Research On Algorithms For Video Moving Objects Detection Based On Fusion Of Spatial-Temporal Information

Because the terroristic events happened frequently these years, people begin to pay their attention to the society security. The video surveillance technology becomes a hot point gradually in related fields for researches and applications as an effective defending measure.As computer vision technology develops much faster, traditional video surveillance system can not satisfy the requirements for security in complex conditions. So the intelligent video surveillance technology for moving objects detection, objects classify, objects tracking and behaviors comprehension appears.Moving objects detection is an important part in intelligent video surveillance system,the results of detection will affect the system's later work directly. So the research on methods of moving objects detection is significative.Several algorithms of moving objects detection are researched in this paper. The Kalman low-pass filter algorithm is mainly studied, and its improved algorithm is proposed, in which gaussian probability density function is used to express the gain factor. Experimental results show that this algorithm is fast, effective and robust.Another improved algorithm, in which spatial and temporal information are used for detection is proposed. The Kalman low-pass filter and edge detection are combined to remove the noises, which are produced by a sudden illumination variation or objects moving and stopping. The new method can detect moving objects in different environments more effectively.