It is possible to use different computer technology, parts of the system can be combined and integrated. Obviously, in many cases, not all parts are necessary. During the development, the division of machine vision systems into four generally accepted categories has stabilized, each of which is focused on a certain group of tasks. The characteristic features of individual categories of machine vision systems are summarized.However, it should be noted that the features and typical uses overlap and the individual categories converge with the development of the field. The categories relate to the commercial production of systems. Machine vision hardware vendors cannot build systems individually, but only need to have a few systems on offer that cover as many tasks as possible.
Camera sensor
It is shown what reduction of the overall scheme is typical for the simplest means of machine vision, the camera sensor. The entire system, including the camera and lighting, is usually integrated into one housing. The number of digital inputs and outputs is limited, and because the camera sensor has very limited programming options, the inputs and outputs have a fixed function (synchronization of image acquisition, good / bad output). Standard PNP or NPN transistor outputs are used with operating voltages up to 30 V and protections against polarity reversal and voltage peak damage when switching inductive loads, sometimes with galvanic isolation. A fast communication interface is usually only used to connect to the MMI to set job parameters. Sometimes the parameters are even set only from the built-in keyboard. The fieldbus interface is seldom available. The design of the camera sensor, including processor power and production costs, is subject to the purpose of creating a machine vision system for very easy use.
Smart camera
With smart camera, the complete machine vision system is already. Its characteristic features are compactness and the use of a powerful single-chip microcomputer as an evaluation unit. Currently, a Texas Instruments DSP digital signal processor with a clock frequency of 400 MHz is standard on smart cameras, and 1 GHz RISC processors are also emerging. The program is stored in flash memory. It is usually built-in, but external MM Cards or SD Cards are also being used. The capacity of the RAM memory is determined mainly by the required size of the program, which after initialization is copied from the program flash memory to the RAM and is run here. So a smart camera with 128 MB of RAM is no exception.
Universal bit interfaces, which can optionally be set as inputs or outputs, are often used as digital inputs and outputs. They comply with industry standards, are in the design with an open collector of the PNP or NPN type, with an operating voltage of up to 30 V and with appropriate protections. Galvanic isolation of inputs and outputs from the supply voltage or their mutual separation often does not occur in smart cameras.
Smart cameras are usually used locally, on devices with a single power supply, where there are no major potential differences. The continuing miniaturization of these cameras often forces manufacturers to deviate from the standards common in industrial electronics. Cameras with inputs and outputs in TTL signal levels are also produced, which still need a galvanically separated level converter to connect to the production process, mostly with an external power supply. Other times, separate inputs and outputs are completely missing, and an I / O adapter connected to the communication interface must be used.
The communication interface almost always performs several functions in a smart camera. The main one is usually the connection to the MMI, which allows you to program the camera. Currently, a standard PC usually works as an MMI. For convenient task development, it is advisable for the MMI to display the image captured by the camera in real time. The transmission speed must then be relatively high, which is why Ethernet is most often used as the basic communication interface. It can also easily transfer data to a higher-level control system or mediate communication between several cameras. It is usually used for servicing, camera firmware replacement and similar purposes.
Sometimes, in addition to the usual Ethernet, the smart camera is also equipped with an auxiliary serial interface RS-232/422/485. The reason is easier connection to most existing PLCs. Interfaces for higher types of industrial buses do not yet appear commonly with these cameras, but the camera firmware sometimes allows access to the industrial bus via a suitable converter from Ethernet.
The number of smart cameras offered on the market is still growing, it seems that, as in the USA, cameras of this class will mostly be used in Europe.
PC systems
As indicated by DZOptics, PC systems are characterized by the use of a personal computer with a standard operating system in place of the evaluation unit. Another characteristic feature is modularity. The camera is separate and most systems allow the connection of several cameras. Cameras with analog output are connected via a frame grabber plug – in card), cameras with digital output via a special camera interface or using high-speed Ethernet. Computing power and memory size are determined by the computer used. Inputs and outputs as well as communication interfaces are implemented in the usual way in PC technology, most often using plug-in cards. Therefore, the scope of delivery of the PC system is usually only a set of software that contains an environment for the development of a machine vision task and a run-time module, often protected by a hardware key. It is also usual to supply a digitizer card or interface for a digital camera, for which the program is optimized. Recently, there are also specialized industrial PCs with an interface for connecting a camera, this solution is sometimes called an embedded vision system.
PC systems were at the beginning of the machine vision era, but are now beginning to give way to smart cameras. In order to really enjoy all the benefits of a PC solution, it is necessary to use a commercially available operating system. However, the machine vision system basically processes the data and performs the control in real time. A commercial operating system is unsuitable for this purpose. In addition, the computing power of smart cameras can already withstand comparisons with a classic PC. The domain of PC systems is rather large-scale tasks requiring the processing of images from several cameras.
Customer systems
This category includes tailor-made systems where none of the mentioned variants can be used. These are usually tasks that require high processing speed, special algorithms or non-standard cameras. The design of the customer system corresponds to the given purpose. Fast signal processors or gate arrays with custom software are most often used for implementation.
