Research On The Depth Acquisiton Algorithm Based On Structured Light

It is convenient to separate the foreground and background of the scene accordingto depth images, and thus can improve the performance of nature human-computerinteraction (HCI) greatly. The popular use of HCI depends on the performance of thedepth sensor such as price and real-time. The depth sensor based on surface structuredlight has high practicality and is applied widely. But now the depth camera is expensive.Depth image calculation technology is still monopolized by a few foreign companies.The research on depth image calculation, not only makes a contribution in the academicfield, but also helps to break the foreign technical barriers and improves the depthcamera manufacturing industry in our country.The depth information is gained at a speed of30frames per second using thematching algorithm of image block, which is improved by the interpolation algorithmand filtering algorithm. What’s more, the image resolution is1cm. Only adding aninfrared transmitter, the Smart TV and personal computer can realize the nature HCIusing the method easily.The matching algorithm is like linear search, which needs to search the optimalmatching block in the image. The depth acquisition method called DepthVH isintroduced based on the pattern recognition of vector field. The depth information ofscene is obtained by feature recognition of vector field directly for every depth point. Inthis method the vector field is firstly generated based on depth feature, then the depthinformation hidden in the vector field is converted to the feature information by thepattern recognition technology, at last the depth information is obtained by featurematching directly after the inverse transformation between the feature information andthe depth information. During the processing of the inverse transformation, theDepthVH method adopts the searching strategy similar to the Hash Mapping, whichavoids the complexity of the linear search. The depth acquisition method calledDepthST uses SVM to realize the decision tree, which can get the depth value of eachpixel point quickly. Though the precision of DepthVH and DepthST algorithm is worsethan matching algorithm, but the amount of calculation is much smaller than matchingalgorithm. DepthVH and DepthST can be applied in the manufacture of low-costcamera.