Trinocular data registration using a three-dimensional self-organizing feature map

A three-dimensional self-organizing feature map (SOFM) that associates redundant and complementary features extracted from images acquired by a trinocular camera system is described. The combined features extracted from three views of the reference parts are used to train the SOFM. The unsupervised learning algorithm ensures that “similar” feature vectors will be assigned to cluster units that lie in close spatial proximity in the 3D feature map. The technique reduces the dimensionality of the input by exploiting hidden redundancies in the training data. During the identification phase, features in the novel test part activate a number of cluster units that have weights similar to the applied training input. If the sum-of-square error (SSE) between the input and weights of the cluster unit with the strongest response is greater than a predefined tolerance, then the test object is labeled as faulty part