Tracking of Multiple Objects Using Optical Flow Based Multiscale Elastic Matching

Abstract

A novel hybrid region-based and contour-based multiple object tracking model using optical flow based elastic matching is proposed. The proposed elastic matching model is general in two significant ways. First, it is suitable for tracking of both, rigid and deformable objects. Second, it is suitable for tracking using both, fixed cameras and moving cameras since the model does not rely on background subtraction. The elastic matching algorithm exploits both, the spectral features and contour-based features of the tracked objects, making it more robust and general in the context of object tracking. The proposed elastic matching algorithm uses a multiscale optical flow technique to compute the velocity field. This prevents the multiscale elastic matching algorithm from being trapped in a local optimum unlike conventional elastic matching algorithms that use a heuristic search procedure in the matching process. The proposed elastic matching based tracking framework is combined with Kalman filter in our current experiments. The multiscale elastic matching algorithm is used to compute the velocity field which is then approximated using Bspline surfaces. The control points of the B-spline surfaces are used directly as the tracking variables in a Kalman filtering model. The B-spline approximation of the velocity field is used to update the spectral features of the tracked objects in the Kalman filter model. The dynamic nature of these spectral features are subsequently used to reason about occlusion. Experimental results on tracking of multiple objects in real-time video are presented.

FMOE-MR: content-driven multiresolution MPEG-4 fine grained scalable layered video encoding

“…Motion tracking is itself an extremely well-researched area, and is clearly beyond the scope of this paper. We have implemented a novel tracking algorithm based on optical flow-based multi-scale elastic matching [13]. The algorithm can detect and track multiple objects moving in a video sequence.…”

“…We have implemented the F-Mask using the Canny edge detector [11], and the M-Mask/O-Mask using optical flow based multi-scale elastic matching algorithm given in [13]. In order to have an objective quantification of video quality, we have used the measure of PSNR.…”

“…A detailed description of the face recognition and tracking algorithm is again beyond the scope of this paper. Faces in a video sequence can be detected using the algorithm described in [16] and tracked using the algorithm described in [13] to create an O-mask based on human faces automatically.…”

“…This figure shows the dependence of PSNR and bit rate on the two MR-Parameters, σ L and σ I , by keeping σ H = 3 (fixing σ H to an empirically chosen value of 3 renders the foreground at a consistently good resolution). The two persons in Video 4 have been tracked using the algorithm described in [13], to create the motion mask (M-Mask), for the video sequence.…”

“…In order to compare the effect of the FMO-Mask on performance of FMOE-MR for Video 4 (Video 4 is chosen arbitrarily), we fix values of the MR parameters empirically as follows: σ L = 21, σ H = 3 and σ I = 1. We have used three kinds of masks: F-mask (edge mask, implemented using the Canny edge operator [11]), M-mask (object tracking mask, implemented using the algorithm described in [13]), and an optimistic combination of the two masks (FM-Mask). Figure 7 shows the multi-resolution re-rendering of the original frame based on the three types of masks mentioned above.…”