People and Object Tracking to increase efficiency in an airport

Area: Transports

Technology: Computer Vision / Tracking

Queues, checks, and waiting time: these are the passengers’ nightmares, in major airports. The difference between a positive experience and a ruined trip is often the hour spent waiting to be noticed at a counter, instead of shopping or dining. How can this problem be solved?

AISent has perfected AI algorithms that can estimate passengers’ numbers and study their behavior, predicting queue waiting times and suggesting corrective actions to airport operators. Cherry on top: the system works with the video stream of surveillance cameras already installed in the airport.

AISent’s solution improves quality, efficiency and safety with no extra Capex.

Self-regulation in a plant producing food products

Area: Food & Beverage

Technology: Machine Learning / Recommendation System

Water temperature, air humidity, viscosity, dough stability: when the variables are too many, it is difficult for a food industry operator to stabilize the process. Often, out of necessity, we proceed by trial and error, modifying one or more parameters according to individual instincts and experience.

AISent programs Machine Learning algorithms which allow the machines to self-adjust in relation to a certain set of parameters, through advanced data-learning processes. Where self-regulation is not possible, a Recommendation System suggests proper actions and adjustments to operators.

AISent’s solution improves production quality under different operating conditions, even in absence of skilled labor. Furthermore, staff training becomes faster, thanks to machine supervision and input.

Quality Inspection of plastic containers

Area: Packaging

Technology: Machine Learning / Quality Inspection

Packaging is the dress by which a product is judged. For a rigid plastic packaging manufacturer, black spots on the bottles could cause an entire batch to be rejected by the customer: considerable inconvenience, leading to waste and inefficiency.

AISent has developed a vision system that integrates cameras and Machine Learning algorithms. Once trained, the algorithms recognize which parts are defective, acting as expert operators.

The system is able to identify and reject defective parts on a fast line, regardless of the position of the black spots, the color of the bottle and external lighting conditions.

Real-time quality checks in a detergent production plant

Area: Consumer Goods

Technology: Quality Inspection / Anomaly Detection / Regression Analysis

Changing one ingredient or altering a random environmental factor is often enough to spoil the quality of liquid detergents. But how can you keep all components under control, when each phase is carried out by a different machine?

AISent has studied and trained a parameter regression algorithm based on the history of quality data: the mathematical model represents how the individual components, generated by different machines, contribute to the final quality of the product.

Thanks to algorithms and an intuitive graphic interface, the customer can see real-time production development, receive reports on anomalies and promptly intervene at any point in the process. The comparison of aggregated data on historical series, identifies the most critical phases of the line, facilitating plant management.

Predictive maintenance on marble cutting machines

Area: Machinery

Technology: Predictive Maintenance

How can you avoid breakdowns, malfunctions and production interruptions? By planning maintenance interventions as carefully as possible. Machines that cut marble slabs have to deal with the tendency of the diamond wire to break, after excessive elongation. How can the manufacturer prevent damage to the various components?

Temperature, pressure and process speed are just some of the conditions that cause damage to machines over time. AISent has studied a Predictive Maintenance system that analyzes environmental and production parameters, to monitor wear conditions and adjust accordingly.

The predictive maintenance system foresees when and on which components the technicians will have to carry out maintenance or replacement interventions. By reducing the risk of unforeseen events, AI preserves industrial assets and avoids sudden production stops.

Computer Vision for the automatic sorting of urban waste

Area: Machinery

Technology: Machine Learning / Object Detection

Sorting waste by material type is not very accurate, if done live; the human eye is often unreliable: there may be gross errors, preventing proper recycling.

AISent has developed algorithms based on multilayer neural networks which identify different objects just as well as expert operators.

Urban waste recycling becomes faster and more accurate, thanks to Artificial Intelligence and Computer Vision. In addition, correct waste sorting improves the circular economy.

AI for Customer Experience

Area: GDO & Retail

Technology: Computer Vision/Data Analysis per il Retail

Clothing stores must attract customers. Maximizing customer experience means not only improving product quality but also understanding people’s desires: we all want to feel understood and pampered.

AISent has implemented the PeopleEngine platform to observe people flow and behavior, analyzing their preferences. PeopleEngine collects customer needs and develops personalized communication strategies.

Vendors organize their spaces and choose which products to display, according to the system's recommendations. They acquire new customers and increase the value of existing clients, thanks to statistical data.

Smart diagnosis for precision medicine

Area: Pharma & Medicine

Technology: Real Time Data Processing

Artificial Intelligence can make a difference even in life sciences. For example, a biomedical device manufacturer would benefit from a system that can estimate blood oxygenation (pO2) for applications in extracorporeal circulation.

AISent has designed an artificial intelligence algorithm that estimates blood pO2 level, at current temperature and at 37 ° C, by observing liquid fluorescence in response to a light pulse.

The system can estimate pO2 level with an error margin comprised within the limits imposed by biomedical devices, regardless of blood temperature, probe and light spot used.