Efficient methods for object recognition using the constellation model
Robert Fergus, Markus Weber, Pietro Perona · 2001
We present efficient methods for object recognition, using the constellation model. This model represents objects as constellations of rigid features, with the variability between them being represented by a joint probability density function. Using simple, exhaustive methods, the computational requirements for this approach quickly become prohibitive. By using A * search methods, accompanied by some heuristics, large performance improvements are achievable in recognition, making real-time recognition on large images possible, using complex object models. 1.