Artificial Intelligence Video
The image processing capabilities constitute the intelligent heart of Colibrì Security Platform and Colibrì Customer Video Insight, solutions based on the continuous development of the most advanced artificial intelligence technologies applied to the fields of security and behavioural analysis.

Artificial intelligence (AI) is a branch of computer science that deals with the intelligent behaviour of machines. It deals with the capacity of a machine to simulate, through the elaboration of particular algorithms and in an automated way, the human behaviour and its reaction models to stimuli coming from external information.
Artificial intelligence is now pervasively present in our daily lives: from weather forecasts to voice recognition, from search engine suggestions to the most advanced humanoid robots. In order to simulate the human experience, one of the most used sources of information is the image, either static or in the form of a video stream. Advanced AI algorithms applied to the large amount of data hidden in a video allow the deduction of multiple pieces of information, some of which are difficult to perceive by the human eye. Computer Vision, Machine Learning and Deep Learning can be considered branches of Artificial Intelligence.

Computer vision is the set of processes that aim to create an approximate model of the real world (3D) from two-dimensional (2D) images. The main purpose of computer vision is to reproduce human vision.

A machine vision system consists of the integration of optical, electronic, and mechanical components that allow for the acquisition, recording, and processing of images both within and beyond the visible light spectrum.

Machine Learning technology aims to simulate on a machine the ability to learn in an automatic way, i.e. the possibility to constantly improve the performance in solving a problem using positive reinforcements (successes) and negative ones (errors). The machine performs a task on the basis of a generalized model, acquiring "experience" thanks to the feedback that the programmer or the operator provides during operation. This allows the machine to deal with tasks that are too general to have an unambiguous solution but require a best effort logic, i.e. the best possible result.


Deep Learning is a further technique of automatic learning that makes extensive use of neural networks, a system of computation that replicates the functioning of the human brain. This type of structure simulates learning processes similar to those of the human mind. In a first phase, called of training, a high number of positive samples is elaborated by a network opportunely dimensioned; the aim is to extract mathematically the general characteristics of this set of information and to assign, to each characteristic, a weight of importance on the basis of how much this recurs in the set, a bit like it is done with an experiential learning. Once this phase is completed, the network is ready to elaborate new information, classifying case histories that it did not receive in the training phase.

The active video surveillance software platform that integrates cutting-edge technologies and COTS security products.

Alerts the operator when a person or vehicle enters a monitored area. You can increase the effectiveness of the feature by setting a number of filters including background noise compensation, object size and object permanence in the scene.

Allows visual control of access and identification for access to a gate, integrating video surveillance and control by physical barriers. It alerts the operator and records alarms in the history due to denied or forced access. It is also able to detect the temperature of people at the entrance.

It allows to identify the identity of a person from a video of him, comparing in real time the images captured with the faces recorded in a database and generating notifications and alerts on the matches detected. 

It is able to identify and extrapolate the license plate number of vehicles in transit, using advanced OCR (Optical Character Recognition) technologies, even in critical situations such as vision limited by darkness or atmospheric agents, dirty or damaged license plates.

The platform that meets the need of businesses open to the public to know the type and behavior of their customers.

Allows you to know demographic information through facial recognition: age range, gender and ethnicity of visitors.

It allows to estimate the occupation of specific areas of the shop and to identify the most frequented routes, through colorimetric graphs. It thus makes it possible to evaluate the positioning of the stands and the effectiveness of the exhibition route.

Allows you to know how many customers enter and leave an access: mall visitors vs shop visitors, visitors vs shoppers, instant/hourly/daily crowding and attendance.

It allows the monitoring and analysis of queues, providing real-time data, historical data sets, useful tools to identify and predict bottlenecks, improve service, sales and resource planning.

Colibrì is an active video surveillance software platform that integrates cutting-edge technologies and COTS (Commercial Off The Shelf) products, creating unique, complete and customizable physical security and customer analytics solutions, adaptable to different markets.