In this paper we propose an algorithm that is able to detect moving objects, returning the number of found objects, together with their position, shape, and approximate distance. The system is based on two cameras, which are supposed to be fixed, a digital processor, and two analog chips, which perform data analysis. The use of a couple of cameras improves the performance in comparison with systems with only one camera, because it can exploit the availability of two images from two different points of view in order to get information on the distance of the objects from the two cameras, in the same way as the human eye does with its so called "binocular vision". We tested our method over several video sequences, both indoor and outdoor. Experimental results show a significantly improved discrimination when multiple objects are moving at different distances. Moreover, the use of stereo images can be exploited to reduce noise, improving performances for clustering.
A CNN-based Algorithm for Moving Object Detection in Stereovision Applications
Giovanni Costantini;
2007
Abstract
In this paper we propose an algorithm that is able to detect moving objects, returning the number of found objects, together with their position, shape, and approximate distance. The system is based on two cameras, which are supposed to be fixed, a digital processor, and two analog chips, which perform data analysis. The use of a couple of cameras improves the performance in comparison with systems with only one camera, because it can exploit the availability of two images from two different points of view in order to get information on the distance of the objects from the two cameras, in the same way as the human eye does with its so called "binocular vision". We tested our method over several video sequences, both indoor and outdoor. Experimental results show a significantly improved discrimination when multiple objects are moving at different distances. Moreover, the use of stereo images can be exploited to reduce noise, improving performances for clustering.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
