Using AI and Machine Learning for Vehicle Inspection
UVeye adds General Motors to its growing list of OEM partners looking to transform the vehicle-inspection process.
At its inception in 2016, UVeye’s deep-learning technology was developed for the security industry to detect weapons and contraband. Applications in the mobility industry followed, with UVeye’s 70-person R&D staff developing high-speed automated vehicle-inspection systems. The Israel-based company has established partnerships with Daimler, Honda, Skoda, Toyota and Volvo, and recently inked a commercial agreement with GM to expand deployment of UVeye’s vehicle inspection systems to GM dealerships, to help identify issues for repair.
UVeye CEO and co-founder Amir Hever recently spoke with SAE Media’s Kami Buchholz about what’s on the horizon for the firm. The interview’s highlights:
What is the technology behind UVeye’s vehicle-inspection systems?
Our systems use a combination of proprietary algorithms, cloud architecture, artificial intelligence, machine learning, high-definition cameras and hardware to provide a high-speed automated visual scan in seconds. We’ve scanned more than a million vehicles via quality-assurance inspections at assembly plants. Our Artemis system checks tires for tread depth, air pressure, sidewall damage and more. Our Helios system is an undercarriage scanner for capturing things like missing parts and fluid leaks. Our 360-degree Atlas system detects exterior defects and damage, even scratches and dents that are as small as 0.5 mm (0.01 inches). All three systems are drive-through.
What’s unique about the scanning systems?
We have 11 patents that address many aspects, including how we acquire the data. Synchronization is very important. That’s why we synchronize the LED illumination with the scanning cameras because we need the camera lens to get the right amount of light projected onto the vehicle. That enables the system to find the right data needed to feed our artificial intelligence. The existing patents cover our core technology, but we’re going to file patents in the very near future that will be more focused on EVs and autonomous vehicles.
What are UVeye’s EV plans?
EVs bring a lot more challenges. From a mechanical perspective, EVs are a bit less complex than ICE-powered vehicles. However, EVs have many sensing technologies, so manual inspection simply isn’t enough anymore. Manufacturers will need automated inspection systems. For one thing, you need to be sure that the sensors and cameras are pointed at the correct angles. With many of today’s EVs having the battery pack residing in the undercarriage, it’s very important to know if there’s underbody damage, because leaks and other issues impact the battery. You definitely want to understand the battery’s health.
What’s the plan for autonomous vehicles?
Autonomous vehicles are going to be connected [vehicles]. We want to communicate with a vehicle to ensure that what we’re seeing with our scanning inspection matches what’s indicated on the onboard vehicle computer. The ability to compare both will give us an ability to understand in-depth what’s going on with the vehicle. Is it performing correctly? What needs to be fixed? We want to provide a more complete image of the vehicle’s overall state. For fleet operators, especially those with autonomous vehicles, it’s important to know if there are any safety or mechanical issues. Automated vehicle inspections can provide a framework for preventive and predictive maintenance, which will help increase a vehicle’s up-time.
Going forward, how do you see UVeye’s future?
We’d like to become the standard for how to inspect a vehicle during its lifecycle. We’d also like to be a provider of data for business insights. For instance, what’s the wear-and-tear on vehicles in cold countries and in hot countries? How does road salt affect the wear-and-tear of EVs with battery packs in the undercarriage? If we’re able to collect enough information, we can start showing insights to help manufacturers take preventative steps and possibly reduce the number of recalls.