Zhongwei Feng

A Method for Extracting the Center of Ring-Structured-Light Stripe

Extracting stripe center accurately is very important in 3D measurement with structured light. A new method for extracting the center of ring-structured-light stripe has been proposed, which combines morphological thinning with improved barycenter method. Convert the original gray-scale image to a binary one, in which stripe stroke at pixel-level can be obtained by morphological thinning. Then the center of ring faculae is gained by curve fitting using the least squares method. Consider the directions pointing from the ring faculae center to every stroke point as normal directions of stripe center locus at its corresponding points. Along each normal direction, barycenter method is applied to the preprocessed gray-scale image. Compared to other stripe center extracting algorithms, the proposed method benefits from its simple intensity segmentation and no curve fitting for the curvature. Experiments show that it can meet industrial real-time applications for its rapidity as well as its sub-pixel level accuracy.

In-pipe profile detection using circular structured light and its calibration technique

A vision sensor utilizing circular structured light is proposed to deal with the in-pipe profile detection. Circular structured light is firstly thrown by the ring laser projector and then reflected by a cone-shaped mirror to intersect with the inner surface of the pipe. The intersection is taken as the representative of the contour of the section by the charge coupled device (CCD) camera and gets processed by a subsequent algorithm to extract relevant information. The imaging principle is analysed to relate the pixel value to the real dimension, and lead to the promotion of an effective and feasible calibration method which adopts a series of pipes with known diameters as calibration objects. This method figures out the mathematic relation between the pixel value and real dimension by interpolation instead of modelling the imaging system directly. With the calibrated system, pipes with known diameters are employed in the experiment. The feasibility and accuracy have been improved compared with the existing techniques.

Applying PSD Analysis in Ring-Structured-Light 3D Measurement

Error elimination is very important in 3D measurement with ring-structured-light. A method has been proposed, which includes the following procedures. First, acquire ring-structured-light images and extract their stripe center. Second, find the base-circle center of the stripe center locus by curve fitting, using the least squares method. Subsequently unfold the stripe center locus along its base-circle. In the waveform of the unfolded view, there exist an eccentricity error component and an ellipse shape error component. The frequencies of these two error components are different from the frequency of detail component. They can be separated in the unfolded view by power spectrum density (PSD) analysis. Based on their formation principles, eccentricity error and ellipse error can be separated and eliminated, and the accuracy of the measurement can be improved. Experiments have demonstrate its applicability.