Fusion Of Multimodal Video Sequences For Human Detection and Tracking

Tracking of multiple moving objects is the process of locating moving objects from a sequence of consecutive frames in a video. In this paper, video fusion is performed for detection of moving objects and the position of the objects are determined using Motion Detection Algorithm. The main difficulty in target tracking is to associate target locations in consecutive video frames, especially when the objects are moving fast relative to the frame rate. In this work, video fusion is based on Multicontourlet Transform (MCT) which improves the geometric resolution of the images.

The multicontourlet is flexible multiscale and multidirection image decomposition. With better direction selectivity and energy convergence compared to that of a multiwavelet, a multicontourlet is suitable for representing remote sensing images bearing abundant detailed and directional information. The fusion weight of the low-pass coefficients is selected adaptively based on the golden section algorithm. For the high-frequency directional coefficients, the local energy feature is employed to select the better coefficients to fusion. Experimental results show that the proposed method achieves better visual quality and objective evaluation indexes than a wavelet-transform based, a contourlet-transform-based, and a multi- wavelet transform-based weighted fusion method.

Tracking of moving object uses an optical flow estimation technique to estimate the motion vectors in each frame of the video sequence. By thresholding and performing morphological closing on the motion vectors, the binary feature images are produced. In this demo, you locate the personsin each binary feature image using blob analysis. Then you draw a rectangle around the cars that pass beneath a reference line. You use a counter to track the number of cars in the region of interest.

Motion Vectors

Thresholding and Region Filtering