 Good day everyone. I'm assistant professor Rishad Mistry from the mechanical engineering department at Wachan Institute of Technology Sholapur. And in this video I'll be talking about machine vision. So this topic in fact is a part of the series industrial robotics and this covers a very important aspect of modern day industrial robots, that is machine vision or robot vision. So the learning outcomes of this session is that the student will be able to define machine vision and its associated terms and explain fundamentals of image processing. So if you actually look up definition for robot vision or machine vision or computer vision in different textbooks or online, you will come across different definitions. I have seen textbooks which use these terms interchangeably as well. So when they say robot vision or computer vision or machine vision, it basically means one and the same thing. So like I said, different sources can give information which is a bit different. And that's why it is more prudent in fact to look at different sources and include in fact all of these. So on the basis of whatever I have read in textbooks and online, this is how I would sort of define robot vision, computer vision and machine vision. So robot vision may be defined as the process of extracting, characterizing and interpreting information from images of a 3D world. In fact this applies to machine vision or computer vision as well. I have seen another definition for computer vision that is it's the study and application of methods which allows a computer to understand the content of the image. And when we talk about machine vision in particular, it's the application of computer vision to industry and manufacturing. So in fact I would say robot vision would be a specific subset within machine vision and machine vision itself would be a subset of computer vision. So in fact robot vision I would define as very specifically the application of computer vision or machine vision in this sense. For robot control tasks either with industrial joint arm manipulators or with mobile robots. Then again this is again it depends upon what the source of your definition is. So what is typically machine vision and what is its scope? What are the different areas of science and technology that it covers? So computer vision or machine vision as indicated in this figure typically includes all these things. So in fact for someone who really wants to make a good career in this particular field he requires understanding of control systems. A bit of artificial intelligence nowadays. Machine learning, mathematics, the mathematics specifically behind vision and signal processing because this is mainly all signal processing. The physics and optics involved because this is where all the cameras come into picture. And to a certain extent if you're looking at very specific application then you need a certain amount of background knowledge regarding that application. For example here is neurobiology if you're looking at physics and atrophysics then in fact that would also come into place. So it's a very multidisciplinary field to begin with. So what is basically machine vision and what it involves? So it involves sensing and acquisition of the image. Followed by what is called as image processing which again may have sub steps within it such as processing, segmentation, description of the image, recognition of a pattern and then interpreting that. And then finally outputting the results either to a control system or to an operator who can take necessary decisions. So when it comes to machine vision like I said sensing is the process which will yield the visual image. Pre-processing is a stage within the image processing which will do noise reduction and enhancement of these details. Segmentation will divide the image into objects of interest and description basically it is involving competition of features such as size, shape, etc. And then recognition involves classifying this and identifying those specifics. So this is in fact a part of the image processing, image analysis and other stuff. In fact I use the word image processing in order to indicate a lot of things that you do with an image using a computer. In fact textbooks on image processing have their own vocabulary and own nomenclature as to what should be called image processing, what is image analysis, what is and so on and so on. Again we continue with our discussion regarding the machine vision process. The process of vision system actually starts with the image acquisition in which the representation of the image data and digitization is accomplished. This is obviously done with the camera. Image sensing is the next step to obtain a proper image from the scene. So now you obtain an image from the scene. Then we digitize that particular image and in the last step the image processing in which a more suitable image is actually prepared. So this is how in fact, in fact a lot of things goes on what appears to us in a simple click. In fact all these things typically happen in that. Now can you think of potential applications of machine vision and automation and manufacturing based on whatever you studied in second and third year of engineering. In fact there is a set up laboratory setup which could definitely use a machine vision system as a part of it. Which is the setup that I'm talking about. Let's see if this rings a bell for you. Let's look into components of a machine vision. So a components of a machine system typically in fact the first part is the camera itself which may be a CCD camera or a CMOS camera. In machine vision applications still CCD cameras are preferred but CMOS cameras are catching up. Then you have specialized light sources with specialized illumination. Then you have the interface for the camera. You have the main processor. You may have a dedicated signal processor for the same. And then you have the device input output communication systems and the software for detecting features and so on. So this is typically what comprises of a complete machine vision system. So it's not just the camera. The camera is one part of it. It includes the lighting. It includes all the accessories. It includes the interfaces. It includes the cables and also the software. So let us discuss some of these components now. There are plenty of good resources available online for machine vision. In fact the manufacturers themselves have some of the best resources that you can find. One of these which I have used. There are others such as Teledyne, Dalsa, Basler and even a lot of Japanese manufacturers. They have very good products and the information contained on their website is very informative, precise and very up to date as a matter of fact. In fact what I have noticed is that some industrial robotics textbook, the information that is contained about machine vision systems is really very old. So in fact these are talking about Vidicon cameras which go back to the 70s. Okay, nobody uses these today. So it is in fact more prudent to seek information from the manufacturers themselves rather than relying on textbooks which typically date back to the 80s and 90s and unfortunately the contents have not been revised in many of these textbooks. So this actually is a typical machine vision camera and this is one of the most critical aspect of the machine vision system. Cameras typically like I said are CCD cameras charge couple devices or CMOS cameras. When it comes to machine vision application let me iterate. A market study which I read a couple of years back indicate that CCD cameras still have the higher market share when it comes to machine vision applications especially space and other defense applications. But CMOS has definitely caught up. When I started looking into the subject in around 2012-13 and by 2018 in fact there was a significant increase in market share when it comes to CMOS cameras. Like I said when it comes to laptops and mobile applications and even SLRs CMOS cameras have pretty much taken over the market. Then an important aspect of machine vision in fact we will be discussing this in detail in a different video is regarding lighting. So lot of information is available regarding what are the different types of lighting classification how to accomplish this. But there is no vision system without the necessary and proper lighting technique based on the application. So this is one example of the lighting system which I have given over here. Another one is in this particular site is called as backlight technique. In fact this website has an interactive set of which tells you how this profile is going to look like in the camera. So like I said more details will be discussed in the dedicated video for lighting. It's a very important aspect of the machine vision system. Another important aspect of machine vision is the software. Nothing can happen without the software. You have open source software based on Python which lot of people widely use. You have open CV, open machine vision and so on and so forth. But many manufacturers have proprietary software which they have fine tuned for certain applications. And in any industrial setup it's highly unlikely that you will find open source software. Though they are extensively used in research work but most industrial applications do deal with proprietary software. So what are the applications of machine vision? In fact we have a dedicated video for this. But generic domains would be control for autonomous mobile robots for quality control applications in industry. Retail automation for example bar core readers. In agricultural automation it's a very upcoming field for detecting plant diseases and so on. Railway track inspection for cracks and flaws is a very specialized application of machine vision. And then you have new developments which involve development of vision systems for blind people. Another applications are biometrics like facial recognition, positioning systems, safety systems for industrial environments. Extensive use in inspection and quality control and like I said in inventory control, counting, etc. Bar code reading especially in supply chain and inventory management. So when it comes to books and references, standard textbooks have good information regarding this. But I would definitely encourage you to look up websites of machine vision manufacturers. These have very good and very relevant information when it comes to cameras and algorithms for machine vision. And in fact a lot of other things. With this we'll call it a day and in the next video we'll talk in more detail about machine vision applications. Thank you.