A shape-based voting algorithm for pedestrian detection and tracking

<p>This paper presents the MOUGH (Mixture of Uniform and Gaussian Hough) Transformfor shape-based object detection and tracking. We show that the edgels of a rigidobject at a given orientation are approximately distributed according to a GaussianMixture Model (GMMs). A variant of the Generalized Hough Transform is proposed,voting using GMMs and optimized via Expectation-Maximization, that is capable ofsearching images for a mildly-deformable shape, based on a training dataset of (possiblynoisy) images with only crude estimates of scale and centroid of the object in eachimage. Further modifications are proposed to optimize the algorithm for tracking. Themethod is able to locate and track objects reliably even against complex backgroundssuch as dense moving foliage, and with a moving camera. Experimental results indicatethat the algorithm is superior to previously-published variants of the Hough transformand to Active Shape Models in tracking pedestrians from a side view.</p>