Research On Algorithms Of Video Tracking Based On Active Contours

The motion analysis to a moving object in sequential images is a hotspot in computer vision.The motion analysis consists of moving object capturing,moving object segmentation,object tracking and object understanding.The object capturing,segmentation and tracking provide the foundation for motion analysis.And a well-segmented and tracked result will lead into a correct object understanding.In recent years,geometric active contour model based on the level sets develops greatly in the image processing domain.The method can extract the object's contour reliably,and it is less dependent on the initial condition.In order to allow computer vision system automatically segmenting and tracking the object continuously and at the meantime automatically getting the object contour information,the geometric active contour model method is used to the studies of moving object segmentation and tracking.The main research contents of this thesis are as follows:1.A new object tracking method that systematically combines region and image information is proposed.Traditional regional feature-based active contour models that use only regional features in the tracking process can achieve good results in some scenes,but can't be accurately tracked in complex scenes.There are great limitations in tracking process when using traditional methods.In this paper,we combine the region feature(color,texture)and gradient feature and propose a new active contour tracking method integrating multiple features into an energy function.The algorithm can effectively tracks the object in low contrast and complex scenes.In order to reduce the computational cost complexity,the information of the previous frame is used in the initialization of level set function in current frame and energy driving function and balloon force is added to the energy function.The experimental results show that the proposed model is more robust than traditional regional feature-based active contour models.2.A novel variational level set framework for contour tracking is proposed.Contour tracking can be implemented by measuring the probability distributions(e.g.color)of both interior and exterior regions of an object contour.Choosing a suitable distance metric for measuring the(dis)similarity between two distributions significantly influences the tracking performance.Most existing contour tracking methods,however,utilize a predefined metric which may not be appropriate for measuring the distributions.The image energy functional is modeled by the distance between the foreground distribution and the given template,divided by the distance between the background distribution and the template.The form of the distance between two distributions is represented by the quadratic distance.To obtain the more robust tracking results,KISS algorithm is employed to achieve the similarity matrix for the quadratic distance.In order to reduce the computational cost complexity,the information of the previous frame is used in the initialization of level set function in current frame and balloon force is added to the energy function.Experiments on several video sequences prove the effectiveness and robustness of our method.3.A multi-object tracking algorithm based on level set method is proposed.The current research on tracking based on active contours is mainly on the tracking of individual targets,and tracking of multiple targets in the actual scene is still immature.In this paper,a multi-object tracking algorithm based on level set is proposed.The improved inter-frame difference method is used to obtain the initial evolution curve of the level set,and the target template is established to identify the objects.We use the correlation between the template and the region of interest to build energy functions for the tracking of multi-object.Experiments on several video sequences prove the effectiveness and robustness of our method.