The Computer Vision includes all those algorithms and applications that allow a normal computer equipped with optical sensors (cameras, lasers, infrared, ...) to visually perceive the space around it and behave accordingly. Applications range from image processing to video analysis, from three-dimensional perception to robot stand alone navigation.
After some pioneering studies in the early 1970s, computer vision was immediately associated with the concept of artificial intelligence, particularly robot related (motion estimation, optical flow, visual odometry, ...). In the 1980s and 1990s, attention was shifted to more stringent mathematical formulas, such as contour extraction, 3D reconstruction, and photogrammetry. From the years 2000 onwards, numerous researchers have joined efforts to study even more advanced techniques, such as wavelet analysis, invariant features, etc...
Today Computer Vision finds space in many application areas, both in specialized environments and in everyday life. Just think about all the algorithms that are designed to process photos taken by our smartphones: facial recognition, HDR images, panoramas, photo filters, etc ...
In recent years, great steps have also been made with 3D cameras, just think of Microsoft's Kinect or Google's Project Tango, which have allowed end users to access localization technologies, augmented reality, and very advanced recognition.
The last frontier is the use of techniques based on artificial intelligence, in particular all those within the category of Deep Learning (deep neural networks). These allow you to analyze the images by following a perceptual and computational process similar to that of the retina and the human brain. There are many applications that automatically accomplish tasks that until some years ago were unthinkable: generic image labeling, black and white photo coloring, etc. ... Other important applications of these algorithms are synthesis and voice recognition, The understanding of natural language, bioinformatics, up to the creation of artificial chess players and Go.
In Zucchetti Centro Sistemi we adopted Computer Vision in various areas. Some examples are the integration of this technology with garment delivery systems developed by the Automation Division and the development of an automated logistics system for large warehouses. The latter, made up of completely autonomous baskets in movements within an environment monitored by cameras and without physical references, stems from the synergy between the know-how accumulated over the years with the Ambrogio robot project and the innovative concepts of Computer Vision.
Among the latest projects of Zucchetti Centro Sistemi SpA, currently under development, there is Agrirobot: the automatic robot for soil management and the treatment of the vineyard (conventional, integrated and organic management) thanks to the information from cameras with Stereo vision treatment is targeted at the plants that need it. Just entered into the vineyard, Agrirobot begins to work by recognizing the end of a spin, moving it to another. The D-GPS system, installed on board, allows Agrirobot to operate with a precision movement of 2/4 cm.
Watch the VIDEO about Agrirobot
In the future there will be many more opportunities to exploit the perceptual and computational capabilities offered by these algorithms. Personally, I am very confident in the potential that high-level software can have if coupled with a solid high-quality firmware / hardware base, such as that designed by ZCS. One way to go might be, for example, the use of low-power integrated video adapters for video recognition for obstacle detection and mapping of the surrounding environment, particularly applied on the ZCS Ambrogio robot's platform.
Sources:
I am a software developer, employed in Ideal Laboratory (Idealab), the research and development department of ZCS. I started my journey by taking up my vision computer robotics, especially in the field of industrial automation.
Altri articoli di Guido Bartoli