to strengthen the foundation of digital intelligence, aokan technology and huawei jointly released a large-scale model solution for transportation and logistics
2024-09-25
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
during huawei connect 2024, huawei's "win-win intelligent transportation summit" was successfully held. aokan technology was invited to attend, and together with huawei vice president and ceo of intelligent transportation corps ma yue and global industry partners, they shared the vision, trends and latest practices of intelligent transformation in the field of large transportation, and officially released the transportation and logistics big model solution to promote the high-quality development of intelligent transportation.
3rd from right, he xin, cto of aokan technology
as a leader in the new quality productivity of visual big data, aokan technology has been complementing huawei's advantages in products, business and market for many years. today, the rapid rise of the big model industry has brought about a paradigm change in ai application and development. by building a general big model + scene fine-tuning, it can achieve better ubiquitous discovery capabilities, higher event recognition accuracy, shorter algorithm development and launch time and event discovery cycle, and efficiently assist in the reform of traffic management methods.
this year, aokan technology focused on traffic accident scenarios, combining view big data and self-developed big model capabilities to buildtraffic video labeling assets, traffic accident detection models, high-speed spilled object detection, traffic accident level determination and descriptionand other business functions to form the aokan traffic safety big model solution, reconstruct the traffic safety management scenario, and help the city move towards refined "smart management".
traffic video asset tags
for the processing and application of massive front-end video resources, the conventional manual sorting method has a long cycle and is not standardized, with low retrieval efficiency, and cannot be used to review videos based on scene changes and business needs. the aokan traffic safety big model can establish a standardized intelligent labeling system for urban traffic video resources, realize ai intelligent labeling, directly match label data with business, and retrieve video resources in seconds.
as a result, video retrieval time has been reduced from minutes to seconds, video review has evolved from manual operation to voice control, and the video tag refresh cycle has been reduced from one month to one hour. activating traffic video resources has become an effective data asset, allowing cameras to be used flexibly for business.
traffic accident detection model
it is understood that the traffic police command center usually has about 5 police officers on duty at the same time, and can handle up to 200 police incidents every day. however, accidents often occur randomly and anywhere, and the lack of patrol police force leads to the reliance on citizens to report incidents, making it difficult to proactively discover and track police incidents in a timely manner.
the aokan traffic safety model has a built-in traffic accident detection model to achieve real-time monitoring and automatic warning of roads. when the system detects abnormal situations, such as vehicle collisions, congestion, etc., it can immediately notify nearby police forces to deal with them, greatly shortening the response time, optimizing the allocation of police resources, and improving the work efficiency and emergency response capabilities of the traffic police command center, thus safeguarding urban traffic safety.
high-speed spill detection
for complex behaviors such as high-speed spillage, special scenarios are difficult to detect with small model algorithms. the aokan traffic safety large model has a built-in training and pushing machine, and highway managers can generate new algorithms at zero cost in a very simple way, truly realizing "ideas are algorithms".
warning information can include detailed information such as the location, size, and type of the spilled objects, so that timely treatment measures can be taken. at the same time, the traffic safety model continuously collects new high-speed spilled object detection data to optimize and update the cv model. by increasing the diversity and quantity of training data, the generalization ability and accuracy of the model are improved to improve the level of highway safety management.
traffic accident grade determination and warning description
the development of large-scale visual models has brought profound changes to the understanding and handling of traffic accidents. traditional judgment methods often rely on manual on-site inspections and relevant witness statements, which is inefficient and easily affected by subjective factors, resulting in inaccurate judgment results.
the aokan traffic safety big model is equipped with cv big model capabilities. through visual learning of a large number of traffic accidents, it can automatically identify and analyze key elements in the accident. for example, the shape of the injured in the accident area, the vehicle overturned and the damaged debris, etc. using advanced image processing technology and deep learning algorithms, the solution can quickly and accurately extract information and convert it into quantifiable data. at the same time, using aokan's self-developed image-to-text big model capabilities, it can automatically generate detailed traffic accident warning descriptions, providing an important basis for subsequent accident handling and responsibility determination.
industry intelligence is a complex system engineering, and only openness and cooperation can lead to win-win results. in the future, aokan technology will work with huawei and global partners to build a stable and reliable digital foundation through technological innovation, thereby enabling intelligent upgrades in transportation business scenarios and helping to realize the vision of "enjoying people's travel, optimizing the flow of goods, and winning with digital intelligence."
(dazhong.com)
