One of the basic “natural” human functions that humans are born with is pattern recognition and identification, which is very easy for us to do. We use it routinely.
There are a huge number of tasks in production, which are associated with the identification and classification of various objects, which the person performs.
Human performance is severely limited in this type of tasks, in addition, routine and monotonous actions can lead to errors. Vision systems allow to solve many tasks of this type in automatic mode.
This will provide a significant increase in productivity and virtually no errors. The cost-effectiveness of modern solutions is very high. In addition, they are virtually maintenance-free.
Different types of sensors can be used:
- based on video analytics,
- based on neural networks,
- based on a laser probe,
- based on 3D-scanner.
Pick&Place – detection of objects, their orientation and displacement for positioning various manipulators and industrial robots for further operations.
Sorting – classifying objects into groups by size, shape, color and many other features.
Object counting – determining the number of randomly distributed items on the conveyor; control of the packing of items.
Detection – determining the presence or absence of objects in the production area, determining the correct use of the object, weld.
Sorting
The sorting process is mandatory for a number of productions in which primary processing of objects is performed, for example, fruit and vegetable crops, meat, leather and fur, wood, valuable minerals. Classification of objects into groups can be carried out by size, shape, color and many other features. Automatic sorting can significantly increase labor productivity. Neural networks with deep learning are used for complicated objects, where expert estimation of goods on the basis of a set of parameters is required. This makes it possible to solve the most complex tasks, which previously could be performed only by a highly qualified specialist.
Counting objects
The problem of determining the number of objects that have passed on the conveyor belt is usually solved by fairly cheap photoelectric sensors. But this solution works only in the case of a single passage of an object per unit time. In case of undirected and random arrangement of objects (for example, in bulk) more complex systems based on video analytics, 3D-sensors or, in complicated cases, neural networks are required. As a rule, the task is solved on the basis of algorithms of matching analytics with the benchmark, which allows to determine and count the necessary objects and their parts with high accuracy.
Detecting
Detection of objects and production processes is one of the most important tasks in modern production. Monitoring compliance with the rules and regulations of the production process and safety rules in the enterprise is a very costly and complex process, which is often impossible to trace. For example, an operator has loaded a mixer with the wrong composition of mixtures, and because of this the production has suffered a defect. Vision systems are able to detect and suppress wrong scenarios in a timely manner, thereby ensuring 100% correctness of the technology performed by the workers.
