type of defect contained in the image of the processed object, and generating Identification results. The fine-judgment device further includes a classification module, and the classification module uses artificial neural networks to identify the types of defects. The training data of the above-mentioned classification module includes the above-mentioned recognition results and manual re-judgment results.

According to yet another embodiment, the above-mentioned processing information acquisition device is connected to the above-mentioned precise judging device for identifying the above-mentioned processing information of the above-mentioned processed object image, and generating the above-mentioned processing parameters.

According to yet another embodiment, the above-mentioned measurement/parameter correction device is connected to the above-mentioned fine-judgment device for measuring the defect of the image of the processed object, and generates the size of the defect, and then calculates according to the size of the defect and the above-mentioned processing parameters. The optimal result of the above-mentioned processing parameters, and then generate the corrected processing parameters, and transmit the above-mentioned corrected processing parameters to the above-mentioned processing device. The above-mentioned measurement/parameter correction device further includes a parameter optimization module, and the above-mentioned parameter optimization module is to apply an algorithm to calculate the best result of the above-mentioned processing parameters. The above-mentioned flaw size may be, for example, a two-dimensional size or a three-dimensional size.

According to yet another embodiment, the above-mentioned measurement/parameter correction device, when the above-mentioned processed object has a design drawing, can further superimpose the above-mentioned design drawing and the above-mentioned processed object image to generate a superimposed image as a measure of the The basis for the size of the defect.

According to yet another embodiment, the above measurement/parameter correction device further includes providing at least one monitoring measurement area in the above design drawing in advance, and then measuring the above monitoring measurement area of the above superimposed image, and generating monitoring Size information. The above-mentioned monitoring measurement area can be regarded as a region of interest or a key region (Region of interest; ROI for short) in terms of image processing.

According to yet another embodiment, the above measurement/parameter correction device can further generate monitoring statistical information through statistics of the above monitoring size information, and the above monitoring statistical information will be used as calculation data for calculating the above processing parameters. The above-mentioned monitoring size information can be, for example, two-dimensional or three-dimensional information.

According to yet another embodiment, when processing according to the above-mentioned processing parameters, the production yield corresponding to the above-mentioned processing parameters can be further calculated through the number of defects found by the above-mentioned precise judging device.

According to yet another embodiment, the qualification standard of the above-mentioned production yield rate can be adjusted according to the needs of users. If the yield rate generated by the above-mentioned processing parameters meets the above-mentioned standard, the above-mentioned processing parameters can be further used as the implementation of virtual metrology (Virtual Metrology; referred to as VM) reference parameters. The above-mentioned virtual measurement refers to estimating the production result (or quality) through processing parameters to achieve the purpose of full inspection.

According to yet another embodiment, the measurement/parameter correction device stops automatically correcting the processing parameters and generates a warning notification when the defect size exceeds a preset threshold.

According to yet another embodiment, upon receiving the warning notification, the above-mentioned fine judgment device will not use the identification result corresponding to the warning notification as the training data for training the classification module.

According to yet another embodiment, the above-mentioned warning notification can be notified to the system management personnel in real time through text messages or emails.

Embodiments of the present invention further provide an intelligent processing method, including the following steps. Preset the processing parameters in the processing device; process the object to be processed and produce the processed object; mark the processing information on the above-mentioned processed object, the above-mentioned processing information includes the above-mentioned processing parameters; perform a preliminary inspection on the above-mentioned processed object with the rough judgment device , and capture at least one processed object image and imaging distance information, wherein the rough judging device can be an automatic visual inspection device; and then compare the above-mentioned imaging distance with a plurality of standard size information contained in the rough judging device Information, to generate the actual size information of the image of the processed object; re-inspect with the precise judgment device, and use the classification module to identify the type of defects contained in the image of the processed object, and generate the identification result, wherein the classification module is Applying the artificial neural network to perform the above-mentioned identification of defect types; superimposing the above-mentioned design drawing of the processed object with the above-mentioned image of the processed object to generate a superimposed image, and then performing measurement of the defect on the above-mentioned superimposed image to generate the size of the defect; Identify the above-mentioned processing information and generate the above-mentioned processing parameters; use the measurement/parameter correction device through the parameter optimization module to calculate the best result of the processing parameters based on the above-mentioned defect size and the above-mentioned processing parameters, and generate corrected processing parameters, and transmit them to The above processing device, wherein the above parameter optimization module uses an artificial neural network to optimize processing parameters.

According to yet another embodiment, the above-mentioned method further includes the step of quality monitoring, which is to set at least one monitoring measurement area in the design drawing of the above-mentioned processed object in the above-mentioned measurement/parameter correction device in advance, and then perform the above-mentioned overlapping The above-mentioned monitoring measurement area of the image is measured, and the monitoring size information is generated.

According to yet another embodiment, according to the above-mentioned measurement/parameter calibration device, the above-mentioned monitoring size information can be further calculated to generate monitoring statistical information, and the above-mentioned monitoring statistical information will be used as calculation data for calculating the above-mentioned processing parameters.

According to yet another embodiment, when processing according to the above-mentioned method and the above-mentioned processing parameters, the production yield corresponding to the above-mentioned processing parameters is further calculated through the number of defects found by the above-mentioned precise judging device.

According to yet another embodiment, according to the above-mentioned method, the qualification standard of the above-mentioned production yield rate can be adjusted according to the needs of users. If the yield rate produced by the above-mentioned processing parameters meets the above-mentioned standard, the above-mentioned processing parameters can be further used as a virtual measurement (Virtual Measurement) Metrology; referred to as VM) reference parameters. The above-mentioned virtual measurement refers to estimating the production result (or quality) through processing parameters to achieve the purpose of full inspection.

According to yet another embodiment, the training data of the above-mentioned classification module according to the above-mentioned method includes the above-mentioned recognition results and manual re-judgment results.

According to yet another embodiment, when the size of the defect according to the above method exceeds a threshold, the measurement/parameter correction device will stop automatically correcting the processing parameters and generate a warning notification.

According to yet another embodiment, upon receiving the warning notification, the fine judgment device according to the above method will not use the identification result corresponding to the warning notification as the training data for training the classification module.

According to yet another embodiment, the preliminary detection method of the rough judgment device according to the above method includes 3D contour comparison, position coordinate comparison, color luster comparison and brightness comparison.

According to yet another embodiment, the plurality of standard size information according to the above method respectively include resolution per unit area, imaging distance, corresponding pixel matrix, and corresponding actual size and film thickness.

According to yet another embodiment, the recording methods of the processing information according to the above method include characters, numbers, symbols and image codes, wherein the image codes include one-dimensional barcodes and two-dimensional barcodes.

Based on the technical features of the above-mentioned embodiments, the following effects can be specifically claimed.

(1) This system is applied in the processing process to achieve the effect of intelligence and full automation. It can also connect the processing equipment of several processing stations in series at the same time, so as to automatically perform the initial judgment of the processed objects of the processing equipment of each station. and re-judgment detection.

(2) For products processed through this system, the processing information on the product can be used to trace the processing parameters of each site in real time, which can quickly clarify which sites cause the insufficient yield rate of the product.

(3) The system can automatically classify the type of defect and perform measurement on the defect based on the above-mentioned preliminary judgment and re-judgment detection results, and through the artificial neural network, and obtain information on the size of the defect.

(4) After collecting the above-mentioned processing parameters and defect sizes, the system can further calculate the optimal result of the above-mentioned processing parameters through an algorithm, and generate corrected processing parameters, and then send the corrected processing parameters back to the above-mentioned processing equipment to achieve real-time The effect of optimizing processing parameters.

100: Intelligent processing system

110: Processing device

120: Marking device

130: Rough judgment device

140: Precise Judgment Device

142: Classification module

160: Processing information acquisition device

180: Measurement/parameter correction device

182:Parameter optimization module

200: Items to be processed

220: processed object

300-322: Steps

402-416: Steps

In order to make the above and other objects, features, advantages and embodiments of the present invention more comprehensible, the accompanying drawings are described as follows:

FIG. 1 is a device schematic diagram of an intelligent processing system according to an embodiment of the present invention.

FIG. 2 is a schematic diagram of an intelligent processing system applied to multiple processing equipment according to an embodiment of the present invention.

FIG. 3 is a flowchart of an intelligent processing system according to an embodiment of the present invention.

FIG. 4 is a flow chart of quality monitoring of an intelligent processing system according to an embodiment of the present invention.

In order to describe various embodiments of the present invention in more detail, the following description is supplemented with accompanying drawings.

Please refer to FIG. 1 . FIG. 1 is a device schematic diagram of an intelligent processing system according to an embodiment of the present invention. In FIG. 1 , an intelligent processing system 100 is provided according to an embodiment. The above-mentioned intelligent processing system 100 is applied to at least one processing device 110. The above-mentioned processing device 110 processes the object to be processed 200 according to the processing parameters, and produces the object to be processed 220. The above-mentioned system 100 includes: a marking device 120, a rough judgment device 130, a fine Judgment device 140 , processing information acquisition device 160 and measurement/parameter correction device 180 . Each component in the above-mentioned intelligent processing system 100 can be implemented through software plus hardware, or through pure hardware, and the present invention is not limited thereto.

The marking device 120 marks processing information on the workpiece 220, wherein the processing information includes at least one processing parameter. The recording methods of the processing information of the above-mentioned marking device 120 include characters, numbers, symbols and image codes, wherein the above-mentioned image codes include one-dimensional barcodes and two-dimensional barcodes. The processing information of the above-mentioned marking device 120 further includes manufacturing batch, product model and manufacturing date. For example: print a two-dimensional image code on the finished or semi-finished printed circuit board, and provide a record of historical processing parameters by scanning or identifying the two-dimensional image code in the future.

The rough judging device 130 is for initially checking whether the processed object 220 has at least one flaw, and generates at least one image of the processed object, wherein the rough judging device is an automatic visual inspection device. The preliminary inspection method of the above-mentioned automatic visual inspection device includes 3D contour comparison, position coordinate comparison, color luster comparison and brightness comparison. The rough judging device 130 of this system is usually stricter in the initial judging standard, so Avoid ignoring products with real defects, but misjudgments often occur. For example, if there is fine dust somewhere in the printed circuit board, it is judged as a defect. Although this can ensure the yield of the product, it is relatively easy to identify false defects. In the past, the above-mentioned false defects could only be eliminated through manual re-judgment, but in the present invention, the re-judgment and defect classification of defects can be carried out through the above-mentioned precise judgment device 140, which can greatly reduce false defects and the need for manual re-judgment quantity.

According to another embodiment of the present invention, the above-mentioned automatic visual inspection device comprises at least one image capturing unit, at least one distance measuring unit, a computing unit and a database. The above distance measuring unit is used to generate image capturing distance information for capturing the processed object. The above-mentioned database is for storing a plurality of standard size information, and each of the above-mentioned standard size information includes the resolution per unit area, the imaging distance, the corresponding pixel matrix, and the corresponding actual size and film thickness. The calculation unit compares the standard size information in the database according to the resolution information of the image of the processed object 220 and the imaging distance information, and calculates and generates the image of the processed object with actual size information. The above-mentioned rough judgment device 130 can be, for example, a computer device (equipped with a CPU and a GPU) that cooperates with an image recognition device.

The fine judging device 140 is connected to the rough judging device 130 for rechecking whether the image of the processed object has at least one defect, identifying the type of defect contained in the image of the processed object, and generating a recognition result. The fine judging device further includes a classification module 142. The classification module 142 uses an artificial neural network to identify the type of the defect. The training data of the above-mentioned classification module 142 includes the above-mentioned recognition results and manual re-judgment results. The artificial neural network applied by the classification module 142 includes a convolutional neural network model (Convolutional Neural Network; CNN for short). The above-mentioned CNN model can be selected from one of R-CNN (Region-based Convolutional Neural Network), Fast R-CNN, Faster R-CNN, RPN (Region Proposal Network) and Mask R-CNN, FCN (Fully Convolutional Network) Identify and classify defects. In addition, the precise judging device 140 In addition to using the convolutional neural network model, other sophisticated algorithms for classifying components in images can also be used to implement the fine judgment device 140, and the present invention is not limited to the implementation of the fine judgment device 140 . The precise judgment device 140 can effectively reduce the number of false defects and the number of manual re-judgment. Whether manual re-judgment is required can be judged by the preset threshold value of the measurement/parameter correction device 180 . Taking the printed circuit board as an example, the above-mentioned false defects may include: false spots on the edge of the board, fine board scraps, dust, and differences between the sub-board to be tested and the mother board. The fine judgment device 140 mentioned above can be, for example, a computer (equipped with a CPU) with computing power.

The above-mentioned processing information acquisition device 160 is connected to the above-mentioned precise judging device 140 for identifying the above-mentioned processing information of the above-mentioned processed object image, and generating the above-mentioned processing parameters. The identification methods can be text identification, image code identification, and induction tags such as RFID. The above-mentioned processing information acquisition device 160 can be, for example, a computer device (equipped with a CPU and a GPU) that cooperates with an image recognition device.

The measurement/parameter correction device 180 is connected to the precise judgment device 140 for the purpose of measuring the defect of the image of the processed object and generating the size of the defect, and then calculating the processing parameters based on the size of the defect and the processing parameters. For the best result, corrected processing parameters are then generated, and the corrected processing parameters are transmitted to the processing device 110 . The measurement/parameter correction device 180 further includes a parameter optimization module 182. The parameter optimization module 182 uses an algorithm to calculate the best result of the above processing parameters. The algorithm applied by the above-mentioned parameter optimization module 182 can apply any algorithm that can calculate or summarize the best result of the above-mentioned processing parameters, such algorithm can be: gradient descent method, Newton method, conjugate gradient method, linear search , Confidence region method, neural network, particle swarm algorithm, analog annealing, support vector machine, ant colony algorithm, differential evolution algorithm, K-nearest neighbor algorithm (K-nearest neighbor). The measurement/parameter correction device 180 mentioned above can be, for example, a computer (equipped with a CPU) with computing power.

According to another embodiment of the present invention, the measurement/parameter correction device 180, when the workpiece 220 has a design drawing, can further superimpose the design drawing with the image of the workpiece 220, which can be used as a measurement The basis for the size of the above defects.

According to another embodiment of the present invention, the above-mentioned measurement/parameter correction device 180 further has a process of performing quality monitoring, which is to set at least one monitoring quantity in the design drawing of the above-mentioned processed object in the above-mentioned measurement/parameter correction device in advance Then measure the above-mentioned monitoring measurement area of the above-mentioned superimposed image, and generate monitoring size information. The above-mentioned quality monitoring process can be carried out in the form of general inspection or random inspection. Taking the detection of printed circuit boards as an example, the above-mentioned monitoring and measurement area can be frame-selected for specific soldering positions of the printed circuit board.

According to another embodiment of the present invention, the above-mentioned monitoring size information can be passed through the above-mentioned measurement/parameter correction device 180 to generate monitoring statistical information through statistics. And the above-mentioned monitoring statistical information will be used as the calculation data for calculating the above-mentioned processing parameters.

According to another embodiment of the present invention, when the size of the defect exceeds a threshold, the measurement/parameter correction device 180 will stop automatically correcting the processing parameters and generate a warning notification.

According to yet another embodiment, the fine judgment device 140 will not use the recognition result corresponding to the warning notification as the training data for training the classification module upon receiving the warning notification.

According to yet another embodiment, the above-mentioned warning notification can immediately notify the system management personnel through text messages or emails, and remind the above-mentioned system management personnel that the identification result corresponding to the above-mentioned warning notification will need manual re-judgment. For example, if it is found that the exposed nickel area on the printed circuit board is greater than 1.0cm ² , if the above-mentioned processing parameters are adjusted according to the defect size, it may be adjusted too much, which will seriously affect the subsequent processing flow. Therefore, the defect size exceeding the threshold is not used as a data source for adjusting the above-mentioned processing parameters and the above-mentioned training data.

According to yet another embodiment, the present invention can be applied in various stages of the processing process, and please refer to FIG. 2 . FIG. 2 is a device for applying an intelligent processing system to multiple processing equipment according to an embodiment of the present invention schematic diagram.

According to the embodiment of the present invention, please refer to FIG. 3 , which is a flowchart of an intelligent processing system according to an embodiment of the present invention.

In FIG. 3, step 300 is to start processing.

In step 302, the processing parameters of the processing device 110 are set.

In step 304 , the processing device 110 executes a processing operation and produces a processed object 220 .

In step 306 , mark the processing information on the workpiece 220 . The recording methods of the above-mentioned processing information include characters, numbers, symbols and image codes, wherein the image codes include one-dimensional barcodes and two-dimensional barcodes.

In step 308, the processed object 220 is initially judged by the rough judgment device 130, and if it is judged that the processed object 220 has at least one defect, the image of the processed object 220 is captured to generate at least one processed object image. The rough judging device 130 can further calculate the actual size information of the image of the processed object. The calculation method is as described above and will not be repeated here.

In step 310 , it is determined that the processed object 220 has at least one flaw through the rough judgment device 130 , and it is judged to be a flaw, and then step 312 is continued. If the rough judging device 130 preliminarily judges that no defect is found, go to step 322 .

In step 312, the image of the workpiece is re-judged by the precise judging device 140, and the type of defect is identified.

In step 314, after the fine judgment device 140 conducts a re-judgment to determine that the image of the workpiece does have at least one defect, the type of the defect is further identified. The method for identifying the defect is as described above and will not be repeated here. If the fine judging device 140 judges again that the image of the processed object is flawless, then skip to step 322, indicating that the above-mentioned flaw determined by the rough judging device 130 is a false flaw.

In step 316a, the measurement/parameter correction device 180 is used to measure the defect and generate the size of the defect. The method of the above measurement is as described above and will not be repeated here.

In step 316b, the processing information is identified through the processing information acquisition device 160, and the processing parameters are generated. The method for identifying the processing information is as described above, and will not be repeated here.

In step 318, the algorithm of the measurement/parameter correction device 180 calculates the best result of correcting the above-mentioned processing parameters based on the above-mentioned defect size and the above-mentioned processing parameters, generates the corrected processing parameters, and transmits the above-mentioned corrected processing parameters to the above-mentioned processing device .

In step 320, the processing equipment 110 adjusts the original processing parameters according to the above-mentioned corrected processing parameters, and jumps back to step 304 to continue processing.

In step 322, it is found that there is at least one defect on the processed product 220 in the preliminary judgment of the rough judging device 130 or the fine judgment of the fine judging device 140, so the processing continues.

According to another embodiment of the present invention, please refer to FIG. 4 , which is a flow chart of quality monitoring of an intelligent processing system according to an embodiment of the present invention.

The process of the embodiment in FIG. 3 of the present invention can cooperate with the quality monitoring process in the embodiment of FIG. 4 to increase the accuracy of calculating the above-mentioned corrected processing parameters, wherein the processes in the embodiments of FIG. 3 and FIG. 4 can be executed simultaneously or separately.

In step 402 , the measurement area to be monitored in the design drawing of the workpiece 220 is set (or framed) in the measurement/parameter correction device 180 in advance.

In step 404 , an image of the workpiece 220 is captured through the rough judgment device 130 .

In step 406 , the design drawing of the workpiece 220 is superimposed on the processed image to generate a superimposed image.

In step 408 , measure the size of the preset monitoring measurement area on the superimposed image.

In step 410, monitor size information is generated according to the measurement results, and monitor statistics information is further calculated.

In step 412, the monitoring statistical information can be used as the calculation data for the measurement/parameter calibration device 180 to calculate and correct the processing parameters.

According to another embodiment of the present invention, in step 414, the processing parameters adjusted in the above steps 402-412 can correspond to the number of defects found in the above steps 300-314 at the same time, and the production quality corresponding to the above processing parameters can be further calculated. Rate. It can also check whether the above-mentioned production yield rate meets the qualification standard set by the user (for example, the yield rate must reach 95.0%)

In step 416 , the above-mentioned processing parameters can be further passed through virtual metrology (Virtual Metrology, VM for short) to increase the above-mentioned production yield. For example: through virtual measurement and calculation, in the process of laying metal copper layer on the substrate when making printed circuit boards, reducing the time of copper sulfate bath electroplating by 0.5 seconds can reduce the probability of circuit short circuit by 1%. Therefore, adjusting production parameters through virtual measurement can further increase production yield. Through the above-mentioned virtual measurement, it is no longer necessary to measure the monitoring measurement area through the above-mentioned steps 402-412. Instead, the virtual measurement is used to calculate the yield rate of the product, so as to achieve the purpose of full inspection.

To sum up the above, the embodiment of the present invention is a method and system of intelligent processing, which has the processing and detection of real-time processing production line. And further implement traceability, defect determination and calibration of processing equipment with automation.

The present invention is only disclosed in preferred embodiments herein, but anyone skilled in the art should understand that the above embodiments are only used to describe the present invention, and are not intended to limit the scope of patent rights claimed by the present invention. All changes or substitutions that are equal or equivalent to the above-mentioned embodiments should be interpreted as falling within the spirit or scope of the present invention. Therefore, the scope of protection of the present invention should be defined by the scope of the following patent application.

100: Intelligent processing system

110: Processing device

120: Marking device

130: Rough judgment device

140: Precise Judgment Device

142: Classification module

160: Processing information acquisition device

180: Measurement/parameter correction device

182:Parameter optimization module

200: Items to be processed

220: processed object

Claims (23)

A method of intelligent processing, the method includes the following steps: preset a processing parameter in a processing device; process a to-be-processed object and generate a processed object; mark a processing information on the processed object, the The processing information includes the processing parameters; a preliminary inspection of the processed object is carried out with a rough judgment device, and at least one image of the processed object and an imaging distance information are captured, wherein the rough judgment device is an automatic visual detection device; The actual size information of the processed object image is generated by comparing the imaging distance information with the plurality of standard size information contained in the rough judgment device, wherein the rough judgment device is based on the resolution of the processed object image Information and the imaging distance information, compared with the standard size information, to calculate and generate the processed object image with actual size information; use a precise judgment device to perform re-inspection, and apply a classification module to identify the processed object image The type of defect contained in the classification module, and generate a recognition result, wherein the classification module uses artificial neural network to perform the recognition of the defect type, wherein the training data of the classification module includes the recognition result and a manual re-judgment result; Superimposing the design drawing of the processed object with the image of the processed object to generate a superimposed image, and then performing measurement of the defect on the superimposed image to generate a defect size; identifying the processing information to generate the processing parameters ; and use a measurement/parameter correction device to calculate the best result of the processing parameter based on the defect size and the processing parameter through a parameter optimization module, and generate a corrected processing parameter, and transmit it to the processing device, wherein The parameter optimization module uses algorithms to optimize processing parameters. Wherein, the measurement/parameter correction device includes a threshold value, and when the size of the defect exceeds the threshold value, the superimposed image corresponding to the size of the defect is regarded as an image requiring manual re-judgment, and the manual re-judgment result will be stored in the in the classification module. The intelligent processing method as described in claim 1, wherein the method further includes the step of quality monitoring, which is to set at least one monitoring measurement area in the design drawing of the processed object through the measurement/parameter correction device, and then The monitoring measurement area of the superimposed image is measured, and at least one monitoring size information is generated. The intelligent processing method as described in any one of request item 1 or request item 2, wherein the measurement/parameter correction device further includes the statistics of the monitoring size information to generate a monitoring statistical information, and the monitoring statistical information will be used as the calculation The calculation data of the processing parameters. The intelligent processing method as described in claim 3, wherein the processing parameter calculated through the monitoring statistical information is used as a reference parameter of a virtual measurement for performing the virtual measurement. In the intelligent processing method described in claim 1, when the defect size exceeds a threshold, the measurement/parameter calibration device will stop automatically correcting the processing parameters and generate a warning notification. In the intelligent processing method described in Claim 5, wherein the fine judgment device receives the warning notification, it will not use the identification result corresponding to the warning notification as the training data for training the classification module. The intelligent processing method as described in Claim 1, wherein the preliminary detection method of the rough judgment device includes 3D contour comparison, position coordinate comparison, color luster comparison and brightness comparison. The intelligent processing method as described in Claim 1, wherein the standard size information includes the resolution per unit area, the imaging distance, the corresponding pixel matrix, and the corresponding actual size and film thickness. The intelligent processing method as described in Claim 1, wherein the recording method of the processing information includes characters, numbers, symbols and image codes, wherein the image codes include one-dimensional barcodes and two-dimensional barcodes. An intelligent processing system, the system is applied to at least one processing device, and the processing device processes a to-be-processed object according to a processing parameter to generate a processed object, the system includes: at least one marking device, on the processed object Marking a processing information, wherein the processing information includes at least one processing parameter; a rough judgment device, used to initially check whether the processed object has at least one defect, and generate at least one processed object image, wherein the rough judgment device is It is an automatic visual inspection device, wherein the rough judging device is based on the resolution information of the processed object image and the imaging distance information, and compares the standard size information to calculate and generate the processed object with actual size information image; a fine judging device, connected to the rough judging device with a signal, for rechecking whether the image of the processed object has at least one defect, and identifying the type of defect contained in the image of the processed object to generate a recognition result, The fine-judgment device includes: a classification module, which uses artificial neural network to identify the defect type, wherein the training data of the classification module includes the recognition result and a manual re-judgment result; a processing information acquisition device, signal Connected to the fine judgment device to identify the processing information of the processed object image and generate the processing parameters; and a measurement/parameter correction device connected to the fine judgment device to measure the processed object image The defect, and generate a defect size, and then calculate the added value based on the defect size and the processing parameters The optimal result of the processing parameters, generating a corrected processing parameter, and transmitting the corrected processing parameter to the processing device, wherein, when the processed object has a design drawing, further superimposing the design drawing with the processed object image, Generating a superimposed image as a basis for measuring the size of the defect, wherein the measurement/parameter correction device includes a threshold, and when the size of the defect exceeds the threshold, the superimposed image corresponding to the size of the defect or The image of the processed object is an image that requires manual re-judgment, and the result of the manual re-judgment will be stored in the classification module. The intelligent processing system as described in Claim 10, wherein the initial inspection method of the automatic visual inspection device includes contour comparison, position coordinate comparison, color luster comparison and brightness comparison. The intelligent processing system as described in Claim 10, wherein the automatic visual inspection device comprises at least one imaging unit, at least one distance measuring unit, a computing unit and a database. In the intelligent processing system as described in Claim 10, wherein the distance measuring unit is used to generate imaging distance information for capturing the processed object. The intelligent processing system as described in claim 10, wherein the database is used to store a plurality of standard size information, and each of the standard size information includes the resolution per unit area, the imaging distance, and the corresponding pixel matrix And the corresponding actual size and film thickness height. The intelligent processing system as described in any one of claims 10 to 14, wherein the calculation unit compares the standard size information in the database according to the resolution information of the processed object image and the imaging distance information , calculate and generate the processed object image with actual size information. In the intelligent processing system as described in Claim 10, wherein the measurement/parameter calibration device further includes a parameter optimization module, the parameter optimization module calculates the best result of the processing parameter by applying an algorithm. The intelligent processing system as described in claim 10, wherein the measurement/parameter correction device further includes providing at least one monitoring measurement area in the design drawing in advance, and then performing the monitoring measurement area on the superimposed image Measure and generate at least one piece of monitoring size information. The intelligent processing system as described in claim 17, wherein the measurement/parameter correction device further includes the statistics of the monitoring dimension information to generate monitoring statistical information, and the monitoring statistical information will be used as calculation data for calculating the processing parameters. In the intelligent processing system as described in claim 18, wherein the processing parameter calculated through the monitoring statistical information is used as a reference parameter for virtual measurement to perform the virtual measurement. The intelligent processing system as described in Claim 10, wherein the recording method of the processing information of the marking device includes characters, numbers, symbols and image codes, wherein the image codes include one-dimensional barcodes and two-dimensional barcodes. The intelligent processing system as described in Claim 10, wherein the processing information of the marking device further includes manufacturing batch, product model and manufacturing date. In the intelligent processing system as described in Claim 10, when the defect size exceeds a threshold, the measurement/parameter calibration device will stop automatically correcting the processing parameters and generate a warning notification. In the intelligent processing system as described in claim 22, wherein the fine judgment device receives the warning notification, it will not use the identification result corresponding to the warning notification as the training data for training the classification module.

Images