Three-Dimensional Hand Reconstruction by Single-Shot Structured Light Line Pattern

Abstract

3-D hand reconstruction is required in many applications, e.g., hand gesture controls, hand modeling, biomechanical analysis, humanoid robot hand design and analysis, hand surgery, and so on. In most situations, it is impossible or difficult for one to keep his hand static. Thus, real-time single-shot 3-D hand reconstruction is usually required in most applications. As a result, the popular phase shiftbased structured light methods could not meet the requirement, because they need at least three images to reconstruct the hand. In this paper, we propose a single-shot 3-D hand reconstruction approach based on the structured light line pattern. The proposed approach calculates a line coordinate shift map from the distorted line pattern in a single image and then reconstructs the 3-D hand with the line coordinate shift map and the calibration parameters. To calculate the line coordinate shift map accurately, we propose a specific line clustering method based on the features of the hand. We also propose a slope-based compensation method to correct the distorted line coordinate shift map. Experimental results verified the effectiveness of the proposed line clustering method based on hand features, the proposed coordinate shift map correction method, and the proposed 3-D hand reconstruction approach.

Keywords

• Image processing
• line segmentation
• line clustering
• three-dimensional hand reconstruction
• line coordinate shift map