A structured light vision measurement with large size M-array for dynamic scenes

In this paper a large size M-array based on pseudo-random numbers is proposed to implement spatial encoding structured light technique for dynamic 3D reconstruction. By using our technique, more accurate M-array can be generated. The large size M-array benefits in providing more information on object surfaces as more primitives can be contained in the pattern. We first discuss a pseudo-random array and point out problems existing ones. Then, we elaborate a new method to generate a large size M-array using pseudo-random numbers, which eliminates the limitation imposed on the coprime of rows and columns of pseudo-random array. To get more robust structured light system, color points are replaced with geometric symbols since color coded pattern are more sensitive to background and noise. The decoding method, in this research, is also improved, which significantly enhances accuracy of neighborhood finding in uneven or unbalance surfaces. Finally, 3D reconstruction results for cylinder and sculpture are presented to show the validity of the proposed large size M-array and decoding method.