Detection of object in motion using improvised background subtraction algorithm

The video streams are analyzed for extracting useful information and that information is further used in other high end applications. This has become a dynamic area of research in the field of computer vision. The growing importance is a result of its application in fields like human tracking and activity recognition systems, video surveillance, traffic monitoring and controlling systems etc. The first step in such applications is differentiating the foreground object i.e. the region or object of interest from background objects that is the complementary set of pixels. Detection of objects in motion from any video stream does not require any prior stored database but only multiple consecutive frames. Background subtraction is one of the simplest and most widely used methods for detection of moving object from a video sequence. Any background subtraction algorithm demands a stable background for efficient performance which makes the use of existing approaches very difficult in complicated real time applications. In this paper an approach based on color features is proposed. The color difference histogram (CDH) is computed in a locally existing small neighborhood which deals with the problem occurring due to dynamic backgrounds (e.g. running water, waving trees). Then, fuzzy c-means clustering (FCM) is employed to obtain fuzzy color difference histogram. This helps to overcome the impact generated due to change in illumination (sudden or gradual). Background modeling is done and foreground is detected using similarity matching. The object is further tracked using frame differencing. The proposed algorithm will work robustly in complex environments such as non-stationary backgrounds, varying illumination, frames with camouflage etc.