Video analysis is a growing research field that has many applications including video surveillance, video indexation etc. Object detection is one of goals in video surveillance applications. In which, one of the main difficulties faced by researchers is the correct object segmentation. Especially in outdoor scene, different phenomena may disturb object detection. These phenomena include variation in brightness level, moving background objects (e.g. ripples of water waves, moving leaves, escalators etc.) and cast shadows of surrounding buildings and trees etc. The methods used for automatic segmentation using stationary cameras are of different types: they can be based intensity/color or shape characteristics of object or background. Background modelling in the fixed camera situations is the first choice for object detection methods. However, when classic background modelling techniques are applied in moving background scenarios they produce low object detection rates. In fact, in moving backgrounds spatiotemporal information must be considered for a good background representation. We address the object detection in moving background with a frequency based background model.