A three-dimensional measurement method based on mesh candidates assisted with structured light

Rendering three-dimensional information of a scene from optical measurement is very important for a wide variety of applications such as robot navigation, rapid prototyping, medical imaging, industrial inspection, etc. In this paper, a new 3D measurement method based on mesh candidate with structured light illuminating is proposed. The vision sensor consists of two CCD cameras and a DLP projector. The measurement system combines the technology of binocular stereo vision and structured light, so as to simplify the process of acquiring depth information using mesh candidates. The measurement method is based on mesh candidates which represent the potential depth in the three dimensional scene. First the mesh grid was created along the direction of axes in world coordinate system, and the nodes were considered as depth candidates on the surface of object. Then each group of the mesh nodes varying along z axis were mapped to the captured image planes of both cameras. At last, according to the similarity measure of the corresponding pixel pairs, the depth of the object surface can be obtained. The matching process is between the pixels in both camera planes corresponding to the spatial mesh candidates. Aided by the structured light pattern, the accuracy of measurement system improved. Appending the periodic sawtooth pattern on the scene by structured light made measurement easier, while the computational cost did not increased since the projector had no need to be calibrated. The 3DS MAX and Matlab software were used to simulate measurement system and reconstruct the surface of the object. After the positioned cameras have been calibrated using Matlab calibration toolbox, the projector is used to project structured light pattern on the scene. Indicated by experimental results, the mesh-candidate-based method is obviously superior in computation and accuracy. Compared with traditional methods based on image matching, our method has several advantages: (1) the complex feature extraction process is no longer needed; (2) the epipolar constraint is replaced by mesh candidates so as to simplify stereo match process; (3) the candidate selection strategy makes unnecessary the process of transformation from two dimensional coordinates to three dimensional coordinates.