Core Insights - The project led by Pingao Co., Ltd. won the national first prize in the "First Comprehensive Transportation Big Model Intelligent Innovation Application Competition" with its video and image intelligent recognition system for smart transportation [1] - The project features an innovative "edge-center-human" three-tier collaborative workflow architecture, significantly enhancing security inspection efficiency while maintaining a high accuracy rate (detection rate ≥ 95%) [2] Technical Highlights - The project utilizes domestically produced high-density computing integrated machines, known as Pingyuan AI integrated machines, which are equipped with domestically designed Jiangyuan D10 acceleration cards, ensuring low power consumption and high computing power [3] - The software architecture is centered around the Pingao Intelligent Computing Cloud Base, enabling collaborative scheduling and resource pooling across edge and center computing platforms [4] - The system incorporates lightweight detection models optimized for edge computing environments, addressing the challenge of limited resources while allowing for parallel processing of multiple AI tasks [4] Model Evolution - The project supports incremental training based on locally added samples, allowing the system to quickly adapt to new threats and complex scenarios without requiring full retraining of the model [5] - The solution has been successfully applied in airport cargo and subway security inspections, transitioning from single item recognition to comprehensive behavior and event monitoring [6] Future Applications - The company plans to expand the solution to new areas such as rail transportation facility health monitoring, urban parking management, and road safety warning systems, leveraging existing successful cases for industry standard formulation and ecological cooperation [6]