Detection of three-dimensional objects based on phase-encoded range images

The problem of three-dimensional (3-D) object recognition is addressed. Range images permit the description of 3-D surfaces with a significant reduction of the complexity of the processing. The encoding of range images as phase gives a simple and straightforward method for parallel processing of these images. In this contribution we propose the use of range images encoded as phase to deal with the detection of three dimensional objects. The use of Fourier transform profilometry can serve for real time optical implementation of correlators based on this encoding, using a joint transform correlator or a VanderLugt one. The extension to scale and in-plane rotation invariant detection is discussed and implemented in a simple optoelectronic system. Furthermore, a method for full 3-D rotation invariant detection is presented. A simplified approach allowing only rotation around an axis perpendicular point of view permits the implementation with a conventional correlator.