Computer vision is a technology that is present both in industry and in other aspects of our daily lives. For example, the facial recognition that our mobiles perform to unlock themselves or the automatic license plate reading that open the way for us in some public car parks without having to enter the ticket. Next, we will explain some aspects of computer vision to delve a little more into what it is and what it is used for.

What is computer vision?

Computer vision is a process by which a machine, usually a computer, can analyze an image, understanding it and understanding the visual information it contains. This process is carried out in a similar way to that of the human body. Our eyes are the cameras that capture the images and that are subsequently analyzed by our brain. As in the process carried out by our body, other variables such as lighting are very important in computer vision systems. Our eyes see things differently depending on the light that falls on them. The same thing happens with computer vision systems.

Since the 70s, many resources have been dedicated to the evolution of computer vision, creating new, efficient and robust algorithms, which we can currently find applied in industry and other sectors.

We can distinguish two different approaches to computer vision. On the one hand, traditional computer vision in which image processing techniques are used, such as feature extraction and, on the other hand, deep learning or neural networks. Here are more details about these two approaches.

 

Traditional approach to machine vision vs deep learning

Traditional machine vision algorithms focus on detecting and extracting features from the image, for example edge detection. The edges are characteristic points of the image since they are areas with a lot of texture, with color changes and that are relatively easy to identify. It is important to identify the characteristics of an image since they are what will allow us to obtain the information and understanding we need about it. From these characteristics, we can perform an analysis and identify possible objects in it. There are many algorithms developed over the years that allow the extraction of different types of characteristics. Therefore, the efficiency of these techniques depends on the quality of the extracted characteristics.

However, the evolution of computer vision has led to deep learning algorithms and neural networks. Deep learning algorithms in machine vision perform the image feature extraction process automatically.

These algorithms allow images to be classified into different classes, for example distinguishing between images of cats and dogs or distinguishing between a person and a car in an image. In addition to classifying images in one category or another, these algorithms can also locate various objects within an image.

For the generation of image detection and classification models, it is necessary to have a labeled image dataset. From these samples, the algorithm learns to distinguish the characteristics of each of the classes automatically and, later, it is able to identify them in a new image.

All these algorithms have many different applications, some of which, we will tell you below.

What is machine vision used for?

Computer vision allows automated inspections of parts or objects, which is very useful in quality processes. In addition to quality processes, there are many other applications in industry.

Computer vision is also used to read codes, both industrially and in everyday life. Currently, there are many cafes or restaurants that present their menu through a QR code. Our mobiles also have an computer vision algorithm to read them.

At the other extreme, computer vision contributes to autonomous driving. The computer vision algorithms allow to identify in real time, the position of the objects that surround the vehicle and not only to identify them but also to locate them in space, which facilitates the interaction with the environment that surrounds it.

¿Necesitas ayuda?

¡Ponte en contacto con nosotros! 
Contacto
close-link