From July 4th to 7th, the 2024 World Artificial Intelligence Conference was held at the Shanghai World Expo Exhibition Center. With the theme of "Promoting Shared Governance through Consultation and Promoting Wise Intelligence through Good Governance," the conference was jointly organized by the Ministry of Foreign Affairs, the National Development and Reform Commission, the Ministry of Education, the Ministry of Science and Technology, the Ministry of Industry and Information Technology, the Chinese Academy of Sciences, the China Association for Science and Technology, and the Shanghai Municipal Government.

Digital technology has had a profound and widespread impact on the transportation industry. As one of the earlier domestic enterprises to explore the implementation of AI and big model industries, Hisense Network Technology has always been concerned about and committed to tightly integrating innovative technologies such as video AI and big models with business applications. At the China AI Industry Innovation Achievement Exhibition Zone in Hall H2-A101 of the Shanghai World Expo Exhibition Center, Hisense showcased the practice of smart transportation construction empowered by AI and big models, discussing with exhibitors the future vision of high-quality development for smart transportation.

Enhancing perception, AI accelerates the construction of smart transportation

Perception is the foundation of smart transportation.

Whether it is alleviating congestion and ensuring smooth traffic flow or safety prevention and control, it cannot be separated from strong perception capabilities. Hisense deeply integrates industry experience, collects massive training materials, and has accumulated more than 200 visual analysis capabilities for four major categories of scenarios, achieving minute-level detection of emergencies with an accuracy rate of over 90%. This has addressed pain points such as poor adaptability of video AI scenes, low implementation indicators, and missed or false alarms. Currently, it has been implemented in scenarios such as congestion identification, accident detection, equipment failure detection, and vehicle-end applications, achieving significant results.

The accuracy rate of accident identification is over 90%. In accident detection scenarios, Hisense has conducted in-depth research on segmented traffic accident scenarios such as rear-end collisions and general collisions, as well as typical accident characteristics such as vehicle stationary state and driver wandering behavior. By accurately identifying eight key elements, events can be quickly determined, with an accuracy rate far higher than the industry.

The accuracy rate of congestion identification is 98%. In congestion identification scenarios, Hisense has made the industry's first breakthrough in congestion detection technology based on high-point video. By linking high and low points and comprehensively perceiving large-scale congestion situations, it provides strong support for congestion mitigation operations. High-point video provides a more comprehensive view of congestion situations, more accurate identification of congestion nodes, and clearer observation of congestion process changes, making it highly favored by frontline commanders. In the National Challenge of Traffic Incident Video Recognition hosted by the Ministry of Public Security's Road Research Center, Hisense won the first prize in the urban traffic scenario with this technology.

Promising Prospects, Big Models Aid in the High-Quality Development of Smart Transportation

From "weak AI" to "strong AI," big models are gradually demonstrating unprecedented capabilities and wisdom, bringing more possibilities to the construction of smart transportation.

What problems can big models solve? Hisense believes that big models play a crucial role in balancing the relationship between people and vehicles, vehicles and roads, and even promoting optimization throughout the entire transportation process. They can bring about four significant values: enhancing proactive traffic cognition, leveraging decision-making, reasoning, and prediction capabilities, improving human-computer interaction, and promoting data governance and efficient application. These values greatly contribute to the construction of smart transportation.

How can the vitality of big models be unleashed? Hisense insists on combining big models with business applications to create scenarios with excellent implementation effects and high application value. As Chen Xiaoming, the chief engineer of Hisense's Smart Transportation Business Unit, previously proposed in a speech: "Big models are not just a show-off of technology. Bringing value increment through technological increment is the future direction, and the wide range of application scenarios in smart transportation is the natural soil for the implementation of big model applications."

Big models make smart transportation construction more efficient. In safety hazard investigations, language big models can extract key knowledge based on brief case descriptions, survey records, and other textual materials, quickly complete and calibrate accident information, autonomously identify safety hazards, recommend hazard management strategies, and provide expert-level knowledge question-and-answer support for management personnel, enabling traffic accidents to be prevented before they occur.

Big models optimize the effectiveness of smart transportation construction. In intelligent optimization scenarios, based on reasoning, learning, and prediction of traffic flow patterns, big models can better identify traffic operation characteristics, assist in diagnosing traffic problems, automatically generate optimization strategies, and automatically evaluate optimization effects, achieving a closed-loop signal optimization system.

Big models reduce the cost of smart transportation construction. In the public transportation sector, the introduction of big models allows for a comprehensive assessment of passenger flow distribution, flow direction, line network optimization, and transportation capacity, ensuring optimal matching of passenger flow and transportation capacity while maintaining service quality. This reduces inefficient mileage and achieves operational cost reduction and efficiency enhancement.

Big models enhance the experience of smart transportation construction. In vehicle-road collaboration scenarios, big models can accurately identify and understand the status and trends of various transportation participants and environments in complex environments, making accurate driving responses to avoid driving risks caused by sudden situations such as blind spots and collisions. This assists in driving with "sharp eyes and quick hands." At the same time, through information sharing and interconnection, it helps to better plan routes and improve travel experience.

As the construction of smart transportation continues to advance, how to combine industry practices with emerging technologies to achieve high-quality development has become an urgent issue to be addressed. As a leader in smart transportation, Hisense will continue to adhere to technological innovation and application implementation, helping transportation move towards a new era of high-quality development, making traffic management smarter and urban operations smoother.