A Preliminary Study on Distance and Speed Estimation Using Event-Based Vision
Leonard Haensel, Torsten Bertram · 2024
This paper presents an approach for distance and velocity estimation based on event-based camera data. Event-based cameras also called Dynamic Vision Sensors (DVS) are bio-inspired sensors with an inherently different operating principle. Instead of conventional image acquisition, each pixel's brightness change is measured asynchronously. This results in a data stream that contains the corresponding events. Each event is clearly defined by its location, timestamp, and the sign of the respective change. This operation method has several advantages over a traditional camera, resulting in a high dynamic range, high temporal resolution, and low energy consumption. These attributes represent great potential in the context of autonomous driving. For this reason, a preliminary method for estimating the distance and velocity of other vehicles based on artificial event data will be presented here. In addition, the estimated values are compared with measured values from a radar sensor in order to gain an impression of their accuracy.