System and method for detecting anomalous conditions during forming of glass ribbon

Abstract

Apparatuses and methods for detecting anomalies during production of glass sheets are described herein. In some examples, a glass forming apparatus pulls down molten glass to produce a glass sheet. In addition, a plurality of cameras capture images of an interior of a housing assembly of the glass forming apparatus and at least a portion of the glass sheet. In addition, the computing device receives the image from the camera and applies a machine vision process to the image. Based on applying the machine vision process to the images, the computing device determines a location of an object in each of the images. The computing device further determines an anomaly based on the location of the object in each of the images. The computing device then generates anomaly detection data based on the anomaly, and stores the anomaly detection data in a database.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the benefit of priority under 35 U.S.C. §119 to U.S. Provisional Application Serial No. 63 / 531,932, filed on August 10, 2023, the contents of which are the basis of this application and are incorporated herein by reference in their entirety. Technical Field

[0003] The present disclosure relates to the production of glass sheets, and more particularly, to apparatus and methods for detecting anomalies during the production of glass sheets. Background Art

[0004] Glass sheets are used in a variety of applications. For example, they can be used in glass display panels such as mobile devices, laptops, tablet computers, computer monitors, and television displays. Glass sheets can be manufactured by a slot draw process, whereby molten glass is drawn through a slot to form a glass sheet. However, in some cases, anomalies may occur in the slot draw process. For example, a glass sheet may crack during the draw, or molten glass may accidentally overflow. These anomalies may result in the generation of unexpected costs. For example, a new glass sheet needs to be produced to replace the cracked glass sheet, and further production of additional glass sheets may be delayed because a replacement glass sheet must be produced. In addition, overflow of molten glass in unexpected places may cause damage to the glass draw system, thereby incurring additional costs. Therefore, there is an opportunity to improve the production of glass sheets. Summary of the invention

[0005] Embodiments disclosed herein are directed to apparatus and methods for detecting anomalies (e.g., abnormal conditions) within a glass forming system during glass sheet production. For example, embodiments can detect when molten glass unexpectedly drips during glass sheet forming. Embodiments can additionally or alternatively detect cracked glass sheets and other anomalies during glass sheet forming.

[0006] For example, embodiments may employ cameras to capture images of glass sheets as they are formed and drawn down by glass drawing equipment (e.g., rollers). The cameras may be positioned inside the cavity of the glass forming apparatus so that their field of view includes the glass and the glass drawing equipment. In some examples, at least two cameras capture images from both ends of the glass as it is being formed. For example, a first camera may be positioned at one end within the cavity of the glass forming apparatus to capture images of the glass drawing equipment and the glass being formed from a first perspective. A second camera may be positioned at another end within the cavity of the glass forming apparatus to capture images of the glass drawing equipment and the glass being formed from a second perspective. In some cases, two, three, or more cameras are positioned at each end of the cavity of the glass drawing apparatus to capture images of the glass drawing equipment and the glass being formed from various perspectives (e.g., opposite perspectives).

[0007] In addition, a computing device (such as a server) can receive an image and can apply a machine vision process to the image to generate anomaly detection data characterizing whether any anomalies are detected. For example, the server can receive images from each of the cameras at corresponding time intervals during the glass sheet forming, and can apply the machine vision process described herein to the image. For example, the machine vision process can include generating features based on the image and inputting the features into a trained machine learning or artificial intelligence model to generate anomaly detection data. Based on the application of the machine vision process to the image, the computing device can detect drips (e.g., dripping trajectories of molten glass) and can generate anomaly detection data characterizing the detected drips. The computing device can perform additional operations based on the detected drips. For example, the computing device can display an image of the detected drips, can send a warning signal (e.g., a message), can cause the rollers of the glass drawing equipment to be disengaged from the glass being pulled down, or can pause the glass forming device from further pulling down the glass, among other examples.

[0008] In some embodiments, a glass forming system includes a glass forming apparatus, at least one camera, and a computing device communicatively coupled to the glass forming apparatus and the at least one camera. The glass forming apparatus is configured to pull down molten glass to produce a glass sheet. In addition, the at least one camera is configured to capture an image of at least a portion of the glass forming apparatus and the glass sheet. In addition, the computing device is configured to receive images from the at least one camera. The computing device is further configured to apply a machine vision process to the image. Based on applying the machine vision process to the image, the computing device is configured to determine the location of at least one object in each of the images. The computing device is further configured to determine at least one anomaly based on the location of at least one object in each of the images. The computing device is further configured to generate anomaly detection data based on the at least one anomaly. In addition, the computing device is configured to store the anomaly detection data in a database.

[0009] In some embodiments, an apparatus includes a memory storing instructions and a processor communicatively coupled to the memory. The processor is configured to execute the instructions to receive images from at least one camera, the at least one camera configured to capture images of at least a portion of a glass forming apparatus and a glass sheet. The computing device is further configured to apply a machine vision process to the images. Based on applying the machine vision process to the images, the computing device is configured to determine a location of at least one object in each of the images. The computing device is further configured to determine at least one anomaly based on the location of the at least one object in each of the images. The computing device is further configured to generate anomaly detection data based on the at least one anomaly. In addition, the computing device is configured to store the anomaly detection data in a database.

[0010] In some embodiments, a method performed by a processor includes receiving images from at least one camera, the at least one camera configured to capture images of at least a portion of a glass forming apparatus and a glass sheet. The method also includes applying a machine vision process to the images. Based on applying the machine vision process to the images, the method includes determining a location of at least one object in each of the images. The method also includes determining at least one anomaly based on the location of the at least one object in each of the images. The method also includes generating anomaly detection data based on the at least one anomaly. Additionally, the method includes storing the anomaly detection data in a database.

[0011] In some embodiments, a non-transitory computer-readable medium stores instructions that, when executed by one or more processors, cause the one or more processors to perform a method, the method comprising receiving images from at least one camera, the camera being configured to capture images of at least a portion of a glass forming apparatus and a glass sheet. The method further comprises applying a machine vision process to the images. Based on applying the machine vision process to the images, the method comprises determining a location of at least one object in each of the images. The method further comprises determining at least one anomaly based on the location of the at least one object in each of the images. The method further comprises generating anomaly detection data based on the at least one anomaly. Additionally, the method comprises storing the anomaly detection data in a database. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The above summary and the following detailed description of exemplary embodiments may be read in conjunction with the accompanying drawings. The accompanying drawings illustrate some exemplary embodiments discussed herein. As will be further explained below, the claims are not limited to the exemplary embodiments. For clarity and ease of reading, views of certain features may be omitted in the accompanying drawings.

[0013] Figure 1 A glass forming anomaly detection system according to some examples is illustrated.

[0014] Figure 2 is a block diagram of an anomaly detection computing device according to some examples.

[0015] Figure 3 Exemplary functionality of an anomaly detection computing device according to some examples is illustrated.

[0016] Figure 4A and 4B Cameras that may be used in a glass forming anomaly detection system are illustrated according to some examples.

[0017] Figure 5 A generated image including a detected drop of molten glass according to some examples is illustrated.

[0018] Figure 6 Exemplary methods for detecting anomalies during glass forming according to some examples are illustrated. DETAILED DESCRIPTION

[0019] The present application discloses illustrative (i.e., example) embodiments. The present disclosure is not limited to illustrative embodiments. Therefore, many implementations of the claims will differ from the illustrative embodiments. Various modifications may be made to the claims without departing from the spirit and scope of the present disclosure. The claims are intended to cover implementations including such modifications.

[0020] Sometimes, this application uses directional terms (e.g., front, back, top, bottom, left, right, side, edge, etc.) to give the reader context when viewing the drawings. However, the claims are not limited to the orientations shown in the drawings. Any absolute terms (e.g., high, low, etc.) can be understood to disclose the corresponding relative terms (e.g., higher, lower, etc.). In addition, any singular term (e.g., processor) can be understood to disclose the corresponding plural term (e.g., processors), etc., and any plural term can be understood to disclose the corresponding singular term.

[0021] Now referring to the accompanying drawings, Figure 1 The glass forming anomaly detection system 100 is illustrated, including a glass forming device 120, a first camera 140, a second camera 142, a third camera 144, a fourth camera 146, an anomaly detection computing device 160, and a database 170. The glass forming device 120 (e.g., a molten glass forming device) includes a cavity 121, the longitudinal sides of which are surrounded by walls 123 and 124. The walls 123 and 124 are respectively integrated with each of the opposite and downwardly inclined glass forming surfaces 133 and 134. The pair of opposite and downwardly inclined glass forming surfaces 133 and 134 terminate and meet at a joint position 135.

[0022] Molten glass (e.g., high purity molten glass) is delivered into the cavity 121 until the molten glass overflows each of the walls 123 and 124 and then flows along the glass forming surfaces 133 and 134 to C. The molten glass flowing over the glass forming surfaces 133 and 134 meets at the joining position 135 and begins to form a glass sheet 150 (e.g., a glass ribbon). Pulling rollers 137A, 137B are located downstream of the joining position 135 and engage with each of the side edges on both sides of the glass sheet 150 to apply tension to the glass sheet 150. The pulling rollers 137A, 137B can be positioned sufficiently below the joining position 135 so that the thickness of the glass sheet 150 is substantially fixed at this position. The pulling rollers 137A, 137B can pull the glass sheet 150 downward at a specified rate, which establishes the thickness of the glass sheet 150 when it is pulled from the joining position 135.

[0023] Thus, the molten glass is pulled down from the joining position 135 until the glass sheet 150 is formed. The glass sheet 150 can be formed into any suitable thickness. For example, in some cases, the thickness of the glass sheet 150 can be any value between 500 microns and 3 millimeters (e.g., 2 millimeters). In other cases, the thickness of the glass sheet 150 can be greater than 3 millimeters, or less than 500 microns.

[0024] As shown, pulling rollers 137A, 137B pull glass sheet 150 down through glass sheet housing 102 (e.g., trough). Each of first camera 140, second camera 142, third camera 144, and fourth camera 146 can be mounted and positioned within housing 102 to capture images of glass sheet 150 being formed and images of corresponding interior portions of housing 102. For example, first camera 140 can be mounted at first end 151 of housing 102 and positioned such that its field of view includes at least a top portion of glass sheet 150. Second camera 142 can also be mounted at first end 151 of housing 102, located below first camera 140, and positioned such that its field of view includes at least a bottom portion of glass sheet 150. The fields of view of first camera 140 and second camera 142 can overlap and can include portions of the interior of the housing, as described herein.

[0025] Likewise, third camera 144 can be mounted at second end 153 of housing 102 and positioned such that its field of view includes at least a top portion of glass piece 150. Fourth camera 146 can also be mounted at second end 153 of housing 102, at a location below third camera 144, and positioned such that its field of view includes at least a bottom portion of glass piece 150. The fields of view of third camera 144 and fourth camera 146 can overlap and can include portions of the interior of the housing, as described herein.

[0026] Additionally, each of the first camera 140 , the second camera 142 , the third camera 144 , and the fourth camera 146 may be any suitable camera, such as a high-definition video camera.

[0027] In addition, the anomaly detection computing device 160 can be communicatively coupled to each of the first camera 140, the second camera 142, the third camera 144, and the fourth camera 146, and in some examples, can also be communicatively coupled to the glass forming apparatus 120. For example, the anomaly detection computing device 160 can be configured to communicate with each of the first camera 140, the second camera 142, the third camera 144, and the fourth camera 146 via the communication network 116. The communication network 116 can be a wired or wireless communication network. For example, in some examples, the communication network 116 can include one or more universal serial (USB) (for example, USB 2.0) or The cables are used to establish a communication channel between the anomaly detection computing device 160 and each of the first camera 140, the second camera 142, the third camera 144, and the fourth camera 146. In some examples, the communication network 116 may include a cellular network, network, network or any other suitable wireless network. In some cases, the anomaly detection computing device 160 communicates with one or more of the first camera 140, the second camera 142, the third camera 144, and the fourth camera 146 using one or more protocols such as the Real Time Streaming Protocol (RTSP). In some examples, the anomaly detection computing device 160 sends a signal (e.g., a message) to any of the first camera 140, the second camera 142, the third camera 144, and the fourth camera 146 to cause the receiving camera to adjust the position of its sensor (e.g., adjust the yaw, pitch, roll), thereby adjusting the camera's field of view (e.g., adjust its field of view up, down, left, or right).

[0028] In addition, anomaly detection computing device 160 may receive images from each of first camera 140, second camera 142, third camera 144, and fourth camera 146. For example, first camera 140 may capture an image of a top portion including first edge 155 of glass piece 150, and may send the image to anomaly detection computing device 160 using communication network 116. Similarly, second camera 142 may capture an image of a bottom portion including first edge 155 of glass piece 150, and may send the image to anomaly detection computing device 160 using communication network 116. Additionally, third camera 1444 may capture an image of a top portion including second edge 157 of glass piece 150, and may send the image to anomaly detection computing device 160 using communication network 116. Similarly, fourth camera 146 may capture an image of a bottom portion including second edge 157 of glass piece 150, and may send the image to anomaly detection computing device 160 using communication network 116. In some cases, each of first camera 140 , second camera 142 , third camera 144 , and fourth camera 146 captures an image of glass sheet 150 during a time interval (eg, while glass sheet 150 is being pulled down) and sends the image to anomaly detection computing device 160 .

[0029] The anomaly detection computing device 160 receives the image and applies a machine vision process to the image to determine the location of at least one object in the image. As described herein, the machine vision process may include generating features based on the received image and inputting the generated features into one or more trained machine learning or artificial intelligence models. Based on the input features, the trained machine learning or artificial intelligence model may generate output data representing one or more detected objects and / or the location of one or more detected objects. For example, the output data may represent the detected objects and their locations, and the objects may be droplets of molten glass within the housing 102 (e.g., dripping from the walls 123, 124 of the glass forming device 120 into the interior of the housing 102). In some cases, the object may be a crack in the glass sheet 150.

[0030] Based on the detected objects, the anomaly detection computing device 160 can determine at least one anomaly based on the position of at least one object in each of the images. For example, the anomaly detection computing device 160 can compare the positions of the detected objects in the respective images to determine whether the object is moving (e.g., a drop of molten glass falling, dripping), and based on its position in each image, it can be determined whether the object is falling through the housing 102. Since it is not expected that an object will fall through the housing 102, the anomaly detection computing device 160 determines that the object (e.g., a drop of molten glass falling through the housing 102) is an anomaly. In addition, the anomaly detection computing device 160 can generate anomaly detection data characterizing the detected anomaly, and the generated anomaly detection data can be stored in the database 170. In some cases, based on the anomaly detection data, the anomaly detection computing device 160 generates an alarm (e.g., displays an alarm, sends an alarm message to another processing device), generates an anomaly map, or moves the glass to a different output line, among other examples.

[0031] In some cases, applying a machine vision process to an image (such as an image captured by first camera 140) includes the following processing steps. Anomaly detection computing device 160 receives the image and processes the image to generate a differential image. For example, anomaly detection computing device 160 may compare each pixel value of the image with a corresponding threshold value. Pixel values ​​less than the threshold value are adjusted to a first value (e.g., 0). For example, the first value may correspond to a black pixel color.

[0032] In addition, the anomaly detection computing device 160 may determine the intersection and / or union between the differential image and the second image, and generate a matching image based on the intersection and / or union. The second image may be an image captured from the same or another camera among the first camera 140, the second camera 142, the third camera 144, and the fourth camera 146. For example, the second image may be an image of the glass sheet 150 captured from the same or different viewing angle at an earlier point in time. In addition, the anomaly detection computing device 160 may determine the connectivity between the pixels of the matching images to generate a first connection image. For example, the anomaly detection computing device 160 may determine the "4-connectivity" or "8-connectivity" between the pixels of the matching images to generate a first connection image.

[0033] In addition, the anomaly detection computing device 160 applies a closing process to the first connected image to generate a closed image. For example, the anomaly detection computing device 160 may dilate the first connected image and then may erode the dilated image to generate the closed image. The closing process allows small holes in the first connected image to be “filled” and preserves the shape and size of large holes and objects in the first connected image.

[0034] Then, the anomaly detection computing device 160 may determine the connectivity between the pixels of the closed image to generate a second connection image. For example, the anomaly detection computing device 160 may determine "4-connectivity" or "8-connectivity" between the pixels of the closed image to generate a second connection image.

[0035] In addition, the anomaly detection computing device 160 can determine the region of interest of the second connected image, and can generate an ROI image based on the region of interest of the second connected image. The region of interest can be a predetermined area in the image that may contain anomalies. For example, the database 170 may include one or more ROI maps that define the ROI of the image. The anomaly detection computing device 160 can obtain the ROI map from the database 170, and can apply the ROI map to the second connected image to generate the ROI image. For example, the pixel values ​​of pixels falling outside the ROI map can be adjusted to a specific value (e.g., 0), while the pixel values ​​of pixels falling within the ROI map are not adjusted.

[0036] In addition, the anomaly detection computing device 160 may perform an operation of scaling the ROI image. For example, the anomaly detection computing device 160 may center the ROI image by subtracting a scaled pixel value (e.g., a predetermined pixel value, such as a predetermined average pixel value) from each pixel value of each pixel of the ROI image to generate a centered image. The anomaly detection computing device 160 may store the centered image in the database 170.

[0037] Anomaly detection computing device 160 may perform the above operations on images received from one or more of first camera 140, second camera 142, third camera 144, and fourth camera 146, and in some examples, on images received within a time interval. The time interval may correspond to the time when glass sheet 150 is being pulled down. Based on the centered image, anomaly detection computing device 160 may generate anomaly detection data characterizing one or more anomalies within the centered image.

[0038] In some cases, when an anomaly is detected, the anomaly detection computing device 160 performs additional operations. For example, when an anomaly is detected, the anomaly detection computing device 160 may provide one or more centered images for display. In some examples, the anomaly detection computing device 160 may send a warning message. For example, the anomaly detection computing device 160 may send a warning message to a mobile device (e.g., a mobile phone, a smart phone, etc.) operated by a technician of the glass forming company. In response to receiving the warning message, the mobile device may display the warning message. In some examples, the anomaly detection computing device 160 may shut down the glass forming apparatus 120, or may disengage the pulling rollers 137A, 137B from pulling down the glass sheet 150 (e.g., sending a signal to the pulling rollers 137A, 137B to move the pulling rollers 137A, 137B away from the glass sheet 150, thereby removing contact with the glass sheet 150).

[0039] Figure 2 An exemplary anomaly detection computing device 200 is illustrated, such as Figure 1 Anomaly detection computing device 160. As shown, anomaly detection computing device 200 may include one or more processors 201, working memory 202, one or more input / output devices 203, instruction memory 207, transceiver 204, and display 206, all of which are operably coupled to one or more data buses 208. Data bus 208 allows communication between various devices. Data bus 208 may include wired or wireless communication channels.

[0040] The processor 201 may include one or more different processors, each having one or more cores. Each of the different processors may have the same or different structures. The processor 201 may include one or more central processing units (CPUs), one or more graphics processing units (GPUs), application specific integrated circuits (ASICs), digital signal processors (DSPs), etc. The processor 201 may be configured to perform the function or operation by executing a code embodying a specific function or operation stored on the instruction memory 207. For example, the processor 201 may be configured to perform one or more of any function, method, or operation disclosed herein.

[0041] Instruction memory 207 may store instructions that may be accessed (e.g., read) and executed by processor 201. For example, instruction memory 207 may be a non-transitory, computer-readable storage medium such as a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), flash memory, a removable disk, a CD-ROM, any non-volatile memory, or any other suitable memory.

[0042] In addition, the processor 201 can store data in the working memory 202 and read data therefrom. For example, the processor 201 can store a set of working instructions in the working memory 202, such as instructions loaded from the instruction memory 207. The processor 201 can also use the working memory 202 to store dynamic data created during the operation of the anomaly detection computing device 200. The working memory 202 can be a random access memory (RAM), such as a static random access memory (SRAM) or a dynamic random access memory (DRAM), or any other suitable memory.

[0043] The input-output device 203 may include any suitable device that can perform data input or output. For example, the input-output device 203 may include one or more of a keyboard, a touchpad, a mouse, a stylus, a touch screen, physical buttons, a speaker, a microphone, or any other suitable input or output device. The input-output device 203 may allow a user to provide input that selects or describes a preferred operating value, such as, for example, a camera position.

[0044] The display 209 may display the user interface 205. The user interface 205 may enable user interaction with the anomaly detection computing device 200. For example, the user interface 205 may be a user interface of an application that allows a user to input preferred operating values ​​to update the positions of cameras, such as, for example, the positions of the first camera 140, the second camera 142, the third camera 144, and the fourth camera 146. In some examples, the user may interact with the user interface 205 by accessing the input-output device 203. In some examples, the display 209 may be a touch screen, where the user interface 205 is displayed on the touch screen.

[0045] The transceiver 204 allows for communication with a network, such as the communication network 116. For example, the transceiver 204 may be connected to The processor 201 may be operable to receive data from or send data to a wireless network via the transceiver 204.

[0046] Figure 3Exemplary functions of the anomaly detection computing device 200, such as machine vision functions, are illustrated. As shown, the anomaly detection computing device 200 includes an image differentiator engine 302, an image matching engine 304, a connection engine 306, a closure engine 308, a ROI engine 310, and a centering engine 312. In some cases, one or more of the image differentiator engine 302, the image matching engine 304, the connection engine 306, the closure engine 308, the ROI engine 310, and the centering engine 312 include instructions stored in the instruction memory 207, which, when executed by the one or more processors 201, cause the one or more processors 201 to perform corresponding operations. In some examples, one or more of the image differentiator engine 302, the image matching engine 304, the connection engine 306, the closure engine 308, the ROI engine 310, and the centering engine 312 can be implemented with logic, such as digital logic, a state machine, an FPGA, an ASIC, or any other suitable logic.

[0047] The image differentiator engine 302 may receive the image data 301. For example, the image differentiator engine 302 may receive the image data 301 representing the captured image from one or more of the first camera 140, the second camera 142, the third camera 144, and the fourth camera 146. The image data 301 may represent an image of an interior portion of the housing 102 and a portion of the glass sheet 150. In addition, the image differentiator engine 302 may filter the image data 301 to generate differential image data 303 representing a differential image. For example, the image differentiator engine 302 may compare each pixel value of the image data 301 with a corresponding threshold value, and may adjust pixel values ​​less than the threshold value to a first value (e.g., 0) to generate the differential image data 303.

[0048] Additionally, image matching engine 304 can receive differential image data 303 and can determine an intersection and / or union between the differential image and a second image. As described herein, the second image can be an image captured from the same or another camera among first camera 140, second camera 142, third camera 144, and fourth camera 146. For example, the second image can be an image of glass piece 150 captured at an earlier point in time from the same or a different viewing angle. Based on the determined intersection and / or union, image matching engine 304 generates matching image data 305 representing the matching image.

[0049] The connection engine 306 receives the matching image data 305 and determines connectivity between pixels of the matching image to generate first image connection data 307 representing a first connected image. For example, the connection engine 306 may determine "4-connectivity" or "8-connectivity" between pixels of the matching image to generate the first connected image.

[0050] The closure engine 308 receives the first image connection data 307 and performs a closure process on the first image connection data 307 to generate closed image data 309 representing a closed image. For example, the closure engine 308 can dilate the first connected image and then erode the dilated image to generate the closed image. The connection engine 306 receives the closed image data 309 and determines the connectivity between the pixels of the closed image to generate the second image connection data 311 representing the second connected image. For example, the closure engine 308 can determine "4-connectivity" or "8-connectivity" between the pixels of the closed image to generate the second connected image.

[0051] The ROI engine 310 may receive the second image connection data 311, and may determine a region of interest of the second connection image based on a ROI map, such as a ROI map stored in the database 170. The ROI engine 310 may generate ROI image data 313 representing the ROI image based on the determined region of interest of the second connection image. For example, the ROI image may include adjusted pixel values ​​(e.g., adjusted to 0) of pixels that fall outside the ROI map.

[0052] In addition, the centering engine 311 can receive the ROI image data 313, and can scale the ROI image data 313 to generate a centered image data 315 representing a centered image. For example, the centering engine 311 can center the ROI image by subtracting the scaled pixel value from each pixel value of each pixel of the ROI image to generate the centered image. The centering engine 311 can store the centered image data 315 in the database 170. In some cases, the centering engine 311 provides the centered image data 315 to the display 209 for display.

[0053] Figure 4A and Figure 4B A view of an assembly 400 including a first camera 402, a second camera 404, and a third camera 406 is illustrated. The assembly 400 may correspond to the first end 151 of the housing 102 including the first camera 140 and the second camera 142. The assembly 400 may also correspond to the second end 153 of the housing 102 including the third camera 140 and the second camera 142. For example, the housing 102 may include the first assembly 400 at the first end 151 and the second assembly 400 at the second end 153. The second camera 404 may be mounted to the assembly 400 at a predetermined distance from the first camera 402. Likewise, the third camera 406 may be mounted to the assembly 400 at a predetermined distance from the second camera 404.

[0054] Figure 5 An example of a generated centered image is shown, such as Figure 3The centered image data 315 is an image of a feature. For example, as described herein, the anomaly detection computing device 160 can receive an image from one or more cameras positioned within the glass housing assembly and apply a machine vision process to the image to determine the location of at least one object in the image. Based on the location of the at least one object in the image, the anomaly detection computing device 160 can detect anomalies during glass forming. For example, Figure 5 As illustrated, the anomaly may be, among other examples, dripping molten glass 702. In this example, dripping of molten glass can be seen through the housing 701 in which the glass sheet is formed.

[0055] Figure 6 An exemplary method 800 is illustrated that may be performed by one or more computing devices, such as anomaly detection computing device 160. In some examples, exemplary method 800 may be performed by one or more processors, such as one or more processors distributed across one or more networks (e.g., processors within one or more cloud servers).

[0056] Beginning at step 802, images are received from cameras. For example, a first camera 140 can be positioned within the glass piece housing 102 and configured to capture images of the glass piece 150 being formed from a first end 151 of the housing 102. Additionally, a third camera 144 can be positioned within the glass piece housing 102 and configured to capture images of the glass piece 150 being formed from a second end 153 of the housing 102. Each of the first camera 140 and the third camera 144 can capture images during glass forming and can send the images to the anomaly detection computing device 160.

[0057] In addition, in step 804, anomaly detection data is generated based on applying a machine vision process to the image. For example, as described herein, the anomaly detection computing device 160 can filter each image to generate a corresponding differential image. For example, the anomaly detection computing device 160 can compare each pixel value of each image with a corresponding threshold. Pixel values ​​less than the threshold are adjusted to a first value (e.g., 0). In addition, the anomaly detection computing device 160 can determine the intersection and / or union between the differential image and the second image, and generate a matching image based on the intersection and / or union. In addition, the anomaly detection computing device 160 can determine the connectivity between the pixels of the matching image to generate a first connected image.

[0058] Additionally, anomaly detection computing device 160 may apply a closing process to the first connected image to generate a closed image. For example, anomaly detection computing device 160 may dilate the first connected image and then may erode the dilated image to generate a closed image. The closing process allows small holes in the first connected image to be “filled” and preserves the shape and size of large holes and objects in the first connected image. Anomaly detection computing device 160 may then determine connectivity between pixels of the closed image to generate a second connected image.

[0059] In addition, the anomaly detection computing device 160 can determine the region of interest of the second connected image, and can generate the ROI image based on the region of interest of the second connected image. In addition, the anomaly detection computing device 160 can perform an operation of scaling the ROI image. For example, the anomaly detection computing device 160 can center the ROI image by subtracting the scaled pixel value (e.g., a predetermined pixel value, such as a predetermined average pixel value) from each pixel value of each pixel of the ROI image to generate a centered image. Based on the centered image, the anomaly detection computing device 160 can generate anomaly detection data characterizing one or more anomalies within the centered image. For example, as described herein, based on the position of the object detected in each image, the anomaly detection computing device 160 can determine that the molten glass is dripping in the glass piece housing 102.

[0060] Continuing to step 806, if no anomaly is detected, the method returns to step 802, where additional images are received from the camera. However, if an anomaly is detected, the method proceeds to step 808, where one or more operations are performed based on the anomaly detection. For example, an image is displayed in step 810, such as a centered image (e.g., image 700) showing a drop of molten glass. In some examples, in step 812, a warning message is sent. In some examples, in step 814, a roller (e.g., pulling rollers 137A, 137B) of a glass forming apparatus (e.g., glass forming apparatus 120) can be disengaged from a glass sheet (e.g., glass sheet 150).

[0061] Although the method described above is with reference to the illustrated flow chart, it should be understood that many other ways of performing the actions associated with the method can be used. For example, the order of some operations can be changed, and some operations described can be optional.

[0062] In addition, the methods and systems described herein may be at least partially embodied in the form of computer-implemented processes and devices for practicing these processes. The disclosed methods may also be at least partially embodied in the form of tangible, non-transitory machine-readable storage media encoded with computer program code. For example, the steps of the method may be embodied in hardware, in executable instructions executed by a processor (e.g., software), or in a combination of the two. The medium may include, for example, RAM, ROM, CD-ROM, DVD-ROM, BD-ROM, hard drive, flash memory, or any other non-transitory machine-readable storage medium. When the computer program code is loaded into a computer and executed by the computer, the computer becomes a device for practicing the method. The method may also be at least partially embodied in the form of a computer, in which the computer program code is loaded or executed, so that the computer becomes a special-purpose computer for practicing the method. When implemented on a general-purpose processor, the computer program code segment configures the processor to create a special-purpose logic circuit. The method may alternatively be at least partially embodied in a dedicated integrated circuit for executing the method.

[0063] The foregoing is intended to illustrate, explain and describe the embodiments of the present disclosure. Modifications and adjustments to these embodiments will be apparent to those skilled in the art and may be made without departing from the scope or spirit of the present disclosure.

Claims

1. A glass forming system, comprising: a glass forming apparatus configured to downdraw molten glass to produce a glass sheet; at least one camera configured to capture an image of at least a portion of the glass forming apparatus and the glass sheet; as well as a computing device communicatively coupled to the at least one camera, the computing device configured to: receiving the image from the at least one camera; applying a machine vision process to the images, and determining a location of at least one object in each of the images based on applying the machine vision process to the images; determining at least one anomaly based on the location of the at least one object in each of the images: generating anomaly detection data based on the at least one anomaly; as well as The anomaly detection data is stored in a database.

2. The glass forming system of claim 1, wherein the computing device is configured to: generating a warning message based on the anomaly detection data; and The alert message is sent to another computing device.

3. The glass forming system of claim 1 , wherein the at least one camera comprises a first camera and a second camera, the first camera having a field of view from a first end of the glass forming apparatus and the second camera having a field of view from a second end of the glass forming apparatus.

4. The glass forming system of claim 3, wherein the at least one camera comprises a third camera and a fourth camera, the third camera having a field of view from the first end of the glass forming device and the fourth camera having a field of view from the second end of the glass forming device.

5. The glass forming system of claim 1, wherein the at least one camera is positioned within a housing of the glass forming apparatus.

6. The glass forming system of claim 1, wherein the at least one anomaly is dripping of molten glass.

7. The glass forming system of claim 1, wherein: To apply the machine vision process to the image, the computing device is configured to: generating features based on the image; inputting the features into a trained machine learning or artificial intelligence model: and Output data characterizing the at least one object and the position of the at least one object is generated.

8. The glass forming system of claim 1 , wherein each of the images comprises a plurality of pixels, and wherein, To apply the machine vision process to the images, the computing device is configured to, for each of the images: comparing each pixel value of each of the plurality of pixels to a corresponding threshold value; When the pixel value is less than the corresponding threshold, adjusting the pixel value of each of the plurality of pixels to generate a differential image; as well as The position of the at least one object is determined based on the differential image.

9. The glass forming system of claim 8, wherein: To apply the machine vision process to the image, the computing device is configured to: determining at least one of an intersection and a union between the differential image and a second image; generating a matching image based on the at least one of the intersection and the union; as well as The position of the at least one object is determined based on the matching images.

10. The glass forming system of claim 9, wherein: To apply the machine vision process to the image, the computing device is configured to: determining a first connectivity between pixels of the matching image to generate a first connected image; as well as The position of the at least one object is determined based on the first connected image.

11. The glass forming system of claim 10, wherein: To apply the machine vision process to the image, the computing device is configured to: applying a closing process to the first connected image to generate a closed image; determining a second connectivity between pixels of the closed image to generate a second connected image; and The position of the at least one object is determined based on the second connected image.

12. The glass forming system of claim 11, wherein: To apply the machine vision process to the image, the computing device is configured to: determining a region of interest of the second connected image; generating a ROI image based on the region of interest of the second connected image; as well as The position of the at least one object is determined based on the ROI image.

13. The glass forming system of claim 12, wherein: To apply the machine vision process to the image, the computing device is configured to: subtracting the scaled pixel value from each pixel value of each pixel of the ROI image to generate a centered image; as well as The position of the at least one object is determined based on the centered image.

14. The glass forming system of claim 1, wherein the computing device is configured to: comparing the position of the at least one object in each of the images; and It is determined that the at least one object is moving.

15. An apparatus comprising: a memory storing instructions; as well as a processor communicatively coupled to the memory, wherein the processor is configured to execute instructions for: receiving an image from at least one camera configured to capture an image of at least a portion of a glass forming apparatus and a glass sheet; applying a machine vision process to the images, and determining a location of at least one object in each of the images based on applying the machine vision process to the images; determining at least one anomaly based on the location of the at least one object in each of the images; generating anomaly detection data based on the at least one anomaly; as well as The anomaly detection data is stored in a database.

16. The apparatus of claim 15, wherein the at least one anomaly is dripping of molten glass.

17. The device of claim 15, wherein: To apply the machine vision process to the image, the processor is configured to execute instructions for: generating features based on the image; inputting the features into a trained machine learning or artificial intelligence model; as well as Output data characterizing the at least one object and the position of the at least one object is generated.

18. The apparatus of claim 15, wherein each of the images comprises a plurality of pixels, and wherein, To apply the machine vision process to the images, the processor is configured to execute, for each of the images, instructions for: comparing each pixel value of each of the plurality of pixels to a corresponding threshold value; When the pixel value is less than the corresponding threshold, adjusting the pixel value of each of the plurality of pixels to generate a differential image; as well as The position of the at least one object is determined based on the differential image.

19. The device of claim 18, wherein: To apply the machine vision process to the image, the processor is configured to execute instructions for: determining at least one of an intersection and a union between the differential image and a second image; generating a matching image based on the at least one of the intersection and the union; determining a first connectivity between pixels of the matching image to generate a first connected image; applying a closing process to the first connected image to generate a closed image; determining a second connectivity between pixels of the closed image to generate a second connected image; determining a region of interest of the second connected image; generating a ROI image based on the region of interest of the second connected image; subtracting the scaled pixel value from each pixel value of each pixel of the ROI image to generate a centered image; as well as The position of the at least one object is determined based on the centered image.

20. A method performed by at least one processor, comprising: receiving an image from at least one camera configured to capture an image of at least a portion of a glass forming apparatus and a glass sheet; applying a machine vision process to the images, and determining a location of at least one object in each of the images based on applying the machine vision process to the images; determining at least one anomaly based on the location of the at least one object in each of the images; generating anomaly detection data based on the at least one anomaly; as well as The anomaly detection data is stored in a database.

Images