When extreme scenarios push the limits of every perception system.

Nanoradar integrates advanced 4D imaging radar with vision in a deep fusion.

Launch In-vehicle Radar-Vision Integrated Unit RVF721 

A new paradigm for all-weather, high-precision, low-cost assisted-driving perception

LiDAR-camera integration

Truly integrated design: no calibration required, plug-and-play 
Compared with the conventional split-type radar-plus-camera solution, the integrated architecture eliminates the need for spatial calibration among multiple sensors, significantly reducing the complexity of installation, calibration, and long-term maintenance, and lowering overall hardware and calibration costs by approximately 30%。

4D imaging radar

The Leap from “Target Point” to “Target Imaging”

4D imaging radar serves as the core perception foundation for the integrated radar–vision system, leveraging VAR virtual aperture and MIMO technologies to increase point-cloud density from 256 to 1,200 points per frame, thereby providing high-density point-cloud input for sensor fusion.

Feature-level early fusion

Enabling true “collaborative operation” between radar and vision 
RVF721 employs feature-level early fusion, fusing raw sensor data and leveraging a unified feature network to enable multi-modal collaborative learning. Compared with late fusion, it achieves more accurate weak-object detection and improves stability in extreme scenarios by 30%.

Core Upgrade 
Three Core Capability Upgrades Enabled by Pre-Fusion of Radar and Vision 
1. Leap in Perceptual Capability — Overall Capability > Single Sensor

Compared with single-sensor systems, radar–vision fusion offers complementary advantages: radar provides range and velocity measurement, while vision enables semantic recognition, supporting the identification of more than 13 object categories, reducing the false-alarm rate by over fivefold, and significantly enhancing overall perception capabilities.

2. All-weather stable perception—reliable even in extreme environments

Leveraging the immunity of millimeter-wave radar to lighting and weather conditions, visual information is fused at the feature level to ensure system stability and reliability even in extreme environments such as rain, fog, nighttime, and backlit scenarios.

3. Feature Fusion Network—Enabling Data to Achieve “1 + 1 > 2”

While post-fusion merely concatenates features, early fusion leverages a deep-learning approach—combining a radar branch (PointPillars), a vision branch (CNN), and Transformer-based fusion—to boost object-detection mAP by 5%–8% and improve long-range detection accuracy by 15%.

Dual Security Assurance

Independent radar redundancy display ensures situational awareness even in the event of visual failure.

To enhance safety redundancy, Nanoradar has upgraded the display interface by integrating a dedicated radar window in the lower-right corner, which presents point-cloud or target data in real time. Even if the visual system fails, the radar can still operate independently to ensure system safety.

Product Specifications

Core Technical Parameters

Application Scenarios 
Comprehensively addresses diverse requirements, including vehicle obstacle avoidance, compliance with UN ECE R158/R159 regulations, and applications in construction and commercial vehicles, thereby driving the safe advancement of intelligent driving!