A Pedestrian Location Estimation Method Based on Multi-Sensor Cooperation

Pedestrian safety is a critical concern in intelligent transportation systems. The earlier a driver becomes aware of a pedestrian’s presence and location, the more promptly they can respond to ensure safety. However, for battery-saving and privacy considerations, pedestrians generally do not enable GPS. As a result, vehicles must rely on their own sensors to detect pedestrians. Previous studies have demonstrated that cameras can be utilized to estimate the distance of objects from the camera. Leveraging this capability, vehicle-mounted cameras can be employed to estimate the distance to pedestrians. However, such distance information is only beneficial to the vehicle that performs the estimation. In contrast, if a pedestrian’s location can be estimated, this information can be transmitted to other vehicles via vehicle-to-everything (V2X) communication. This would enable other vehicles to be aware of the pedestrian’s presence and location, even if they do not directly detect the pedestrian. In this study, we integrate sensor and image data to estimate pedestrian locations. The sensor data include the vehicle’s GPS location, orientation, and speed, which can be obtained from vehicle sensors or driver’s smartphone. The image data are captured using off-the-shelf dashcams without requiring knowledge of the cameras’ detailed parameters. The minimal hardware requirements enhance the feasibility of our proposed approach. Finally, we validate the effectiveness of the proposed method using data generated from the CARLA simulator.