Schmalz and Angelopoulou Structured light decoding

Schmalz and Angelopoulou Structured light decoding

This method is able to cope with colored and textured objects as well as with challenging im-age quality. We give a short summary of the method here. 

First, a watershed transform of the input image is performed. From the superpixels in the resulting oversegmented image a region adja-cency graph is built.Each superpixel becomes a vertex in the graph. The color assigned to the vertex is the median of the original image pixels belonging to that superpixel.The edges of the region adjacency graph describe the color changes between neighboring vertices. They are scored according to how well they fit possible color changes in the projected pattern. To find correspondences be-tween the projected pattern and the camera image, unique se-quences of edges representing the color changes in the projected pattern have to be found in the region adjacency graph. If a match-ing sequence is found, the correspondence information is propa-gated to the neighboring vertices in a best-first-search. Once all possible regions have been mapped to projected stripes, the stripe edge locations in the original image are localized to subpixel preci-sion. The triangulation between the projected light planes (or cones in our case) and the viewing rays from the camera can then be performed to obtain 3D data. 


The advantages of this graph-based decoding method are the robust color assignments of the superpixels, the absence of fixed thresholds for color changes and the ability to sidestep any disruptions of the pattern in the camera image by finding alternative paths in the region adjacency graph.

Furthermore, the implementation is fast and can be run in real-time on current hardware.

Chinese Translation Vision

---

版权声明
本博客所有的原创文章，作者皆保留版权。转载必须包含本声明，保持本文完整，并以超链接形式注明作者编程随想和本文原始地址：
https://idreamboats.blogspot.com/2013/03/schmalz-and-angelopoulou-structured.html

---