Research On The Boundary Detection Of Coded Structured Light Stripe

In structured light 3D vision, the stripe features, such as edges and corners, of the coded structured light images should be extracted before the 3D reconstruction. Because the three dimensional shapes of the objects and the noises during the collection and transmission of the images, the coded structured light stripes in the collected images may have deformation. The positions, directions and scales may also change. The desired feature detectors should deal with these changes. The precision of the detected features will determine the accuracy and the correctness of the following camera calibration and 3D reconstruction. The proposed steerable filters will solve these problems.The steerable filters can be rotated to any direction by using a set of basis filters and interpolation functions. They have good direction selectivity. We can shift the steerable filters arbitrarily in both phase and orientation with the help of Hilbert transform. In this paper we will adjust the parameters of the filters to detect features in different scales. The paper will firstly take pictures of the coded structured light images. The main task of the project is to accurately extract the structured light stripe features from the pictures we take using the steerable filters. We also further design the steerable filters based on the cosine-function-like edge model.In order to reconstruct the 3D information of the objects precisely, we need to get the sub-pixel positions of the stripe features. In the grid model the corners and the edges are extracted respectively. Then the paper will match the corners with the edges and link the broken edges near their corresponding corners, and remove the false edges out.