This interconnection within one program environment significantly simplifies not only the development of new applications, but also the resulting solution is cheaper than other concepts. In most cases, we can use a single computer to which both cameras and industrial input / output units are connected and which is also connected to a computer network and company information system. This computer usually runs the graphical user interface of the application, the database SQL server, the web server, visual inspection and machine vision tasks, as well as the entire logic of the automation application.

If additional remote automation racks are needed, each can be connected via Ethernet and TCP/IP networks. An example of such a remote switchboard is shown in Fig. 2. The efficiency of the solution and the advantageous price of all hardware and software is obvious in such conceived systems.

With the advent of the new generation of Control Web 8, several features have been added to further enhance machine vision integration capabilities.

A new type of data data objects for transferring general binary data has been added to machine vision. This type corresponds to data elements of the data type in the Control Web environment. Virtual instruments in the application program and steps in the machine vision chain can now efficiently exchange blocks of arbitrary data. The type of data and the format of its storage is not prescribed in any way. In some tasks, such as further work with detected significant image points in an application program, this technique of transferring binary data blocks brings significant acceleration. New possibilities of cooperation of automation application with a chain of machine vision steps can be well demonstrated eg by virtual instruments for displaying spatial scene in the form of 3D object. The image data is captured by a stereoscopic pair of cameras. This virtual instrument has several peculiarities. It depicts a scene captured by cameras, but does not communicate directly with the cameras itself. It needs not only two images from cameras, but also information about feature points and their descriptors found in these two images. It must therefore cooperate with machine vision, which, by means of the point detection step, fills data elements of the data type with found descriptors of these points. Similar to the point descriptors, the actual image from each camera is transmitted to the virtual instrument via data elements of the data type. A virtual instrument can be very useful in applications where we need information about the distances of objects in an image, when we have to find objects at different distances or avoid collisions with objects in space while moving. By means of data elements of the data type, any image can also be transmitted, for example, to display devices located anywhere in the 3D space of the scene.

We can also use VisionLab's ability to locate feature image points and calculate descriptors describing their surroundings efficiently and with maximum use of massively parallel GPU performance, to detect general objects in images. Feature points are searched on a scale of several scales, then objects can be detected independently of their rotation and brightness, but also independently of their size. A set of descriptors that characterize objects that we want to identify further in the image data can provide detection steps in both files and data elements of the data type.

Now that the new eighth generation of Control Web is available, it is also worth noting that the development version of Control Web can be downloaded free of charge from www.controlweb.eu and used on an unlimited number of free installations. The license is only required if the application is required to run continuously. All VisionLab drivers and vision systems can be used as well.

_{Roman Cagaš}