We present a novel method to estimate the surface normal of an object at high-speed with high dynamic range using an event camera and a rotating light source. Conventional photometric stereo methods use RGB frame cameras and at least three light sources. As the number of light sources increases, the accuracy increases, but the frame that needs to be captured increases, which means that the speed of a normal map estimation is slower. Also, conventional photometric stereo methods are implemented in the dark room to avoid ambient light illumination. To overcome these limitations, the novel method employs an event camera. The event camera operates at microseconds resolution with negligible motion blur, and outputs a continuous stream of events that measures log intensity changes of each pixel asynchronously. Applying these properties, we add a light source rotating around the event camera to have the same effect as numerous light sources. Our method estimates normal vectors by analyzing how the light intensity changes as the light source rotates. Experiment results demonstrate the novel approach for the 3D normal vector estimation using the event camera based on photometric stereo.
|