A Novel Peak Detection Method of Structured Light Stripes for 3D Reconstruction

We present a novel stripe peak detection method for structured light based 3D reconstruction. With such approach, we achieve accurate peak location even for weak stripes. The process consists of two steps. In first step, an algorithm named Local Trend Measurement is proposed to segment stripe areas. Instead of detecting peaks by amplitude and width, we detect stripe peak by its illuminance rising edge and falling edge, which are derived by Local Trend Measurement. Pixels between the edges are labeled as stripe area. In second step, a modified weighted average method is applied to determine accurate peak position in stripe area. Experimental results demonstrate that our method is effective in both stripe segmentation and peak position derivation.