What Should be Kept in Mind While Choosing a Smart CameraOver the last decade, a new space has opened up in the machine vision market for Smart Cameras, exploitation of which is increasing remarkably. Currently the advanced smart cameras can replace the more massive and premium Machine Vision Systems in many applications. On the other hand, many of the PC-based Machine Vision Systems are getting more competitive with the smart cameras also in price terms due to the ever-decreasing cost and advancement of new technologies (i.a. FireWire, USB) So how should we act if we want to choose the right kind of solution for applications demanding machine vision? Smart Cameras compared to PC-based Machine Vision Systems Smart cameras are described as fully independent devices not exploiting PC-, VME- or some other similar kind of architecture. The most eminent differences between smart cameras and PC-based machine vision systems are in the following features: Speed: as the PC-based machine vision systems use remarkably faster processors, they usually remain faster than smart cameras. Tool accuracy: many of the algorithms used by PC-based machine vision systems are more accurate than the ones in smart cameras. Still, considering the positioning tools, the advanced smart cameras are capable of reaching sub-pixel accuracy in most of the applications. The use of multiple cameras: in case of applications that demand several cameras, the PC-based machine vision system can be more cost-effective than the usage of several smart cameras in the same application. Specialized cameras: if an application demands a high resolution camera, a 3-chip-colour camera, an UV sensitive camera or e.g. a thermal camera, the realization option will most probably be a PC-based machine vision system. Usage of own algorithms: for some applications it might be useful to implement algorithms developed by the client. Although some smart cameras enable the implementation of own algorithms, the choice will usually be a PC-based vision system if application- or client-specific tools have to be implemented with the application. Despite the limitations mentioned above it can safely be said that smart cameras can be a true alternative to most of the machine vision applications that have traditionally been implemented by taking advantage of the PC-based machine vision systems. With the advancing capabilities of the equipment and tools, the gap between the solutions will grow constantly smaller. The most remarkable advantage of smart cameras is perhaps their ease of use. This applies most of all to users or businesses with no previous experience in implementing the machine vision and then the smart camera would be a recommendable starting option. Selecting a Smart Camera Vendor While selecting equipment to the application, people most commonly compare only the technical characteristics - frame rate, camera resolution, memory limit, etc based on datasheets. With the machine vision systems, the suitability of the equipment for the purpose depends mostly on the power and sophistication of algorithms and tools. Since the different vendors use different approaches and algorithms even for similarly named tools, it is almost impossible to compare the tools only on the bases of datasheets. Under these conditions, almost the only way for true evaluation is to evaluate different application solutions with the piece in question. The most typical differences you should address to are, among others, differences in position and orientation, inconsistent lighting, blur and process variations as well as reflections and shadows. The other noteworthy aspect to keep in mind is user-friendliness and the simplicity of programming the system. When the smart camera should be integrated as part of a production process, it is important to address the connection of the system to the control- and higher-level systems. In addition to the basic interfaces (I/O, RS232), there is a good reason to sort out the Ethernet- interface and different supported protocols (FTP, Telnet, SMTP, Modbus TCP, Ethernet IP) as well as the possible connections to the field buses. The availability of support services is also an important factor in selecting the system. It is important to know whether the vendor can provide professional technical support for the solution and whether they provide local support. Sometimes it is necessary to ensure also the availability of global technical support. When you have progressed this far in your evaluation of equipment, it is often best to select the system that offers the widest, most effective and reliable tools instead of choosing the least expensive solution. There are several reasons for this recommendation, especially when implementing the very first machine vision application. The first and probably the most important thing is that later on – almost certainly – further requirements will arise to add more functionality to the system. If the system solution is selected only to fulfil the originally determined requirements, you will soon find out that the device cannot be expanded to new application targets. As far as price is concerned, the difference between the cheapest and most expensive smart cameras is about EUR 4000. The cheapest application solution can quickly turn out to be a more expensive one if the used tools are not compatible with the application and deployment requirements. Applications development with a Smart Camera Now that you have selected your equipment, it is time to start building the application. There are two ways of doing it - you can either outsource the job and hire a vision system integrator to deploy your system or you can handle it yourself. If you are going to outsource the deployment, you might want to consult your vendor. Usually the vendors can recommend a company who is familiar with the special requirements of the application field or domain. If you choose to handle the deployment yourself, the first step towards successful result is to get educated on the chosen system. This can save considerable amount of time during the first stages of building the application and, above all, create better conditions for correct deployment. Before starting the deployment process, it is necessary to choose illumination and optics for the respective application. For instance, a reliable vendor can determine the necessary illumination method and optics by completing a pre-study. They are likely to have previous experience with application of such type and can therefore recommend the best working solutions. It is useful to conduct the deployment by starting with offline tests for instance in laboratory environment. You should try to imitate the production conditions while the system in running offline by regulating the lighting and optics. Different tools and various approaches should be used in testing and eventually choose the most reliable ones. The possibility to save the rejected test images for example over the Ethernet-network is a very useful function in the most advanced vision sensors available. Taking advantage of this opportunity will allow determining if the rejected products were in fact defective and make the necessary modifications in the parameters of the tools. Recognizing the risks If properly exploited, the machine vision is a very powerful technology that can remarkably reduce manufacturing costs, improve the quality of products, increase productivity, and, in some cases, even solve applications that previously would have seemed impossible. However, machine vision is also a technology that is not always easy to apply and there will always remain a possibility of miscalculations. With today’s smart cameras costing EUR 3000 – 8000, the economical risks are quite small. Still, the risks of application failures are still there. Smart cameras or machine vision systems of different vendors are often regarded as identical because of their characteristics or performance. The prevailing idea is often that if an application cannot be solved with one “camera”, it cannot be solved with another. However, in reality the equipment spectrum is wide and there are big differences in their performance and, further on, the compatibility to different applications.
Smart cameras can be extremely powerful tools and fulfil a wide range of different applications. Because of their low purchase costs and relatively easy programming, they are well suitable for beginners as well as machine vision experts.
|
