Solving pain points and bringing in revenue — OmniEyes uses AI image recognition to create an all-round traffic management system

3 Dec

Fleet management today faces many challenges. There are hundreds of thousands of vehicles on the street at the same time, covering hundreds of kilometers. Management centers must monitor driving information in real-time and take appropriate action for operations management to be truly carried out. The startup founded by professors from NTU’s College of Electrical Engineering and Computer Science has developed “OmniEyes – Next Generation Mobile Video Platform” that can fully collect and analyze live street views. It can increase fleet management efficiency and work with third parties through innovative application services, establishing a three-way win-win model.

As society and mobile technology progresses, fleet arrangement will also diversify, making fleet management improvements vital. Zhou Jun-ting, the CEO of OmniEyes and a professor at NTU’s Department of Electrical Engineering, has studied several fleets since the team was founded a year ago. He observed three pain points in fleet management: “The first is fuel consumption, including idle management and path planning, since many taxi drivers do not shut off the engine. The second is dangerous driving, which causes vehicle damage and personal injury. The third is fines for traffic violations.” These three carry a large expense ratio in fleet management. Therefore, when profit growth is limited, many fleets hope to reduce operating expenses by using traffic management to solve the pain points.

At present, traffic management has become digitized and mostly relies on GPS and OBD2 (On-Board Diagnostics). GPS can record the route and OBD2 is similar to a traffic computer that records driving information, like how hard the driver steps on the gas pedal or brakes, and sends it to the traffic control center. However, Zhou Jun-ting points out that these two tools do not allow the fleet to carry out active and verifiable management and auditing. He said, “Heavy braking may be due to the driver’s poor driving habit, but it may also be problems on the road or a motorcycle dashing out all of a sudden.”

OmniCam makes fleet management easier

How can we help fleets become more active? Zhou Jun-ting mentions that image collection and automatic identification technology must be introduced into the current traffic management system. OmniCam, developed by OmniEyes, is an AI image-based traffic management system that combines technologies such as edge computing, AI image recognition and cloud computing. It captures street views in real-time and tracks useful images, converts them into information and sends them to the cloud platform.

OmniCam is not a physical camera. It is software introduced into a fleet-compatible OBU or traffic management system. Through local OBU identification results or images sent to the cloud for real-time analysis, the fleet can determine whether the driver has committed any of the dozen violations such as reverse parking, parking on a crosswalk, illegal left turn, and not yielding at an intersection. “For example, we have recently detected drivers who constantly go in the opposite direction and run red lights. When the traffic control center receives a warning and pulls up the footage, the driver cannot make any excuses.” OmniCam can also detect the lane position (e.g. the second inner lane) and lane changing during driving and check for other problems through the images.

As for the technical advantages of OmniEyes, Zhou Jun-ting points out that current competitors cannot provide real-time recognition on the front-end and handle various real-time recognition requirements. OmniEyes can adjust its computational model to fit different scenarios and can be used in dynamic and complex environments. It is a technological barrier established by the team after years of research and development.

Zhou Jun-ting has also shared the case studies of the technical advantages applied in the public sector. Since July of this year, the Taipei City Government has stipulated that passengers swipe their cards both when getting on and off the bus. However, the concern is that this will increase the time it takes to get on and off. So, the government requested OmniEyes to design a system that can calculate the time.

“It took us just two weeks to complete the design, construction and analysis. We calculated not just the total time but also the time at every stop,” Zhou Jun-ting said proudly of this achievement. The solution proposed was very intuitive. The camera was pointed at the door and the software was trained to detect the door opening and closing. The camera’s built-in timer started when the door opened and stopped when the door closed. The conclusion was that the total dwell time of the entire trip was increased by two minutes.

OmniLabel increases overall fleet revenue

In addition to fleet management, OmniEyes has developed another B2B product called OmniLabel. It can add markers to the information collected by the fleet and share them with third parties, thereby increasing overall fleet revenue.

“Commercial fleets transport items from point A to B. The more items transported, the higher the revenue. To increase revenue, the fleet must be expanded. If the information collected by the fleet is shared with third parties and a data usage fee is charged, the fleet will experience organic growth,” Zhou Jun-ting explained.

OmniEyes has also reached out to these third parties and discovered instances where consultants were requested by customers to find a location with many pet shops, so the customer can open a pet shop nearby. They were also requested to open a dental clinic at a location without dentists within 500 meters. In addition, taxation departments were looking for vehicles that evaded taxes, police departments looking to seize stolen cars, vendors wanting to know the visitor traffic at intersections, passersby looking for parking information or the queue status at popular stores.

“In the past, yellow pages were used to look up information or a part-time employee was assigned to stand on the street with a stopwatch. But now, OmniLabel can identify the information on signs, the type of business, the number of stores, etc. It can manage all of these in real-time.

Establishing an all-round ecosystem

At present, there are about 200 vehicles with OmniCam on the streets of Taipei, including 100 buses on the 6 lines of the Metropolitan Transport Corporation and 40 cabs from Taiwan Taxi. Zhou Jun-ting revealed that OmniEyes would be able to work with a medium-sized fleet at the end of the year. By then, there will be 1000 more vehicles equipped with OmniEyes.

“Omni-” is the root of the word omnipresent. As its name suggests, OmniEyes means omnipresent eyes and is also Zhou Jun-ting’s vision for the entire industry. From 200 to 1000 vehicles at the end of the year and the 400 000 dash cam demands of the country, Zhou Jun-ting has gradually built an all-round ecosystem from which fleets and third parties can benefit as more application services are created, furthering the realization of a smart city.

Facebook Comments