Abstract:
Moving object detection has been widely used in
intelligent surveillance. This paper proposes a new moving
object detection scheme in based on gradient information for
real time detection. In the proposed moving object detection
scheme, the input and background image are converted into
gradient map. Then gradient difference map is calculated and
proper masking over the map extracts moving objects. The
proposed moving object detection scheme is tested by various
video sequences to demonstrate the robustness of moving object
detection. Simulation results suggest that the moving object
detection scheme embedded with the proposed method are
highly robust to illumination change, presence of noise and
distracting motions. We observe that the proposed method
provides better performance than existing edge based methods
in terms of displacement error and false detection because of
improved edge localization by new gradient detection and
gradient directional masking. Our proposed scheme achieves
25% reduction in displacement error than traditional edge
based methods.
