The Institute for Information Industry's Software Technology Institute (STI), under the subsidy of the technology project supported by the Department of Industrial Technology, Ministry of Economic Affairs, has developed the world's largest thermal imaging deep learning dataset, Formosa Dataset, with over 5 million vehicle and pedestrian objects. Additionally, the project staff have developed AI thermal imaging object recognition technology based on the dataset. The developed "AI Thermal Imaging Smart Driving Warning System" outperforms traditional image warning technology in poor low light driving conditions such as on rainy or foggy days. The system can achieve a collision-warning rate of over 95% for forward collision and blind spot detection, ensuring all-weather driving safety.
The technology, AI Thermal Imaging AR-HUD Smart Driving Warning System, which is developed in collaboration with the domestic company First International Computer (FIC), made its debut at the "2024 Taipei International Automobile & Motorcycle Parts & Accessories Show (2024 Taipei AMPA). The Institute for Information Industry (III) combined locally developed automotive AR HUD products with thermal imaging cameras, pioneering the "AI Thermal Imaging AR-HUD Smart Driving Warning System" that integrates AI recognition technology, thermal imaging cameras, and automotive AR HUDs. The system offers the advantages of all-weather adaptability, high precision recognition, and intuitive display for intelligent driving safety warnings.
The Institute for Information Industry (III) has long been committed to the development of intelligent vehicle technologies. To address the high accident rates caused by poor night vision, III conducted relevant research and found that many image-based Advanced Driver Assistance Systems (ADAS) on the current market predominantly use visible light cameras as sensing components. The operating principle of this technology is similar to that of the human eye, making it susceptible to changes of environmental lighting conditions, which leads to limitations in visibility. These systems are especially problematic at night, when urban lighting can interfere and reduce the driver's visual range. Additionally, adverse weather conditions such as heavy rain or fog during the day also restrict the driver's field of vision and limit the recognition capabilities of ADAS.
For the reason, the Software Technology Institute (STI) of the Institute for Information Industry (III) and First International Computer (FIC), have jointly developed the "AI Thermal Imaging AR-HUD Smart Driving Warning System." The AR HUD, an augmented reality head-up display, utilizes optical technology to project ADAS information, navigation details, vehicle speed, speed limit indicators, road warnings, and other data into the driver’s field of vision. Therefore, drivers' focus can be shifted from the instrument panel and navigation screen to the view ahead. By integrating AI thermal imaging recognition technology, III enhances the system’s capabilities to detect surface temperatures of pedestrians, vehicles, and various objects under different temperature conditions. The system is now unaffected by environmental conditions such as day or night, light or shadow, rain or fog, etc., and can project the location of objects that drivers should to be warned about directly onto the AR HUD. The adjustment of the system enables drivers to concentrate more on the road ahead, achieving an effect that guarantees intelligent all-weather safety warnings for driving.
Henry, I-Heng Meng, Acting Director General of the Software Technology Institute of III, stated that after integrating the research and development results of the department with AR-HUD modules developed by First International Computer (FIC), the development team is able to directly project warning information onto the windshield in the manner of augmented reality. The improvement allows drivers to effectively focus on the projected warnings ahead in various challenging visual environments, enabling them to promptly identify and pay attention to the positions of pedestrians and vehicles and slow down accordingly, significantly reducing the likelihood of traffic accidents. Furthermore, the intelligent cockpit systems are also expected to be implemented in foreign passenger buses to provide professional services by the end of 2024, with an estimated economic impact of around 1.2 billion.
It is worth mentioning that in the past, MIT (Made in Taiwan) thermal imaging products were lacking, and automotive applications were only seen in imported luxury cars. The market situation was mainly due to cost considerations, making the technology difficult to enter the affordable market. However, the results showcased this time are entirely MIT, both software and hardware. Thanks to the participation and investment of Taiwanese manufacturers, safety features previously limited to luxury cars are now being implemented in public transportation vehicles such as passenger buses. In the near future, it is believed that this safety service will become widespread in various vehicles in Taiwan.