KR102065893B1 - Method and apparatus for controlling wiper manufacturing process for cleanroom - Google Patents

Abstract

Disclosed are a method of controlling a wiper manufacturing process for a clean room, and an apparatus thereof. According to one embodiment of the present invention, an apparatus of controlling a wiper manufacturing process for a clean room is configured to: photograph each point in a wiper fabric for a clean room, desired to be manufactured, by using cameras corresponding to different constants; obtain first images photographed by the cameras; generate an input vector corresponding to an input layer of a convolutional neural network which has been previously learned to generate a panoramic image of a wiper cutting specification for a clean room from a plurality of images based on the first images; obtain an output vector by applying the input vector to the convolutional neural network; generate, based on the output vector, position control data to align photographing positions of the cameras in the formation based on the constants; align the cameras in the formation based on the position control data; obtain second images photographed by the aligned cameras; generate a target panoramic image of each point in a wiper fabric for a target clean room based on the second images; generate control information for manufacturing from information related to a wiper for a clean room; and control a packaging device. According to embodiments of the present invention, control information associated with a process of cutting, washing, drying and packing is estimated based on artificial intelligence to optimize process control.

Description

METHOOD AND APPARATUS FOR CONTROLLING WIPER MANUFACTURING PROCESS FOR CLEANROOM

The following examples relate to techniques for controlling the wiper manufacturing process for clean rooms.

Clean room refers to a workshop that removes fine dust and bacteria in the microscopic industry such as semiconductor devices, precision electronic products such as LCD, and genetic engineering. Clean room wipers are often used to clean the surface of equipment during manufacturing in clean rooms or to prevent the spread of liquids.

Therefore, in order to remove the contamination of the equipment as effectively as possible, clean room wipers should also be manufactured to minimize the contact of fine particles in the clean room environment. Therefore, it is important to maintain this environment as a whole, from cutting of fabric to washing, drying, and packaging, which has many difficulties. Therefore, there is a need for research of a technology for manufacturing a wiper for a clean room to maintain a minimum pollution state and to maximize the effect.

KR100908217 KR101933517 KR20000008677 KR1020080028925

Embodiments are intended to increase the precision of the cutting by generating an image of an object to be cut in a clean room wiper fabric based on artificial intelligence.

Embodiments aim to optimize process control by inferring control information related to the process of cutting, washing, drying, and packaging a clean room wiper based on artificial intelligence.

Embodiments aim to streamline the process steps through inter-facility feedback commands required for the process, manage locations and areas where errors frequently occur during the process through the feedback command, and improve the quality of the process.

Embodiments attempt to increase the precision of process equipment control by tying the information necessary for the manufacturing process of a wiper for a clean room.

According to one or more exemplary embodiments, a method for controlling a wiper manufacturing process for a clean room includes photographing points of a clean room wiper fabric to be manufactured using cameras corresponding to different constants; Obtaining first images photographed by the camera; Generating an input vector corresponding to an input layer of a convolutional neural network that has been previously learned to generate a panoramic image of a wiper cutting standard for a clean room from a plurality of images based on the first images. step; Applying the input vector to the convolutional neural network to obtain an output vector; Based on the output vector, generating position adjustment data for aligning the photographing positions of the cameras in formation based on the constants; Aligning the cameras in the large format based on the position adjustment data; Acquiring second images taken by the aligned cameras; Generating a target panoramic image of each of the points in the wiper fabric for the target clean room based on the second images; Obtaining an identifier corresponding to each of the points in the wiper fabric for the target clean room based on the target panoramic image; Obtaining a cutting type and manufacturing process of the target clean room wiper corresponding to the identifier from a database matching identifiers, cutting types of clean room wipers and manufacturing processes of the clean room wiper; Generating a first input signal based on the identifier, the cutting type of the target cleanroom wiper and the manufacturing process; Applying the first input signal to an input layer of a first learned neural network to generate control information for manufacturing from information associated with a clean room wiper; Obtaining a first output signal generated from an output layer of the first neural network; Generating first control commands based on the first output signal; Based on the first control instructions, controlling manufacturing process facilities; Generating a first loading signal for loading products according to a control result of the manufacturing process facilities into a cutting locker in a clean room; Controlling a first loading facility based on the first loading signal; Generating a second loading signal based on a signal of the self loading evaluation device in the cutout box; Controlling a second loading facility based on the second loading signal; Generating a second control command for simultaneously washing products according to the control result of the second loading facility; Controlling a laundry machine based on the second control command; Generating a third loading signal for transferring the products to a drying apparatus according to the control result of the washing apparatus; Controlling a third loading facility based on the third loading signal; Generating a third control command for simultaneously drying the products according to the control result of the third loading facility; Controlling a drying device based on the third control command; Generating a fourth loading signal for transferring the products to a conveyor belt according to a control result of the drying apparatus; Controlling a fourth loading facility based on the fourth loading signal; Generating a fourth control command for individually vacuum packaging the products on a conveyor belt according to the control result of the fourth loading facility; And controlling the packaging apparatus based on the fourth control command.

According to an embodiment, the output vector groups position and size data of portions of the first image photographed by one of the cameras, which overlap with the first images photographed by the other cameras, for each camera. A first overlapping partial data, wherein the large size is a large size in which the cameras respectively corresponding to the constants are sequentially arranged in the azimuthal direction of the reference position according to the order defined based on the constants; The position adjustment data is based on the first overlapping partial data, and acquires first relative positions of the cameras, and is photographed by adjacent cameras when the cameras align the photographing position in the large format. Portions overlapping each other in the images are taken to the second images taken by the adjacent cameras. Compute the second relative positions of the cameras to make the occupied position and size coincide with the predefined second overlapping partial data to control the cameras to move from the first relative positions to the second relative positions, respectively. Data.

According to one embodiment, the step of controlling the manufacturing process facilities is based on the first control commands, the structural specification information for manufacturing the wiper for the target clean room-the structural specification information is in the wiper fabric for the target clean room Structural specifications for positioning the cutting tool in space, including position coordinates, connection relationships, hierarchies, and depths, machine specification information, wherein the machine specification information includes physical, wiper fabric and cutting tools for the target cleanroom. Creating a specification according to chemical action and reaction, including size, length, shape, thickness, weight, material, proper pressure, proper temperature, specification and strength, pH, degree of oxidation; Generating test signals for manufacturing the wiper for the target clean room and time stamps for applying the test signals based on the structural specification information and the machine specification information; And controlling the manufacturing process facilities based on the sequence of test signals and the time stamps.

According to one embodiment, the controlling of the manufacturing process equipments may include querying a database that matches structural specification information, machine specification information, and cutting tools, such that the cutting tool corresponds to the structural specification information and the machine specification information. Controlling a first manufacturing process facility for obtaining the food; Generating a manufacturing process sequence condition corresponding to the structural specification information, the condition for placing the tools in a position that matches the sequence of the manufacturing process, based on the control result of the first manufacturing process facility; Controlling a second manufacturing process facility to place the tools based on the manufacturing process sequence conditions; Comparing the control result of the second manufacturing process facility with the manufacturing process sequence conditions to generate a feedback command for adjusting the placement of the tools; And controlling the second manufacturing process facility based on the generated feedback command.

According to one embodiment, the step of generating the manufacturing process sequence condition is based on a first history of tools clustered according to the structural specification information, the first of the obtained tools, a second tool, a third Generating a sequence of position instructions and time stamps for placing a tool and a fourth tool in a first position, a second position, a third position and a fourth position, respectively, in time series on the space in the wiper fabric for the object cleanroom step; And generating the manufacturing process sequence conditions based on the generated sequence.

The apparatus according to one embodiment may be controlled by a computer program stored in a medium in combination with hardware to carry out the method of any one of the aforementioned methods.

Embodiments may generate an image of an object to be cut in a clean room wiper fabric based on artificial intelligence to increase the precision of the cutting.

Embodiments may infer control information related to a process of cutting, washing, drying, and packaging a wiper for a clean room based on artificial intelligence to optimize process control.

Embodiments can streamline the process steps through inter-facility feedback commands required for the process, manage locations and areas where errors frequently occur during the process through the feedback command, and improve the quality of the process.

Embodiments can type information necessary for the manufacturing process of the wiper for a clean room to increase the precision of the process equipment control.

1 is a flowchart illustrating a method for controlling a manufacturing process of a wiper for a clean room according to an embodiment.
FIG. 2 is a diagram for describing a method of generating a panoramic image for cutting of a wiper for a clean room, according to an exemplary embodiment.
3 is a diagram for describing a method of generating a panoramic image for cutting of a wiper for a clean room, according to an exemplary embodiment.
4 is a diagram illustrating a learning process of a convolutional neural network employed to generate a panoramic image for cutting a wiper for a clean room, according to an exemplary embodiment.
5 is a diagram for describing a method for controlling a manufacturing process of a wiper for a clean room, according to an exemplary embodiment.
6 is a diagram for describing a method for controlling a manufacturing process of a wiper for a clean room, according to an exemplary embodiment.
7 is an exemplary diagram of a configuration of an apparatus according to an embodiment.

Hereinafter, exemplary embodiments will be described in detail with reference to the accompanying drawings. However, various changes may be made to the embodiments so that the scope of the patent application is not limited or limited by these embodiments. It is to be understood that all changes, equivalents, and substitutes for the embodiments are included in the scope of rights.

Specific structural or functional descriptions of the embodiments are disclosed for purposes of illustration only, and may be practiced in various forms. Accordingly, the embodiments are not limited to the specific disclosure, and the scope of the present specification includes changes, equivalents, or substitutes included in the technical idea.

Terms such as first or second may be used to describe various components, but such terms should be interpreted only for the purpose of distinguishing one component from another component. For example, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component.

When a component is referred to as being "connected" to another component, it should be understood that there may be a direct connection or connection to that other component, but there may be other components in between.

The terminology used herein is for the purpose of description and should not be construed as limiting. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "comprise" or "have" are intended to indicate that there is a feature, number, step, action, component, part, or combination thereof described on the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in the commonly used dictionaries should be construed as having meanings consistent with the meanings in the context of the related art and shall not be construed in ideal or excessively formal meanings unless expressly defined in this application. Do not.

In addition, in the description with reference to the accompanying drawings, the same components regardless of reference numerals will be given the same reference numerals and duplicate description thereof will be omitted. In the following description of the embodiment, when it is determined that the detailed description of the related known technology may unnecessarily obscure the gist of the embodiment, the detailed description thereof will be omitted.

Embodiments may be implemented in various forms of products, such as personal computers, laptop computers, tablet computers, smart phones, televisions, smart home appliances, intelligent cars, kiosks, wearable devices, and the like.

1 is a flowchart illustrating a method for controlling a manufacturing process of a wiper for a clean room according to an embodiment.

According to one embodiment, a control device (hereinafter, referred to as a control device) for a manufacturing process of a wiper for a clean room uses cameras corresponding to different constants of respective points in the fabric of the wiper for a clean room to be manufactured. It is possible to take a picture (101). The control device is a device for controlling the manufacturing process of the wiper for a clean room, and may be implemented, for example, as a software module, a hardware module, or a combination thereof.

The control device according to an embodiment may be an electronic device including a communication function. For example, the electronic device may be a smart phone, a tablet personal computer, a mobile phone, a video phone, an e-book reader, a desktop personal computer, a laptop. Laptop personal computer (PC), netbook computer, personal digital assistant (PDA), portable multimedia player (PMP), MP3 player, mobile medical device, camera, or wearable device (e.g., Include at least one of a head-mounted-device (HMD), such as electronic glasses, an electronic garment, an electronic bracelet, an electronic necklace, an electronic accessory, an electronic tattoo, a smart car, or a smartwatch. Can be.

According to an embodiment, the electronic device may be a smart home appliance having a communication function. Smart home appliances are, for example, electronic devices such as televisions, digital video disk (DVD) players, audio, refrigerators, air conditioners, cleaners, ovens, microwave ovens, washing machines, air purifiers, set-top boxes, TVs. It may include at least one of a box (eg, Samsung HomeSync ™, Apple TV ™, or Google TV ™), game consoles, electronic dictionaries, electronic keys, camcorders, or electronic photo frames.

According to an embodiment of the present disclosure, the electronic device may include various medical devices (for example, magnetic resonance angiography (MRA), magnetic resonance imaging (MRI), computed tomography (CT), an imaging device, an ultrasound device, etc.), a navigation device, and a GPS receiver ( global positioning system receiver (EDR), event data recorder (EDR), flight data recorder (FDR), automotive infotainment devices, marine electronic equipment (e.g. marine navigation systems and gyro compasses), avionics, security At least one of a device, a vehicle head unit, an industrial or household robot, an automatic teller's machine (ATM) of a financial institution, or a point of sales (POS) of a store.

According to one embodiment, an electronic device is a part of a furniture or building / structure including a communication function, an electronic board, an electronic signature receiving device, a projector, or various measurements And at least one of a device (eg, water, electricity, gas, or a radio wave measuring device). An electronic device according to an embodiment may be one or a combination of the above-described various devices. Also, an electronic device according to an embodiment may be a flexible device. In addition, it will be apparent to those skilled in the art that the electronic device according to an embodiment is not limited to the above-described devices. The term user used in various embodiments may refer to a person who uses an electronic device or a device (eg, an artificial intelligence electronic device) that uses an electronic device.

An electronic device according to an embodiment includes a processor, a memory, a user interface, and a communication interface, and may be connected to another electronic device through a network. The communication interface may transmit / receive data with other electronic devices within a predetermined distance through wired, wireless network, or wired serial communication. The network enables wired and wireless communication between an electronic device and various entities according to an embodiment. The electronic device may communicate with various entities over the network, and the network may use standard communication technologies and / or protocols. In this case, the network includes, but is not limited to, the Internet, a local area network (LAN), a wireless local area network (WLAN), a wide area network (WAN), a personal area network (PAN), and the like. It will be appreciated by those skilled in the art of communication that there may be other types of networks capable of transmitting and receiving information.

According to one embodiment, the wiper for the target clean room is a fiber product used to remove foreign substances in the clean room and the product in the clean room, and may be classified into a wiper and a roll wiper depending on the shape. In the case of a wiper, it may be classified into at least one of poly, microfiber, and nonwoven fabric according to the type of fabric. In the case of roll wipers, they may be classified into at least one of a microfiber fabric, a knit, and a twill depending on the application.

According to an embodiment, the control device may control cameras photographing respective points in the wiper fabric for the target clean room. Cameras can be represented by different constants, for example, through the constants 1, 2, and 3, each camera can be represented in the same manner as camera 1, camera 2, camera 3. Constants are not necessarily numbers, but may consist of letters, special characters, numbers, or a combination thereof. The position of each camera can be represented through a cylindrical coordinate (O). The origin of the cylindrical coordinate system O may be predefined, and for example, may be a position separated by a predefined distance with respect to the center of the wiper fabric for the clean room. Centering on the cylindrical coordinate system O, camera 1 captures the detail region of the region of interest with the largest value of azimuth angle φ, and camera 2 captures the detail region of the region of interest with the next largest value of azimuth angle φ. The camera 3 may capture a detailed region of the ROI having the smallest value of the azimuth angle φ. The region of interest may be a combination of at least one of the target points to be photographed in the wiper fabric for the target clean room.

According to one embodiment, the control device may collect the image taken by the cameras and display it on the display. The control device may control the camera operation such as changing the position of the cameras or adjusting the image collection period. Data transfer between the cameras and the control device may be via a wireless network. Cameras and control devices may include communication antennas, hardware, software, and combinations thereof for a wireless network, which may use standard communication techniques and / or protocols.

According to an embodiment, each of the cameras, which are represented by different constants, may respectively photograph the first images of the detail region of the ROI. Centering on the cylindrical coordinate system O, camera 1 captures the detail region of the region of interest having the largest value of the azimuth angle φ as the first image, and camera 2 has the next region of interest having the next largest value of the azimuth angle φ. The detailed region of may be photographed as a first image, and the camera 3 may photograph the detailed region of the region of interest having the smallest value of the azimuth angle φ as the first image.

According to an embodiment, the control device may acquire 102 first images photographed by the cameras. The first images acquired by the control device may be displayed by the display.

According to one embodiment, the control device comprises an input layer of a convolutional neural network that has been previously learned to generate a panoramic image of a wiper cutting specification for a clean room from a plurality of images based on the first images. An input vector corresponding to may be generated (103). The control device can apply an input vector to the convolutional neural network to obtain an output vector (104). The control device can generate 105 position adjustment data based on the output vector to align the shooting positions of the cameras in formation based on the constants. When generation of the position adjustment data is completed, the control device may transmit the position adjustment data to the cameras.

According to one embodiment, the control device can align the cameras 106 on a large scale based on the position adjustment data. Each camera identifies the data assigned to itself in the position adjustment data and executes a movement command assigned to each of them, thereby forming a shooting position in formation based on constants corresponding to each camera in a one-to-one correspondence. Can be sorted. In the aligned state, the cameras may take a second image. After the photographing, the cameras may transmit the second images to the control device.

According to one embodiment, the control device can acquire second images captured by the aligned cameras (107). The control device may generate a target panoramic image of respective points in the far end of the wiper for the target clean room based on the second images (108). The control device may generate a target image, such as a panorama image of the ROI, based on the second images.

According to an embodiment, the control device may acquire an identifier corresponding to each point in the wiper far end for the target clean room based on the target panoramic image (109). The identifier corresponding to each point in the wiper fabric for the target cleanroom is information identifying each point in the wiper fabric for the target cleanroom, e.g., a code relating to the shape, size and material of the cleanroom wiper to be subjected to the manufacturing process. Although not limited thereto, the process may be defined in various ways depending on the type or design intent.

According to an embodiment, the control device may obtain the cutting type and manufacturing process of the target cleanroom wiper corresponding to the identifier from a database that matches the identifiers, the cutting types of the cleanroom wiper and the manufacturing processes of the cleanroom wiper. (110). For example, the database may be implemented with a table that matches the identifiers and cutting types and manufacturing processes of the wiper for the clean room, and the control device may query the database.

According to one embodiment, the cutting type of the wiper for clean room may be classified according to the processing method such as laser cutting, ultrasonic cutting, thermal cutting, etc., but is not limited thereto and may be defined in various ways according to the shape and roughness of the cutting finish. have.

According to an embodiment, the control device may generate the first input signal based on the identifier, the cutting type of the wiper for the target clean room, and the manufacturing process (111). The control device may generate a vector by applying weights to the identifier, the cutting type of the target clean room wiper, and the manufacturing process, and each weight may be information retrieved from a database. The control device may generate a first input signal corresponding to the generated vector.

According to one embodiment, the control device may apply 112 a first input signal to an input layer of a first neural network that has been previously learned to generate control information for manufacturing from information associated with a wiper for a clean room. The control device can generate a control command using a deep learning technique.

According to one embodiment, the control device may utilize a neural network learned to generate control commands optimized according to the identifier of the cleanroom wiper, the cutting type of the cleanroom wiper, and the manufacturing process of the cleanroom wiper. The control device may preprocess the variables corresponding to the identifier, the cutting type of the wiper for the target clean room, and the manufacturing process to generate the first input signal. Here, the input signal may be defined in various forms according to the design intention, such as one-hot vector, real vector. Weights may be applied to variables when generating an input signal. Here, the weights may be optimized when learning the neural network.

According to an embodiment, the control device may apply the first input signal to the first neural network that has been previously learned based on control commands for the manufacturing process facilities. The input signal may correspond to an input layer of the neural network.

According to an embodiment, the control device may acquire a first output signal generated from an output layer of the first neural network (113). The neural network may generate output signals by processing values passing through an intermediate layer into an output layer using a nonlinear activation function.

According to one embodiment, the output signal may comprise nodes corresponding to control commands of the manufacturing process facilities. The output layer may include a first node that outputs a value corresponding to the specific control command, a second node that outputs a value corresponding to the specific control command, and the like. The value output by the node can be represented by discrete locations or by successive values such as probability.

According to one embodiment, the control device can generate 114 first control commands based on the first output signal. The first control commands may include mechanical, physical and electronic command profiles for controlling the inspection process facilities. The first output signal may be represented by information associated with a probability corresponding to each of the optimized control commands.

According to one embodiment, the neural network includes an input layer to which training samples are input and an output layer to output training outputs, and can be learned based on the difference between the training outputs and the labels. Here, the labels can be defined based on the estimation results for the control commands for controlling the process facilities. A neural network is connected to a group of nodes and is defined by weights between the connected nodes and an activation function that activates the nodes.

According to an embodiment, the learning apparatus may train a neural network using a gradient decent (GD) technique or a stochastic gradient descent (SGD) technique. The learning device may use a loss function designed by the outputs and labels of the neural network.

According to an embodiment, the learning apparatus may calculate a training error using a predefined loss function. The loss function can be predefined with labels, outputs and parameters as input variables, where the parameters can be set by weights in the neural network. For example, the loss function may be designed in the form of Mean Square Error (MSE), entropy, or the like. Various embodiments or methods may be employed in embodiments in which the loss function is designed.

According to an embodiment, the learning apparatus may find weights that affect the training error by using a backpropagation technique. Here, the weights are relationships between nodes in the neural network. The learning apparatus may use the SGD technique with labels and outputs to optimize the weights found through the backpropagation technique. For example, the learning apparatus can update the weights of the loss function defined based on the labels, outputs and weights using the SGD technique.

According to an embodiment, the control device may acquire an output signal using the neural network that has been learned, and estimate information related to control commands for controlling the manufacturing process facilities.

According to one embodiment, the control device may control 115 the manufacturing process facilities based on the first control commands. The control apparatus can control the manufacturing process facilities based on first control instructions that include policies or profiles for controlling the manufacturing process facilities. The control device may monitor the operation of each manufacturing process based on the operations and the monitoring results of the manufacturing process facilities responding to the first control commands to check the speed and the result of the manufacturing process.

According to an embodiment, the control device may control at least one of physical operation, logical operation, and change of setting information of the manufacturing process facility based on the first control command. The control device may control the manufacturing process facility in series on the basis of a signal obtained from the manufacturing process facility in response to the control signal.

According to an embodiment, the control device may generate structural specification information and machine specification information for manufacturing the wiper for the target clean room based on the first control commands. According to an embodiment, the control device may generate test signals for manufacturing the wiper for the target clean room and time stamps for applying the test signals, based on the structural specification information and the machine specification information. The control device may generate an input vector based on the structural specification information and the machine specification information, and apply the generated input vector to the input layer including nodes corresponding to the structural specification information and the machine specification information, respectively. The control device can obtain an output vector from the output layer of the neural network. Nodes of the output layer may correspond with test signals and time stamps, respectively. The control device can generate test signals and time stamps based on the output vector. The above description may be applied to embodiments related to inference and learning related to test signal and time stamp generation.

According to one embodiment, the control device may control the manufacturing process facilities based on the sequence of test signals and time stamps. The control device can streamline the manufacturing process and increase the processing speed through the speed of the manufacturing process equipment and the feedback of the results.

According to an embodiment, the control device may infer control information related to a process of manufacturing a wiper for a clean room based on artificial intelligence to optimize process control. The control device can streamline the process steps through the inter-facility feedback command required for the process, manage the locations and areas where errors frequently occur during the process through the feedback command, and improve the quality of the process. The control device can increase the precision of process equipment control by tying the information necessary for the manufacturing process of the wiper for clean room.

FIG. 2 is a diagram for describing a method of generating a panoramic image for cutting of a wiper for a clean room, according to an exemplary embodiment.

According to an embodiment, the control device may generate an input vector 300 corresponding to an input layer of the convolutional neural network 400 based on the obtained first images. In generating the input vector 300 from the first images, there may be a process of processing the horizontal length and the vertical length of each first image to be the same. The length unit may be a physical length unit such as cm, mm, or inch, or may be an image unit such as dpi or ppi.

By unifying the size of each first image, the control device can generate an input through which the convolutional neural network 400 can easily compare each image. In addition, in generating the input vector 300 from the first images, there may be a process of processing the first images as a black and white image or a specific color among RGB colors. Through this, the input vector 300 may be an image file of a single channel in which color information is unified. When an image file of a single channel is input to the convolutional neural network 400, the convolutional neural network needs to calculate a feature map having only height and width and no depth. With less computing power, you can achieve the target output at higher speeds. Through this, even when the control device is not a dedicated server or a supercomputer, the calculation of the convolutional neural network 400 may be easily performed.

According to an embodiment, the control device may obtain the output vector by applying the generated input vector 300 to the convolutional neural network 400. The output vector may be first overlapped partial data 311 in which position and size data of parts overlapping with the first images captured by the remaining cameras in the first image photographed by each camera are grouped for each camera. .

Looking at the first overlapping partial data 311, which is the output of the convolutional neural network 400, the preset data per pixel unit is shown as a circle, and the pixel-specific data of the image captured by the camera 1 is shown. Are shown in the left rectangle, the pixel-by-pixel data of the image photographed by the camera 2 in the middle rectangle, and the pixel-by-pixel data of the image photographed by the camera 3 in the right rectangle.

Meanwhile, the data for each pixel unit may have a value of 0 or 1. It is shown as an uncolored circle when the pixel-by-pixel data has a value of 0 and as a colored circle when the pixel-by-pixel data has a value of 1. For the partial image corresponding to one preset pixel unit in the first image captured by one camera, the convolutional neural network 400 overlaps the partial image in the pixel unit with the first images captured by the other cameras. When it is inferred that a pixel unit having an image exists, the pixel unit data of the pixel unit has a value of 1 and is shown as colored circles. Meanwhile, with respect to the partial image corresponding to one preset pixel unit in the first image photographed by one camera, the convolutional neural network 400 overlaps the partial image in the pixel unit in the first images photographed by the remaining cameras. If it is inferred that there is no pixel unit having the partial image to be present, the pixel unit data of the pixel has a value of 0 and is shown as an uncolored circle.

According to an embodiment, in the case of the first overlapped partial data 311, five columns on the right side of the left rectangular pixel unit data corresponding to the image photographed by the camera 1 based on a preset pixel unit The pixel-specific data of all rows of) correspond to one. This means that a partial image consisting of pixels corresponding to all rows of the five columns on the right, based on preset pixel units in the image taken by Camera 1, overlaps with a portion of the image taken by Camera 2 It means.

According to an embodiment, among the data of the pixel unit of the rectangle corresponding to the image photographed by the camera 2, five rows of the middle part of the three columns of the right and the left column based on the preset pixel unit ), Data per pixel unit corresponds to 1. This means that partial images consisting of five rows of pixels in the middle of the three columns on the right and left sides of the image taken by the camera 2 are preset to the camera 1 and camera 3. It means that it overlaps with some of the images taken by it.

Further, among the pixel data of the right rectangle corresponding to the image photographed by the camera 3, pixel-by-pixel data of all rows of the five columns on the left side is set to 1 based on a preset pixel unit. Corresponding. This means that in the image captured by Camera 3, a partial image consisting of pixels corresponding to all rows of the five columns on the left, based on a preset pixel unit, overlaps with a portion of the image captured by Camera 2 It means.

According to an embodiment, the control device generates an input vector 300 based on the first images photographed by the respective cameras, and then applies the input vector 300 to the convolutional neural network 400 to output the input vector 300. The first overlapped partial data 311 may be obtained, and the first overlapped partial data 311 may be the location and size of the overlapped portions of the first images photographed by the remaining cameras in the first image photographed by each camera. The data may be grouped by each camera. In this case, the first overlapped partial data 311 is not fixed to one value, and may have different values depending on how the first images photographed differently according to the relative positions of the respective cameras overlap each other (312, 313). ). The detailed configuration and operation of the convolutional neural network 400 for obtaining the input vector 300 and generating the output vector will be described later.

According to one embodiment, the control device position adjustment data 320 to arrange the shooting position in a form based on the constants to the cameras having a one-to-one correspondence with different constants based on the obtained output vector Can be generated. Here, the large size based on the constants may be a large size in which multiple cameras respectively corresponding to the constants are sequentially arranged in the azimuthal direction of the cylindrical coordinate system O in the order defined based on the constants. That is, based on the cylindrical coordinate system O, the cameras 1 corresponding to the constant 1 have the largest azimuth angle φ while the multiple cameras have the same distance r and height z values. The corresponding camera 2 has the next largest azimuth angle φ, and the camera 3 103 corresponding to the constant 3 may be large having the smallest azimuth angle φ value. In this case, the origin of the cylindrical coordinate system O may be predefined based on the target elevator.

According to an embodiment, the position adjustment data 320 for controlling each camera to move to a large size based on the constants may be generated through the following process. The control device may calculate first relative positions of the cameras based on the first overlapped partial data 311. When the output vector of the convolutional neural network 400 is the same as the first overlapped partial data 311, the controller may calculate which positional relationship each camera has with respect to each other from the overlapping relationship of the images taken by the respective cameras. Based on this, the relative positions of the respective cameras can be calculated.

For example, when the output vector of the convolutional neural network 400 is the same as the first overlapped partial data 311, the control device may determine that the relative photographing position of each camera is based on the cylindrical coordinate system O. It can be calculated that the height z of 2 and camera 3 is the same. It is also possible to calculate that camera 1 has the largest azimuth angle φ value, camera 2 has the next largest azimuth angle φ value, and camera 3 has the smallest azimuth angle φ value. Further, it can be calculated that camera 1 and camera 3 are located at a distance r farther from the origin of the cylindrical coordinate system O by the same distance r, and camera 2 is closer to the origin than them. The control device can calculate the first relative positions of the multiple cameras based on these calculation results.

When the cameras arrange the photographing positions in a large size based on the constants, the second device photographed by the cameras in which the overlapping portions of the second images photographed by the cameras in the vicinity are adjacent to each other. Second relative positions of the cameras may be calculated such that the position and size occupied in the images coincide with the predefined second overlapping partial data.

The second overlapping partial data may be captured by the camera 1 when the preset data per pixel unit may be illustrated as a circle, and each of the cameras arranges the shooting positions in a large size based on constants. The pixel-by-pixel data of the image to be imaged may be shown in the left rectangle, the pixel-by-pixel data of the image to be photographed by the camera 2 in the middle rectangle, and the pixel-by-pixel data of the image to be photographed by the camera 3 in the right rectangle. The data for each pixel unit may have a value of 0 or 1.

According to one embodiment, the control device, when each of the cameras arranged the shooting position in a large-scale based on the constants, the second predefined portions of the overlapping each other between the second images to be taken by each camera in the aligned state The second relative positions can be calculated to match the duplicate partial data. More specifically, the control device is characterized in that each camera has the same distance r and height z values relative to the cylindrical coordinate system O; Camera 1 has the largest azimuth value, camera 2 has the next largest azimuth value, and camera 3 has the smallest azimuth value; The difference value of the azimuth angle between camera 1 and camera 2 and the difference value of the azimuth angle between camera 2 and camera 3 are the same; The difference between the azimuth angles between the camera 1 and the camera 2 is that the overlapping portions of the images to be photographed by the camera 1 and the camera 2 are right or left adjacent to each other of each image based on a preset pixel unit as the second overlapping portion data. A value that makes it part of the pixels corresponding to all rows of a column of; The second relative positions of the respective cameras may be calculated and obtained.

According to one embodiment, the control device based on the first relative positions and the second relative positions, position adjustment data 320 to control each camera to move from the first relative positions to the second relative positions, respectively Can be generated.

According to one embodiment, the control device may transmit position adjustment data 320 to the cameras, each of the cameras identify the data assigned to each of them in the position adjustment data 320, so that each movement assigned to itself By executing the command, it is possible to align the photographing position to the exact large size based on the constants corresponding to the respective cameras in a one-to-one correspondence. The cameras may take pictures of the second images in the aligned state. Portions of each second image that are overlapped with the remaining second images may correspond to portions of pixel-specific data having a value of 1 in the second overlapped portion data. After the photographing, the cameras may transmit the second images to the control device. The control device obtaining the second image may acquire the target image based on the second image.

3 is a diagram for describing a method of generating a panoramic image for cutting of a wiper for a clean room, according to an exemplary embodiment.

According to one embodiment, the convolutional neural network 400 based on the input vector 300 generated through the images taken by each camera, the data of the position and size of the portions of each image overlapping with the remaining images The first duplicated data 311 may be generated as an output. The convolutional neural network 400 may load the kernel or the input vector 300 from a pre-built database, and the database may be embodied as a memory included in the control device or may be connected to the control device and wired, wireless, or network. It can be implemented as an external device such as a server that can be connected.

In machine learning, a convolutional neural network 400, a type of neural network, includes convolution layers 410 and 420 designed to perform convolution operations. The convolution layers 410 and 420 of the convolutional neural network 400 may perform a convolution operation associated with an input using at least one kernel. When the convolutional neural network 400 includes a plurality of convolutional layers 410 and 420, the control device performs a plurality of convolutional operations by performing respective convolution operations corresponding to the respective convolutional layers 410 and 420. You can do that. The sizes of the inputs, kernels, and outputs of the respective convolution layers 410 and 420 may be defined according to aspects in which the corresponding convolution layers are designed.

According to an embodiment, the first convolutional layer 410 serves as an input layer that receives the input vector 300, and the input vector 300 is convolved with the kernel to form a first feature map. (feature map, 412) can be created. The kernel may be configured as a kernel that detects a brightness feature or an edge feature of the input vector 300, and kernels in the known AlexNet model, ConvNet model, LeNet-5 model, and GoogLeNet model may be used. However, the present invention is not limited thereto, and kernels used in various convolutional neural network models may be used.

According to an embodiment, the control device may apply an activation function 411 to the first feature map 412. This allows for machine learning to be nonlinear deep learning. The activation function may be a known sigmoid function, a hyperbolic tangent function, a rectified linear unit, or the like. However, the present invention is not limited thereto, and activation functions used in various convolutional neural network models may be used.

According to an embodiment of the present disclosure, the control device may generate a first pooled feature map 414 by pooling the first feature maps 412. For pooling, known maximum pooling, mean pooling, and the like may be used. However, the present invention is not limited thereto, and poolings used in various convolution neural network models may be used.

The pooling 413 can reduce the total number of nodes and the total amount of computations in the convolutional neural network 400 while leaving meaningful information, so that the control device can output faster targets with less computing power. Can be obtained. Through this, even when the control device is not a dedicated server or a supercomputer, the calculation of the convolutional neural network 400 may be easily performed.

According to an embodiment, the process of generating a feature map through a convolution layer, applying the activation function to the feature map, and then pooling may be performed a plurality of times. In detail, the second convolutional layer 420 of the convolutional neural network 400 may obtain the first pooled feature map 414 as an input and generate the second feature map 422 through a convolution operation. have. Subsequently, an activation function 421 may be applied to the second feature map 422, and a second pooled feature map 424 may be generated by pooling 423 the second feature map 422 to which the activation function is applied. have.

Also, although not shown, there may be an nth convolutional layer. The n th convolution layer may obtain the (n-1) th pooled feature map as an input and generate an n th feature map through a convolution operation. Subsequently, an activation function may be applied to the n-th feature map, and the n-th pooled feature map may be generated by pooling the n-th feature map to which the activation function is applied. This process of generating the pooled feature maps multiple times reduces the total number of nodes and the total amount of computations in the convolutional neural network, while leaving meaningful information, so that the target output can be generated at higher speed with less computing power. You can get it.

According to one embodiment, the control device connects all of the pooled feature maps 424 to a fully connected layer 490, and the first redundant partial data 311 that is output from the pulley connected layer 490. ) Can be created.

4 is a diagram illustrating a learning process of a convolutional neural network employed to generate a panoramic image for cutting a wiper for a clean room, according to an exemplary embodiment.

According to one embodiment, the learning of the convolutional neural network may be performed by the learning apparatus. A learning device is a device that trains a convolutional neural network that infers the location and size of overlapping portions in a plurality of images that may include overlapping portions, such as software modules, hardware modules, or a combination thereof. Can be implemented.

According to an embodiment, the learning apparatus acquires images, which may include portions overlapping each other, as training data, and inputs training data corresponding to an input layer of a convolutional neural network based on the acquired training data. Can be generated. The learning device can then apply the input of training data to the convolutional neural network to obtain an output of the training data generated by the convolutional neural network, and then compare the output of the training data with a label to convolution. Neural network can be optimized.

According to an embodiment, the learning apparatus may use images, which may include portions overlapping each other, as training data 1100. The learning apparatus may obtain the output 1120 of the training data by applying the input 1110 of the training data generated based on the training data 1100 to the convolutional neural network 400. Specifically, the convolutional neural network 400 generates a feature map by convolutional operation of the data and the kernel input through the convolutional layer; Applying an activation function to the feature map; And generating a pooled feature map by pooling the feature map to which the activation function is applied. The convolutional neural network 400 may generate an output 1120 of training data from the feature pool last pulled through the pulley connected layer.

According to an embodiment, the learning apparatus may compare the output 1120 of the training data with the label 1130 to optimize the convolutional neural network 400 (1140).

According to an embodiment of the present disclosure, the label 1130 may be data that pre-records positions and sizes of portions of the training data 1100 that overlap with the remaining images. In detail, the label 1130 may be data to which data per pixel unit consisting of 0 or 1 is preset for each image of the training data 1100, for each pixel unit. have. More specifically, the label 1130 has a value of 1 as pixel data for each image of the training data 1100, for pixel units overlapping with other images in the image for a preset pixel unit. For pixel units that are provided and do not overlap with other images in the corresponding image, a value of 0 may be data that is previously given as data for each pixel unit.

According to an exemplary embodiment, the output 1120 of the training data may be data obtained by the convolutional neural network 400 inferring values of positions and sizes of portions of the training data 1100 that overlap with the remaining images. have. In detail, the output 1120 of the training data is 0 or 1 pixel unit for each preset pixel unit for each image of the training data 1100 according to the inference result of the convolutional neural network 400. It may have data per pixel unit. More specifically, the convolutional neural network 400 outputs training data for each pixel of the training data 1100, for pixel units inferred to overlap with the rest of the images in the preset pixel unit. A value of 1 is recorded in the corresponding pixel unit data of 1120, and 0 for the corresponding pixel unit data of the output 1120 of the training data for pixel units inferred not to overlap with the remaining images in the image. By recording a value of, the output 1120 of training data may be generated.

According to an embodiment, the learning apparatus may optimize the convolutional neural network 400 by comparing the output 1120 of the training data corresponding to the above inference result with the labeled data 1130 corresponding to the correct answer (1140). ). The comparison of the training data output 1120 and the labeled data 1130 may be made through a loss function. The loss function may use a known mean squared error (MSE), a cross entropy error (CEE), or the like. However, the present invention is not limited thereto, and a loss function used in various convolutional neural network models may be used as long as the deviation or error of the output 1120 and the labeled data 1130 of the training data can be measured.

According to an embodiment, the learning apparatus may optimize the convolutional neural network 400 by estimating the minimum value of the lossy function and repeating the process of resetting the weight of the convolutional neural network such that the lossy function becomes the minimum value. . In order to optimize the convolutional neural network 400, a known backpropagation algorithm, stochastic gradient descent, and the like may be used. However, the present invention is not limited thereto, and an optimization algorithm of weights used in various convolution neural network models may be used.

5 is a diagram for describing a method for controlling a manufacturing process of a wiper for a clean room, according to an exemplary embodiment.

According to one embodiment, the control system may include a control device 201, a first manufacturing process facility 202, a second manufacturing process facility 203, and a network 205. The subjects in the control system may be connected to each other via the network 205, and embodiments are described based on the control operation of the inspection process of the control device 201, but at least two subjects in the system cooperate with each other to perform the control operation. May be The control device 201 may control the first manufacturing process equipment 202 and the second manufacturing process equipment 203 for the manufacturing process of the wiper 206 for the clean room.

According to an embodiment, the control device 201 may obtain structural specification information and machine specification information for manufacturing the wiper 206 for the clean room. For example, the structural specification information is a structural specification for positioning a cutting tool on a space in a wiper fabric for a target clean room, and may include position coordinates, connection relations, hierarchies, and depths. Machine specification information is a specification according to the physical and chemical reactions and reactions of the wiper fabric and cutting tools for the clean room, including size, length, shape, thickness, weight, material, proper pressure, proper temperature, specification and strength, pH and oxidation degree. It may include.

According to one embodiment, the control device 201 queries a database that matches the structural specification information, the machine specification information, and the cutting tools, and obtains cutting tools corresponding to the structural specification information and the machine specification information. Process equipment 202 may be controlled. For example, the cutting tools may be a laser cutter, hot wire cutter, ultrasonic sealer, or the like.

According to an embodiment, the control device 201 may generate a manufacturing process sequence condition corresponding to the structural specification information based on the control result of the first process equipment. Manufacturing process sequence conditions are conditions for placing the cutting tools in position to match the sequence of the manufacturing process.

According to one embodiment, the control device 201 may control the second manufacturing process facility 203 for placing the tools based on manufacturing process sequence conditions.

According to one embodiment, the control device 201 may compare the control results of the second manufacturing process facility 203 with the manufacturing process sequence conditions to generate a feedback command to adjust the placement of the tools. According to an embodiment, the feedback command is a command for modifying an existing command to correct an in-process error. If the feedback command corresponds to a structural specification of a wiper for a clean room, a mechanical specification of a wiper for a clean room, and a position It may include at least one of an error of an instruction, a delay of a process time according to a time stamp, an incompatibility of cutting tools, or a case in which the next process cannot be performed only by an existing instruction.

According to one embodiment, the control device 201 may control the second manufacturing process facility 203 based on the generated feedback command.

According to an embodiment, the control device 201 may query a database to obtain a first history of clustered tools according to the structural specification information. The control device 201 is based on the first history, the first, second, third and fourth of the acquired tools in time series in the space in the wiper fabric for the target clean room, A sequence of position instructions and time stamps may be generated for placement at the second position, the third position, and the fourth position, respectively. The generated position command and time stamp may be modified later. The modified position command and time stamp information can be recorded in history.

According to an embodiment, the time stamp is a time displacement parameter indicating a starting point of time with respect to a commonly referred time. According to one embodiment, based on the generated sequence, the control device 201 may generate manufacturing process sequence conditions.

According to one embodiment, the control device 201 may generate a first position command and a first time stamp for positioning the first tool at a first position in the wiper fabric for the clean room. The control device 201 may generate a second position command and a second time stamp for positioning the second tool at a second position in the wiper fabric for the clean room. The control device 201 may generate a third position command and a third time stamp for positioning the third tool at a third position within the clean room wiper fabric. The control device 201 may generate a fourth position command and a fourth time stamp for positioning the fourth tool at the fourth position in the wiper fabric for the clean room.

According to one embodiment, the control device 201 controls the second manufacturing process facility 203 to place the first tool, second tool, third tool and fourth tool in a first position, a second position, a third position. And a fourth position.

According to one embodiment, the control device 201 uses a position discrimination sensor in the second manufacturing process facility 203 to produce a first tool, a second tool, a third tool and a third tool of the second manufacturing process facility 203. Four can identify whether the tool is correctly positioned in the first position, the second position, the third position and the fourth position.

Position discrimination sensor according to an embodiment is a kind of position sensor that is on-off at a certain position, a mechanical, electrical, magnetic, optical sensor according to the principle classification or contact or non-contact sensor according to whether or not contact It may include at least one of.

According to one embodiment, the control device 201 uses the infrared sensor of the second manufacturing process facility 203 when the difference between the position coordinates and the actual position is within a threshold range, the first corresponding to the first tool. The temperature can be determined.

According to one embodiment, the control device 201 may compare the first temperature and the appropriate temperature corresponding to the first tool in the machine specification information.

According to one embodiment, the proper temperature means a temperature for changing the machine specifications of the cutting tool, except for the proper temperature, from the inside of the critical range. It may include at least one of the denaturation potential of the room wiper fabric.

According to one embodiment, the control device 201 may calculate the time required for preheating or cooling when preheating or cooling of the first tool is required, based on the comparison result between the proper temperature and the first temperature.

According to one embodiment, when the processing time according to the first time stamp is delayed by the required time, the control device 201 is based on a connection relationship, a hierarchical structure, and a depth corresponding to the first tool in the structural specification information. Among the candidate second tool and the third tool, the cutting tool can be selected in the region closest to the first position while maintaining the connection relationship, hierarchy and depth. If the process time according to the first time stamp is not delayed, the control device 201 may perform the next process.

According to one embodiment, the control device 201 is configured to control the first tool from the first position to the second position, the second tool from the second position to the third position, and the third tool from the third position to the fourth position. The first cutting process signal may be generated to move and operate the fourth tool from the fourth position to the fifth position.

According to an embodiment, the control device 201 may operate the first tool, the second tool, the third tool and the fourth tool based on the first cutting process signal.

According to an embodiment, the operation may include at least one of laser cutting, thermal cutting of a hot wire cutter, and ultrasonic cutting of an ultrasonic sealer.

According to one embodiment, the control device 201, if any one of the first position, the second position, the third position and the fourth position at the end of the cutting is kept uncut, the corresponding position command and A feedback command can be created to modify the location stamp.

According to an exemplary embodiment, the control device 201 may manage a location and an area where errors frequently occur during the manufacturing process of the wiper for the clean room through a history command to improve the quality of the subsequent process.

6 is a diagram for describing a method for controlling a manufacturing process of a wiper for a clean room, according to an exemplary embodiment.

According to one embodiment, the control system may include a control device, a manufacturing process facility 501 for fabric cutting, a laundry device 502, a drying device 503 and a packaging device 504.

According to one embodiment, the control device may generate a first loading signal for loading the products according to the control results of the manufacturing process equipment 501 in the cutting locker in the clean room. According to an embodiment, the control device may control the first loading facility based on the first loading signal. According to one embodiment, the control device may generate a second loading signal based on the signal of the self loading evaluation device in the cutout.

According to one embodiment, there are 10, 100, 1000 classes, etc. in a clean room for manufacturing a wiper for a clean room, and are generally manufactured in a 10 class environment. Class 10 refers to an environment free of less than 10 dust of 0.5 μm and 5 μm of dust particles per cubic foot.

According to one embodiment, the self-loading evaluation device in the cutting box creates the overall loading condition based on the size, weight and size in the machine specification information of the wiper fabric for the clean room, and compares the loading result of the products with the overall loading condition. To determine whether the product has been loaded. Overall loading conditions are generally based on 7,000 to 8,000 products, but are not limited to these and can be defined in various ways depending on the limitations and convenience of the manufacturing process.

According to one embodiment, the second loading facility may be controlled based on the second loading signal. According to one embodiment, it is possible to generate a second control command for washing products simultaneously as a result of the second loading facility. According to an embodiment, the washing apparatus 502 may be controlled based on the second control command.

According to one embodiment, the laundry apparatus 502 is in a clean room and uses DI water (Deionized water). When washing may add an additive, such additive may include at least one of an antistatic agent, a hygroscopic agent, a detergent, a protective agent, an antioxidant.

According to an embodiment, a third loading signal may be generated for transferring products to the drying apparatus 503 according to the control result of the washing apparatus 502. According to one embodiment, the third loading facility may be controlled based on the third loading signal. According to one embodiment, a third control command for simultaneously drying the products according to the control result of the third loading facility can be generated. According to an embodiment, the drying apparatus 503 may be controlled based on the third control command.

According to one embodiment, the drying apparatus 503 is present in a clean room and may be classified into at least one of convective heat transfer, conduction heat transfer, radiant heat transfer, high frequency use, and desiccant use according to a heating method. The drying apparatus 503 may be used in various ways depending on design intention, such as drying, moisture absorption, and sterilization of laundry.

According to one embodiment, it is possible to generate a fourth loading signal for transferring the products according to the control result of the drying apparatus 503 to the conveyor belt. According to an embodiment, the fourth loading facility may be controlled based on the fourth loading signal. According to one embodiment, it is possible to generate a fourth control command for individually vacuum packaging products on the conveyor belt according to the control result of the fourth loading facility. According to one embodiment, the packaging device 504 may be controlled based on the fourth control command.

According to one embodiment, the packaging device 504 generates the overall packaging conditions, based on the size and thickness in the machine specification information of the wiper for the clean room, and compares the packaging results of the products and the overall packaging conditions whether the products are finished packaging Determine. Overall packaging conditions are generally based on 100 products, but are not limited thereto and may be defined in various ways depending on the limitations and convenience of the manufacturing process.

7 is an exemplary diagram of a configuration of an apparatus according to an embodiment.

Device 1201 according to one embodiment includes a processor 1202 and a memory 1203. The apparatus 1201 according to an embodiment may be the server or the terminal described above. The processor may include at least one of the devices described above with reference to FIGS. 1 through 5, or may perform at least one method described above with reference to FIGS. 1 through 5. The memory 1203 may store information related to the above-described method or a program in which the above-described method is implemented. The memory 1203 may be a volatile memory or a nonvolatile memory.

According to one embodiment, the processor 1202 may execute a program and control the device 1201. Code of a program executed by the processor 1202 may be stored in the memory 1203. The device 1201 may be connected to an external device (eg, a personal computer or a network) through an input / output device (not shown), and may exchange data.

The embodiments described above may be implemented as hardware components, software components, and / or combinations of hardware components and software components. For example, the devices, methods, and components described in the embodiments may include, for example, processors, controllers, arithmetic logic units (ALUs), digital signal processors, microcomputers, field programmable gates (FPGAs). It may be implemented using one or more general purpose or special purpose computers, such as an array, a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to the execution of the software. For convenience of explanation, one processing device may be described as being used, but one of ordinary skill in the art will appreciate that the processing device includes a plurality of processing elements and / or a plurality of types of processing elements. It can be seen that it may include. For example, the processing device may include a plurality of processors or one processor and one controller. In addition, other processing configurations are possible, such as parallel processors.

The method according to the embodiment may be embodied in the form of program instructions that can be executed by various computer means and recorded in a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the media may be those specially designed and constructed for the purposes of the embodiments, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

The software may include a computer program, code, instructions, or a combination of one or more of the above, and configure the processing device to operate as desired, or process it independently or collectively. You can command the device. Software and / or data may be any type of machine, component, physical device, virtual equipment, computer storage medium or device in order to be interpreted by or to provide instructions or data to the processing device. Or may be permanently or temporarily embodied in a signal wave to be transmitted. The software may be distributed over networked computer systems so that they may be stored or executed in a distributed manner. Software and data may be stored on one or more computer readable recording media.

Although the embodiments have been described with reference to the accompanying drawings, those skilled in the art may apply various technical modifications and variations based on the above. For example, the described techniques may be performed in a different order than the described method, and / or components of the described systems, structures, devices, circuits, etc. may be combined or combined in a different form than the described method, or other components. Or even if replaced or substituted by equivalents, an appropriate result can be achieved.

Therefore, other implementations, other embodiments, and equivalents to the claims are within the scope of the following claims.

Claims (5)

Photographing each point in the clean room wiper fabric to be manufactured using cameras corresponding to different constants;
Acquiring first images photographed by the cameras;
Generating an input vector corresponding to an input layer of a convolutional neural network that has been previously learned to generate a panoramic image of a wiper cutting standard for a clean room from the plurality of images based on the first images. step;
Applying the input vector to the convolutional neural network to obtain an output vector;
Based on the output vector, generating position adjustment data for aligning the photographing positions of the cameras in a formation based on the constants;
Aligning the cameras in the large format based on the position adjustment data;
Acquiring second images taken by the aligned cameras;
Generating a target panoramic image of the respective points in the clean room wiper fabric based on the second images;
Obtaining an identifier corresponding to each of the points in the clean room wiper fabric based on the object panoramic image;
Obtaining a cutting type and manufacturing process of the cleanroom wiper corresponding to the identifier from a database matching identifiers, cutting types of cleanroom wiper and manufacturing processes of the cleanroom wiper;
Generating a first input signal based on the identifier, the cutting type of the cleanroom wiper and the manufacturing process;
Applying the first input signal to an input layer of a first learned neural network to generate control information for manufacturing from information associated with a clean room wiper;
Obtaining a first output signal generated from an output layer of the first neural network;
Generating first control commands based on the first output signal;
Based on the first control instructions, controlling manufacturing process facilities;
Generating a first loading signal for loading products according to a control result of the manufacturing process facilities into a cutting locker in a clean room;
Controlling a first loading facility based on the first loading signal;
Generating a second loading signal based on a signal of the self loading evaluation device in the cutout box;
Controlling a second loading facility based on the second loading signal;
Generating a second control command for simultaneously washing products according to the control result of the second loading facility;
Controlling a laundry machine based on the second control command;
Generating a third loading signal for transferring the products to a drying apparatus according to the control result of the washing apparatus;
Controlling a third loading facility based on the third loading signal;
Generating a third control command for simultaneously drying the products according to the control result of the third loading facility;
Controlling a drying device based on the third control command;
Generating a fourth loading signal for transferring the products to a conveyor belt according to a control result of the drying apparatus;
Controlling a fourth loading facility based on the fourth loading signal;
Generating a fourth control command for individually vacuum packaging the products on a conveyor belt according to the control result of the fourth loading facility; And
Controlling a packaging device based on the fourth control command
Containing
Method for controlling the wiper manufacturing process for a clean room.
The method of claim 1,
The output vector is
In the first image taken by any one of the cameras includes first overlapping partial data grouping the position and size data of the portions overlapping with the first image taken by the remaining cameras for each camera,
The large is
A large size in which the cameras respectively corresponding to the constants are sequentially arranged in the azimuthal direction of the reference position according to the order defined based on the constants,
The position adjustment data
Based on the first redundant partial data, obtain first relative positions of the cameras,
When the cameras arrange the photographing positions in the large size, positions overlapping each other in the second images photographed by the adjacent cameras occupy the second images photographed by the adjacent cameras; and Calculating second relative positions of the cameras such that the size matches the predefined second overlapping partial data,
Data that controls the cameras to move from the first relative positions to the second relative positions, respectively,
Method for controlling the wiper manufacturing process for a clean room.
The method of claim 1,
Controlling the manufacturing process facilities
Based on the first control commands, structural specification information for manufacturing the wiper for the cleanroom—the structural specification information is a structural specification for positioning the cutting tool on the space in the wiper fabric for the cleanroom, the position coordinates, Machine specification information—the machine specification information is a specification according to the physical, chemical action and reaction of the wiper fabric and cutting tools for the clean room, including size, length, shape, thickness, and weight. Creating a material, including appropriate material, proper pressure, proper temperature, specification and strength, pH, degree of oxidation;
Generating test signals for manufacturing the wiper for the clean room and time stamps for applying the test signals based on the structural specification information and the machine specification information; And
Controlling the manufacturing process facilities based on the sequence of test signals and the time stamps
Including,
Method for controlling the wiper manufacturing process for a clean room.
The method of claim 3,
Controlling the manufacturing process facilities
Querying a database that matches structural specification information, machine specification information, and cutting tools to control a first manufacturing process facility for obtaining the cutting tools corresponding to the structural specification information and the machine specification information;
Generating a manufacturing process sequence condition corresponding to the structural specification information, the condition for placing the tools in a position that matches the sequence of the manufacturing process, based on the control result of the first manufacturing process facility;
Controlling a second manufacturing process facility to place the tools based on the manufacturing process sequence conditions;
Comparing the control result of the second manufacturing process facility with the manufacturing process sequence conditions to generate a feedback command for adjusting the placement of the tools; And
Controlling the second manufacturing process facility based on the generated feedback command
Including,
Method for controlling the wiper manufacturing process for a clean room.
The method of claim 3,
Generating the manufacturing process sequence conditions
Based on the first history of the tools clustered according to the structural specification information, a first tool, a second tool, a third tool and a fourth tool of the obtained tools are watched on the space in the wiper fabric for the clean room. Generating a sequence of position instructions and time stamps for thermally placing at a first position, a second position, a third position, and a fourth position, respectively; And
Based on the generated sequence, generating the manufacturing process sequence conditions
Including,
Method for controlling the wiper manufacturing process for a clean room.

Images