Mapping & Visual Localization

Abstract/Technology Overview

Autonomous vehicle technology is aiming to increase road safety, increase transportation efficiency, relieve travelers from driving chores, and enhance mobility of all-age people and the disabled. As one of the essential modules, self-localization technology enables the autonomous vehicle to know its position along the journey and decide the movements for the next moments. 

For self-localization, the commonly used methods are based on the expensive GPS system or 3D LiDAR. However GPS system will fail inside tunnels or near to skyscrapers. Lidar will also fail in the environment with few changes like tunnels. The proposed technology enables a low-cost solution to the self-localization. The technology builds prior map by cameras only or fusing the information from GPS, Lidar and cameras. For localization, only cameras are used to reliably determine the locations of autonomous vehicles in lane. The technology can work in challenging scenario like rain, night, tunnel and flyover. It can also be used in conjunction with other localization technologies to achieve fail-safe operation.

Technology Features, Specifications and Advantages

The proposed technology mainly comprises cameras installed on a vehicle and a processing unit. In practice, the cameras can be installed at multiple positions such as in the front, side, and back of the vehicle. The software algorithm on the processing unit takes in the camera frames and derives the position of the vehicle with respect to the map. Such information can then be used in decision making module such as path planning for the vehicle. The mapping software is running offline. The localization software algorithm is running in real-time.

A) Mapping System Requirements

  • Monocular cameras (single or multiple)
  • Processing Unit (PC)
  • Optional sensors: Lidar (for high-precision localization up to 20cm), GPS (for geo-reference localization)


B) Localization System Requirements

  • Monocular cameras (single or multiple)
  • Processing Unit (PC or embossed board)
  • Optional sensors: CAN bus (for speeding up computation)


C) Accuracy

  • Mapping: 20 cm on average
  • Localization: 20cm - 50cm on average


D) Processing Time

  • 15 Frames per seconds on average (4 cameras on PC or embedded board)

Potential Applications

The proposed technology can be applied to Advanced Driving Assistance System (ADAS), Autonomous Vehicles (AV), and robotics applications. The technology can provide the vehicle position that is more accurate than normal GPS and can be used for navigation system, path planning in AV and robotics. It can be extended to smart logistic solution to improve productivity and efficiency with automated vehicle.

  • ADAS
  • Autonomous Vehicle
  • Robotics
  • Logistics

Customer Benefit

Provides accurate localization of 20cm-50cm with real-time processing, which enables the autonomous system to make accurate decision of the vehicle path.

Low cost. The camera-based solution is lower in cost compared to 3D LiDAR or expensive GPS system.

Additional applications. Other applications can also be developed such as motion flow analysis. With self-localization information, the customer can derive the movement information of the objects in the surrounding such as object pose, moving direction and speed. With such information, the customer can predict the potential risk of objects and provide warning or taking action before the accident occurs.

Technology Owner

Pongsak LASANG


Panasonic R&D Center Singapore

Technology Category
  • Artificial Intelligence
  • Human-Computer Interaction
  • Mobility
  • Robotics & Automation
  • Video/Image Analysis
  • Video/Image Processing
Technology Status
  • Available for Licensing
Technology Readiness Level
  • TRL 6

Autonomous vehicle, Self-localization, Visual mapping