Research On Visual Target Tracking Based On Convolution Filter

Visual object tracking is a task of computer vision and has been studied deeply.Research results have been used in the field of manless driving,video indexing,video surveillance and so on.The basic mission of visual tracking is that an object is identified manually in the first frame of a video sequence(using a rectangular bounding box),then a tracker is constructed to predict the location of the object in the subsequent frames by estimating its trajectory as it moves around.A typical tracking system consists of three components:(1)an appearance model,which can predict the likelihood that the object of interest is at some particular location based on the local image appearance;(2)a location model,which evaluates the prior probability that the object is present at a particular location;and(3)a search strategy for finding the maximum a posteriori location of the object in the current frame.Current tracking algorithms could be categorized into generative and discriminative models.Recently,a class of tracking techniques called synthetic exact filters has been used to construct appearance model and shown to give promising results at impressive speeds.Synthetic exact filters are trained using a large number of training images and associated continuous labels,however,there is not much theory behind it.In this paper,we theoretically explain the reason why synthetic exact filters based methods work well and propose a novel visual object tracking algorithm based on convolutional filters,which are trained only by training images without labels.Compared with the prior methods such as synthetic exact filters which are trained by training images and labels,advantages of the convolutional filters training include: faster and more robust than synthetic exact filters,insensitive to parameters and simpler in pre-processing of training images.Convolutional filters are theoretically optimal in terms of the signal-to-noise ratio.Furthermore,we utilize spatial context information to improve robustness of our tracking system.Experiments on many challenging video sequences demonstrate that our convolutional filters based tracker is competitive with the state-of-the-art trackers in accuracy and outperforms most trackers in efficiency.