M. Leo, P. Spagnolo, and G. Attolico (Italy)
Pattern Recognition, feature extraction, neural network, video surveillance
People detection in outdoor environments is one of the most important problems in the context of video surveillance. Many example-based learning techniques have been used in the last years, the most diffuse of which are based on the analysis of different features of the shapes segmented in the scene. The choice of the features is a crucial and an hard step. In order to overcome this obstacle, this work compares the classification performance of a people detection system when two of the most used types of features are supplied as input. The features compared are: the coefficients of the Discrete Wavelet Decomposition and the geometrical measures of the objects in the scene. The experiments have been performed on real image sequences acquired in a parking area. The results have shown that the system performances are good with both kinds of features. However, the implementation simplicity and computational speed encourage to use geometrical features rather than the coefficients of the wavelet decomposition.
Important Links:
Go Back
