Nowadays, vision sensor has been introduced in detecting object in day time or night time. Vision sensor can be applied in many areas, such as in medical system, military system, transportation system, robotic and control system, and surveillance system. Using vision, such systems can detect, recognize and actuate accurately depend on how good the image has been processed.
In the past, a frame difference using online K-means approximation is one of the most popular methods. In this technique, computation cost and memory requirements are low. However, the foreground segmentation accuracy inevitably decreases due to the loss of color information.
To solve the problem of accuracy and resource requirements of video surveillance detection, this paper proposes a background subtraction method by means of Kalman filtering that can be used on both dynamic and static videos.
Background subtraction is a low-level task and we consider two aspects: accuracy and computational resources (time and memory). Background subtraction performance depends mainly on the background modeling technique.
The experiments were done using MALAB 7.8 in a 2.8 GHz Intel Dual core 860 CPU. The grayscale images are generated from the original color images. Secondly, Kalman filter images are generated from the original color images.
To read the external content in full, download the complete paper from the author online archives.