| 摘要 |
<p>The invention is a passive method to measure the translational speed of a visual scene using the distribution of light intensities. Measuring the speed of translation is useful for control, safety, management of resources, fuel efficiency, and many more application fields. It is however technically challenging because a wide-field translating scene projects on an image plane as a heterogeneous field of apparent speeds. The invention solves this problem by combining two principles: perspective distortion matching over a broad field of view, and temporal filtering variation. In conventional systems, an acquired image is calibrated to obtain linear coordinates. Instead, the invention uses the perspective distortion of the image to sample the visual scene at different linear wavelengths over the visual field. The result is a spatial sensitivity map of the visual scene. The obtained signal is then temporally filtered with cutoff frequencies proportional to the spatial sensitivity. The final result is the wide- spectrum computation of the ratio between temporal and linear spatial frequencies, in other words linear speed. The technique is passive because it does not require the emission of a reference signal. This is an advantage over active speed sensors mainly because of reduced power consumption, but it is an enabling factor for other applications. Where it is difficult or impossible using standard device-centered techniques, like on aircrafts or in fluids, the invention enables measuring absolute linear speed. The advantage over non-device-centered techniques -like GPS- is the independence from external infrastructures. The small computational overhead makes it ideal for mobile applications.</p> |
