Contenido

Los usos principales de la visión artificial son la inspección y clasificación automáticas basadas en imágenes y la guía del robot.;[6]​[7]​ en esta sección la primera se abrevia como "inspección automática". El proceso general incluye la planificación de los detalles de los requisitos y el proyecto, y luego la creación de una solución.[9]​[10]​ Esta sección describe el proceso técnico que ocurre durante la operación de la solución.

Methods and sequence of operation.

The first step in the automatic inspection operation sequence is the acquisition of an image, typically using cameras, lenses, and lighting that have been designed to provide the differentiation required by subsequent processing.[11]​[12]​ MV software packages and the programs developed on them employ various digital image processing techniques to extract the required information and often make decisions (such as pass/fail) based on the extracted information.[13]​ The components of an automatic inspection system typically include lighting, a camera or other image generator, a processor, software and output devices.[7]​.

Images

The imaging device (for example, a camera) may be separate from or combined with the main image processing unit, in which case the combination is generally called a smart camera or smart sensor.[14]​[15]​ The inclusion of the entire processing function in the same enclosure as the camera is often called embedded processing.[16]​ When separated, the connection may be made to specialized intermediate hardware, a custom processing device, or a frame grabber within a computer via an interface Analog or standardized digital (Camera Link, CoaXPress). multispectral, hyperspectral imaging, multi-band infrared imaging,[22]​ linear scan imaging, 3D surface imaging, and

These cameras need to have very particular specifications for this application, since they need to be in use for long periods of time and often in a small space. Additionally, these cameras normally have high definition and FPS to capture dimensions and errors in production. As mentioned above, interfaces such as CoaXpress are used to transfer this information efficiently. These interfaces can reach speeds of 12.5Gbps (gigabits per second) per transmission channel.[24]​ "Garbage in, Garbage out" is a computer science concept that is applied here. Cameras require these features, since without them, mediocre results would be produced. High speed and image quality are key to machine vision. [25]​ Environmental conditions such as temperature, shock, etc. They also need to be considered in the industrial context. [26]​.

Although the vast majority of machine vision applications are solved using two-dimensional images, machine vision applications that use 3D images are a growing niche within the industry.[27]​[28]​ The most widely used method for 3D imaging is scan-based triangulation, which uses the movement of the product or image during the imaging process. A laser is projected onto the surfaces of an object. In machine vision, this is accomplished with a scanning motion, either by moving the workpiece or by moving the camera and laser imaging system. The line is seen by a camera from a different angle; Deviation from the line represents variations in shape. Lines from multiple scans are assembled into a depth map or point cloud.[29]​ Stereoscopic vision is used in special cases involving unique features present in both views of a pair of cameras.[29]​ Other 3D methods used for computer vision are time-of-flight and grid.[29]​[27]​ One method is grid array-based systems that use a pseudo-random structured light system such as that employed by the Microsoft Kinect system. around 2012.[30][31]​.

Image processing

Once an image is acquired, it is processed.[19]​ Core processing functions are typically performed by a CPU, GPU, FPGA, or a combination of these.[16]​ Deep learning inference and training impose higher processing performance requirements.[32]​ Typically, multiple processing stages are used in a sequence that ends up as a desired result. A typical sequence might begin with tools such as filters that modify the image, followed by extracting objects, then extracting (e.g., measurements, reading codes) data from those objects, followed by communicating that data or comparing it to target values ​​to create and communicate “pass/fail” results. Computer vision image processing methods include;

Outputs

A common output of automatic inspection systems is pass/reject decisions.[13] These decisions can, in turn, trigger mechanisms that reject failed items or sound an alarm. Other common outputs include object position and orientation information for robot guidance systems.[6]​ In addition, output types include numerical measurement data, data read from codes and characters, object counts and classification, process or result visualizations, stored images, alarms from automated MV spatial monitoring systems, and process control signals.[9]​[12]​ This also includes user interfaces, interfaces for multi-component system integration, and data exchange. automated.[43]​.

The term "deep learning" has varying meanings, most of which can be applied to techniques used in computer vision for over 20 years. However, use of the term in machine vision began in the late 2010s with the advent of the ability to successfully apply such techniques to full images in the industrial machine vision space.[44] Conventional machine vision typically requires the "physics" phase of an automated machine vision inspection solution to create simple and reliable differentiation of defects. An example of "simple" differentiation is that the defects are dark and the good parts of the product are light. A common reason why some applications were not feasible was when "simple" was impossible to achieve; Deep learning eliminates this requirement, essentially "seeing" the object more like a human does, which now makes it possible to achieve such automatic applications.[44]​ The system learns from a large number of images during a training phase and then performs inspection at runtime, called "inference".[44]​.

Machine vision commonly provides location and orientation information to a robot to allow the robot to correctly grasp the product. This capability is also used to guide motion that is simpler than robots, such as a 1- or 2-axis motion controller.[6]​ The overall process includes planning the details of the requirements and project, and then creating a solution. This section describes the technical process that occurs during the operation of the solution. Many of the process steps are the same as with automatic inspection, except that they focus on providing position and orientation information as a result.[6]​.

As recently as 2006, an industry consultant reported that MV represented a $1.5 billion market in North America.[45]​ However, the editor-in-chief of a MV trade magazine stated that "computer vision is not an industry in itself," but rather "the integration of technologies and products that provide services or applications that benefit real industries such as automotive or consumer goods manufacturing, agriculture, and defense."[4]​.

Contenido

Los usos principales de la visión artificial son la inspección y clasificación automáticas basadas en imágenes y la guía del robot.;[6]​[7]​ en esta sección la primera se abrevia como "inspección automática". El proceso general incluye la planificación de los detalles de los requisitos y el proyecto, y luego la creación de una solución.[9]​[10]​ Esta sección describe el proceso técnico que ocurre durante la operación de la solución.

Methods and sequence of operation.

The first step in the automatic inspection operation sequence is the acquisition of an image, typically using cameras, lenses, and lighting that have been designed to provide the differentiation required by subsequent processing.[11]​[12]​ MV software packages and the programs developed on them employ various digital image processing techniques to extract the required information and often make decisions (such as pass/fail) based on the extracted information.[13]​ The components of an automatic inspection system typically include lighting, a camera or other image generator, a processor, software and output devices.[7]​.

Images

The imaging device (for example, a camera) may be separate from or combined with the main image processing unit, in which case the combination is generally called a smart camera or smart sensor.[14]​[15]​ The inclusion of the entire processing function in the same enclosure as the camera is often called embedded processing.[16]​ When separated, the connection may be made to specialized intermediate hardware, a custom processing device, or a frame grabber within a computer via an interface Analog or standardized digital (Camera Link, CoaXPress). multispectral, hyperspectral imaging, multi-band infrared imaging,[22]​ linear scan imaging, 3D surface imaging, and

These cameras need to have very particular specifications for this application, since they need to be in use for long periods of time and often in a small space. Additionally, these cameras normally have high definition and FPS to capture dimensions and errors in production. As mentioned above, interfaces such as CoaXpress are used to transfer this information efficiently. These interfaces can reach speeds of 12.5Gbps (gigabits per second) per transmission channel.[24]​ "Garbage in, Garbage out" is a computer science concept that is applied here. Cameras require these features, since without them, mediocre results would be produced. High speed and image quality are key to machine vision. [25]​ Environmental conditions such as temperature, shock, etc. They also need to be considered in the industrial context. [26]​.

Although the vast majority of machine vision applications are solved using two-dimensional images, machine vision applications that use 3D images are a growing niche within the industry.[27]​[28]​ The most widely used method for 3D imaging is scan-based triangulation, which uses the movement of the product or image during the imaging process. A laser is projected onto the surfaces of an object. In machine vision, this is accomplished with a scanning motion, either by moving the workpiece or by moving the camera and laser imaging system. The line is seen by a camera from a different angle; Deviation from the line represents variations in shape. Lines from multiple scans are assembled into a depth map or point cloud.[29]​ Stereoscopic vision is used in special cases involving unique features present in both views of a pair of cameras.[29]​ Other 3D methods used for computer vision are time-of-flight and grid.[29]​[27]​ One method is grid array-based systems that use a pseudo-random structured light system such as that employed by the Microsoft Kinect system. around 2012.[30][31]​.

Image processing

Once an image is acquired, it is processed.[19]​ Core processing functions are typically performed by a CPU, GPU, FPGA, or a combination of these.[16]​ Deep learning inference and training impose higher processing performance requirements.[32]​ Typically, multiple processing stages are used in a sequence that ends up as a desired result. A typical sequence might begin with tools such as filters that modify the image, followed by extracting objects, then extracting (e.g., measurements, reading codes) data from those objects, followed by communicating that data or comparing it to target values ​​to create and communicate “pass/fail” results. Computer vision image processing methods include;

Outputs

A common output of automatic inspection systems is pass/reject decisions.[13] These decisions can, in turn, trigger mechanisms that reject failed items or sound an alarm. Other common outputs include object position and orientation information for robot guidance systems.[6]​ In addition, output types include numerical measurement data, data read from codes and characters, object counts and classification, process or result visualizations, stored images, alarms from automated MV spatial monitoring systems, and process control signals.[9]​[12]​ This also includes user interfaces, interfaces for multi-component system integration, and data exchange. automated.[43]​.

The term "deep learning" has varying meanings, most of which can be applied to techniques used in computer vision for over 20 years. However, use of the term in machine vision began in the late 2010s with the advent of the ability to successfully apply such techniques to full images in the industrial machine vision space.[44] Conventional machine vision typically requires the "physics" phase of an automated machine vision inspection solution to create simple and reliable differentiation of defects. An example of "simple" differentiation is that the defects are dark and the good parts of the product are light. A common reason why some applications were not feasible was when "simple" was impossible to achieve; Deep learning eliminates this requirement, essentially "seeing" the object more like a human does, which now makes it possible to achieve such automatic applications.[44]​ The system learns from a large number of images during a training phase and then performs inspection at runtime, called "inference".[44]​.

Machine vision commonly provides location and orientation information to a robot to allow the robot to correctly grasp the product. This capability is also used to guide motion that is simpler than robots, such as a 1- or 2-axis motion controller.[6]​ The overall process includes planning the details of the requirements and project, and then creating a solution. This section describes the technical process that occurs during the operation of the solution. Many of the process steps are the same as with automatic inspection, except that they focus on providing position and orientation information as a result.[6]​.

As recently as 2006, an industry consultant reported that MV represented a $1.5 billion market in North America.[45]​ However, the editor-in-chief of a MV trade magazine stated that "computer vision is not an industry in itself," but rather "the integration of technologies and products that provide services or applications that benefit real industries such as automotive or consumer goods manufacturing, agriculture, and defense."[4]​.