International Journal of Computer Applications (0975 – 8887) Volume 116 – No. 20, April 2015 15 Moving Object Extraction through a Real-World Variable- Bandwidth Network using KFDA-based RBF Ramakant Verma Assistant Professor, CSE Dept. BIT Meerut Maitreyee Dutta, Ph.D Professor& Head, CSE Dept. NITTTR, Chandigarh ABSTRACT Motion detection has become one of the most important applications in traffic monitoring systems. Video communication in traffic monitoring systems may suffer network congestion or unstable bandwidth over real-world networks with definite bandwidth, which is dangerous in motion detection in video streams of variable-bit-rate. In this paper, we propose a unique Kernel Fisher’s linear discriminant (KFLD)-based radial basis function (RBF) network for motion detection approach for accurate and complete detection of moving objects in video streams of both high and low bit rates. The proposed method will be accomplished through a combination of two stages: pattern generation (PG) and motion extraction (ME). In the PG stage, the variable - bit- rate video stream properties will be accommodated by this new technique, which subsequently distinguishes the moving objects within the segmented regions belonging to the moving object class by using two devised procedures that is Background Discriminant Procedure and Object Extraction Procedure during the ME stage. The accuracy result evaluations can show that the new method exhibits superior when compared to the old methods. General Terms Motion detection, Spatiotemporal, Gaussian Function Keywords PCA, RBF, KFDA,KDA 1. INTRODUCTION Identifying moving objects from a video sequence is a fundamental and critical task in many computer-vision applications. A common approach is to perform background subtraction[1], [2], which identifies moving objects from the portion of a video frame that differs significantly from a background model. There are many challenges in developing a good background subtraction algorithm. First, it must be robust against changes in illumination. Second, it should avoid detecting non-stationary background objects such as moving leaves, rain, snow, and shadows cast by moving objects. For example, by spatial and temporal smoothing, we can remove snow from a video. Small moving objects, such as moving leaves on a tree, can be removed by morphological processing of the frames after the identification of the moving objects. The problem with background subtraction is to automatically update the background from the incoming video frame and it should be able to overcome the following problems: • Motion in the background: Non-stationary background regions, like branches and leaves of trees, a flag waving in the wind, or flowing water in background, can be identified as part of the background. • Illumination changes: The background model should be able to tolerate, gradual changes in illumination over a period of time. Because since morning to evening the sun light change. • Memory:To detect moving object the background module should not use much resource, in terms of computing power and memory. • Shadows: Shadows cast by moving object should be identified as part of the background and not foreground, so to remove shadows extra logic and processor cycle is required. • Camouflage: Moving object should be detected even if pixel characteristics are similar to those of the background. • Bootstrapping: The background model should be able to maintain background even in the absence of training background. 2. PCA In PCA [3], the main concept is to express the available dataset to extract the relevant information by reducing the duplicate attributesand minimize the noise. PCA didn’t concern about whether this dataset represent features from one or more classes, i.e. the discrimination power was not taken into consideration in PCA.In PCA, it had a dataset matrix X with dimensions mxn, where columns represent different data samples.Zero mean dataset is created by subtracting the mean from each sample of dataset matrix X, after that covariance matrix Sx = XX T is created.Now Eigen values and Eigen vectors were then computed from Sx. So the new basis vectors are those Eigen vectors with highest Eigen values. Eigen vectors having lower Eigen values are eliminated.Thus, using the new basis, algorithm can project the dataset onto a less dimensional space with more powerful data representation. This paper presents concepts and working program to detect motion in a video sequence using Kernel Fisher Discriminant Analysis (KFDA). Prior to using KFDA for motion detection, a histogram-based [6],Sigma DifferenceEstimation(SDE) [4], MultiSigma Difference Estimation(MSDE)[5]Mean based[6],Median and Histogram based [7]sand PCA and many more technique were used. Thesetechniquesdid not provide as good results as the KFDA based system where speeds of computation and sensitivity to motion were of primary concern. Using KFDA to provide the feature descriptor of consecutive video frames along with dynamic thresholding provided very good results in detecting object larger than the video segmentation block used. The PCA algorithm presented here used two different PCA matrix generation methods: global and local; and two thresholding methods: static and dynamic. Collected data shows that the local PCA generation with dynamic thresholding may be the best combination of