Videointellect, a Skolkovo company that provides AI, data analysis and machine vision systems for the transport, sports, industrial and financial sectors, reached the semifinal of the prestigious “Straight to Wuzhen” Global Internet Competition 2020 in China.
The Straight to Wuzhen competition is an important part of the year-long World Internet Conference, attracting over 1,000 companies and 1,000 high-quality digital projects, the ultimate goal being to advance the digital economy. A total of 27 foreign projects out of more than 800 were selected to go through to the competition semifinal and Videointellect’s AI, data analysis and machine vision system was among these. While the company did not make it into the final, the achievement is nonetheless impressive and represents a great opportunity for it to find partners in the Chinese market.
Videointellect’s business development representative Anton Chertkov said in a statement to Sk.ru: “We are grateful that we are part of that prestigious event. The competition is quite tough as we competed with the best startups from all over the world, and we demonstrated to the judges the distinctive features of our technology. It’s quite unfortunate that we can’t visit Wuzhen during this competition, but the event was highly organized even online.”
What does Videointellect do?
Closed-circuit television (CCTV) has been around for decades, but with the advent of artificial intelligence, predictive analytics, data analysis and machine vision it has been brought to new levels of advancement. Videointellect’s unique selling point is to provide CCTV systems with all of these features built in. The purpose of predictive analytics is to detect and predict events that require a fast and rapid response from the operator.
Today’s society faces new security challenges, especially with regards to terrorism or enforcing social distancing in the context of the coronavirus pandemic. Videointellect’s system focuses on four key areas: transportation, industrial sector, safe & smart city and the financial sector. Videointellect’s video analytical systems (VIDEOINTELLECT software) has been designed for efficient work both on premises as well as in places of mass gatherings: in transport infrastructure (subways, railway stations, airports, bus stations), as well as in parks, gardens, embankments, water protection zones, tourist areas, educational institutions, sports facilities and shopping centers, etc. The system analyzes video stream in real-time and reports when an atypical action takes place (vandalism, fights, crime, etc.), identifies unattended objects in public spaces and can monitor factory floors for PPE use and theft prevention.
According to Mr. Chertkov, the idea came into being after the 2010 Moscow metro terrorist attack, which killed forty people and injured dozens more.
“Our team decided to enhance the security standards of underground stations to prevent possible future attacks and ensure people's safety,” he said. “Our team put their knowledge into practice and began developing video analytics modules that would improve security systems.”
The Videointellect team, with its extensive experience working with artificial intelligence and video analytics systems, saw an opportunity in the Russian market and countries elsewhere. The team won a grant for the new technology and installed their first pilot project in 2015 before becoming a part of the Skolkovo ecosystem.
The global market for video analytics is valued at approximately $3 billion according to Fortune Business Analytics and is projected to grow to nearly $12 billion by 2026. The Russian market alone is valued at $300 million in 2019. According to Mr. Chertkov, Videointellect is the only company on the Russian market to have all four modules certified – abandoned items, movement in restricted areas, movement in restricted direction, and atypical changes in behavior.
Criminal Pattern Recognition
While the company operates in multiple sectors, the core of the business is in transport security and its first milestone project was in the St. Petersburg underground where the project showed high-accuracy results for detecting abandoned items.
“Such places are very difficult to analyze, because there are lots of people and items, and the camera angles can be tricky to get right. The camera should always spot the smallest objects and we are currently updating our system so that it can understand more patterns, identify more items and so on,” said Mr. Cherkov. “The chances of the system identifying a bad situation are increasing daily, making it one of the best on the market, if not the best.”
The key element of the system is its self-learning mechanism, which is particularly useful for crowded places, and underground railway platforms serve as good training for the AI. The more the camera observes a busy metro station, the more the system updates and improves itself on human behavior and the nature of abandoned items. While the system is already trained to identify many different items, it can also be trained to notice particular items.
“The system we have today is a whole different level of identification when compared to the pilot project we did in 2015,” said Mr. Chertkov, “and it has been installed into more than 80% of metro stations in St. Petersburg. The aim is to expand this even more.”
The system utilizes a complex process of pattern recognition to identify preparation for an illegal act, autonomously updating itself while also being updated by the Videointellect team. According to Mr. Chertkov, all human behavior follows a pattern, whether it involves getting ready for work in the morning or preparing to commit a crime.
“When it comes to illegal behavior, the patterns are still there but they’re just different to everyday acts. In talking about AI and video analytics, human activity is characterized by two components of behavior. The first is the sequence of human action to achieve a result. The second is the behavioral characteristics of a person,” he said.
Every person has individual motor skills that affect and portray the emotional state. In a shop scenario, the Videointellect system can identify through a person’s body language what the intent is, whether it is to purchase an item, to steal it, or simply to browse. In doing so, it reduces the human error factor, because a security officer observing multiple screens at once might not spot odd behavior. A person intending to steal in a shop setting would deviate from the expected patterns that would show intent to purchase an item or to browse. When the system identifies behavior associated with intent to shoplift, it notifies the operator and displays a percentage chance of theft. In this way it acts as an assistant to human operators rather than a replacement.
Applying Self-learning CCTV
One of the main advantages of the Videointellect system is its adaptability across multiple industries. For instance, while it has been applied to the St. Petersburg underground system, it can also be used to monitor goods in a factory in real-time.
In a factory setting, it can act as a quality control and a potential Asian client hopes to use the system to monitor the quality of paper and plastic cutlery products during production. Right now, the company in question uses factory workers for quality control, but this adds human error to the equation. When an individual is looking at thirty thousand items a day, the chances of error invariably increase. An automated system removes the human error factor.
The term “preparation for an illegal act” is quite broad, but in the context of a public setting it encompasses a number of crimes such as pickpocketing, fighting, arson, and terrorism. An example of the system in operation in a sports setting is at a soccer match, where violence between football fans is commonplace. Even amidst the crowded setting of a soccer stadium, Videointellect’s video analytics system can identify the first signs of an altercation before it has even begun and can spot fan banners that might incite violence.
“We began with a focus on transport security and we saw that there is great demand for our types of products. We also saw opportunities in the retail, industrial and banking sectors and we have had a lot of communication and pilots with companies from those sectors,” said Mr. Chertkov.
In 2018, all of the company’s contracts were with the transport sector. As of 2020, they had already diversified into the industrial sector and are focusing on getting contracts in safe and smart cities by the end of the year. Videointellect has closed several industrial contracts with domestic companies, one of which works in the petrochemical sector and another that works in the metals and mining sector.
“Over the last few months we’ve pitched in ten different markets, including the US, Indonesia, France, Singapore, South Korea, Japan, and so on. There has been a steady stream of demand from the industrial sector and the problems facing those companies are the same as those facing domestic companies. One of these is controlling the use and availability of PPE – vests, gloves, helmets. The second challenge relates to sabotage, fire safety, and theft. Theft on the factory floor is a major issue for many companies, compelling them to spend a lot of money to combat it. You can hire a security person to prevent theft, but what if he himself gets involved? You have to hire a second security person to monitor the first. Smart CCTV solves that.”
At Home and Abroad
While building the company name in the Russian and CIS markets is currently the main goal, the next stage in its business development is to enter other markets, particularly in Asia.
“We see this kind of demand across the world and we are now thinking of opening a branch in a country of our interest. We are currently seeking customers that would be interested in this kind of solution. Of course, it’s very challenging right now, because our technicians need to check the facility where the system will be installed in order to understand the requirements - the number of cameras, the area, what technologies are already installed, what resolution is needed, understanding the production line, where the cameras can be placed and where they can’t. Because the borders are generally closed, it’s not as simple as sending a team out to do an inspection.”
By the end of the year, Videointellect aims to have a bigger picture of what countries are the most realistic candidates for market entry. The company has attracted interest from an Indonesian business as well as the French corporation Total. Another European firm has also approached Videointellect with regards to using its system for airport security.
“We want to find new pilot projects with different challenges to solve, tailoring our system to suit each client’s demands. Give us a problem and we will solve it.”