The Fusion Perception Suite is a combination product independently developed by CICTCI, which integrates perception fusion algorithms, calibration tools, deployment supervision tools, and evaluation tools. This kit includes calibration tools that are compatible with cameras, millimeter wave radars, and LiDAR sensors, enabling semi-automatic sensor calibration; Including roadside perception and lightning vision fusion algorithm, it can achieve functions such as intersection or road target fusion, trajectory continuation, etc; Includes deployment monitoring tools that can achieve fast algorithm deployment, remote OTA, device management, and other functions; Contains algorithm evaluation tools, making it easy to quickly output algorithm performance indicators. Featuring strong adaptability, ease of use, and convenience.

Key Features

• One-stop perception calibration, fusion, deployment supervision and evaluation
• Supports fast and convenient calibration, improving calibration efficiency
• Support integrated deployment, remote OTA and supervision, reduce operational difficulties and costs
• Rich data chart presentation, clear at a glance
• Fully functional and high-performance perception fusion algorithm
• Multiple evaluation parameters, simple methods, and support for customization
• Support real-time viewing of visual effects
• Adapt to the standard testing method of "Technical Requirements and Testing Methods for Vehicle Road Collaborative Roadside Sensing System"

Algorithm performance

• Target object recognition, event recognition, and traffic detection accuracy: ≥ 95%
• Target recognition recall rate: ≥ 90%
• Target positioning accuracy: ≤ 1m
• Algorithm processing delay: ≤ 80ms
• Track angle: ≤ 5 °
• Speed deviation: ≤ 1m/s

Applications

• Chongqing Western Science City Intelligent Networking Project

The Multi-access Edge Computing Unit (MEC) utilizes raw data collected by roadside sensing devices such as cameras and radars, to effectively identify key information in the road environment, including traffic participants, traffic events, and traffic flow. Traffic participants encompass various target types such as motor vehicles, non-motorized vehicles, and pedestrians, with simultaneous detection of crucial details like speed, heading, and position. Traffic events include over a dozen types such as abnormal parking, wrong-way driving, speeding, and pedestrian intrusion, providing a holistic perception capability for vehicle-road collaboration systems. This MEC (Multi-access Edge Computing) technology can be widely applied to various business scenarios such as vehicle-road collaboration, smart transportation, intelligent expressways, and smart parks.

Key Features

• Supports access to multiple heterogeneous sensors, providing high-precision integrated perception of traffic participants, traffic events, and traffic flow
• Supports intelligent decision-making and traffic infrastructure interaction for various business needs such as traffic management, traffic safety, transportation, supervision, and smart driving
• Designed for roadside edge computing with specifications including small size, high computing power, low power consumption, and wide temperature range
• Supports multiple communication interfaces including Ethernet, USB, RS232, RS485, CAN, and DIO
• Supports ultra-low latency computing, achieving millisecond-level holistic perception through efficient hardware-software integrated algorithms
• Equipped with comprehensive toolsets for automatic calibration, integrated deployment tools, and perception effect evaluation

The intelligent signal light collector is composed of three parts: signal light detection unit, internal calculation and analysis unit,and data transmission unit. It is installed in the intersection of signal machine to obtain the remaining time of the traffic movement phase of the signal light, and informs the vehicle communication unit in real time through the roadside communication unit to ensure that the vehicle passes the intersection safely and efficiently.

Key Features

• Can effectively obtain the current intersection phase status of Traffic light
• Can accurately obtain the remaining phase time of Traffic light and Support for obtaining 12 lights data at the same time, which can be flexibly expanded
• Support real-time data updates at 10Hz network port
• Support visual software debugging of basic parameters of signal light collector
• Support time synchronization through NTP protocol and Support adding timestamp to messages