Determining object shape from local velocity measurements

摘要

In this paper we discuss the following question: what can be said on the movement and shape of a planar rigid object from sparse samplings of edge crossing measurements. It is shown that if the object moves in translation over an array of detectors, providing local edge crossing information, one can determine the translation velocity and use a back projection process to obtain an equivalent sampling of the object along a set of parallel lines. Furthermore, if information from several such passes (with different translation velocities) over detector arrays is available one can combine the corresponding backprojected data-sets to determine classes of objects that are consistent with all the data. Such an array of detectors is shown to be a useful alternative way of acquiring information on dynamic scenes when the tasks are limited to positioning, velocity measurement and shape recognition of rigid two-dimensional objects moving on a contrasting background.