Research On Target Segmentation And Tracking Technology Based On Depth Information

In the process of segmentation and tracking, for the weather, illumination, complex background and other traditional methods are not well resolved. This paper proposes a method to split the fusion target depth information for tracking parts in an Efficient Mean Shift tracking algorithm similarity measure as the main object of study. Segmentation and tracking through Matlab software simulation to validate the algorithm to achieve the effect.Split part, the traditional image segmentation methods can be broadly divided into two categories:edge-based segmentation and region-based segmentation method. Traditional segmentation algorithm over-segmentation seriously interfere with the extraction of the real goals. This paper presents a new approach, the first image fusion depth information combined statistical area, which can reduce interference when dividing complex background, get initial segmentation region. if the target with a similar background is high, there will be over-segmentation phenomenon. in order to solve the over-segmentation problem that requires further processing, we get the initial segmented regions of color fill, prepare the next segmentation, segmentation with graph cuts off part of the deal, get the precise target. This approach is the integration of a variety of algorithms to obtain satisfactory results. Depth information fusion segmentation image segmentation is a new technology, it turns out, integrate multiple segmentation method is easier to get the desired results. Experimental results show that the method can solve many practical problems, a certain versatility, especially for more complex background image, a significant treatment effect, reduces interference color similar.Tracking section, an overview of the basic theory of mean shift, introduced the efficient mean shift tracking method based on similarity measure, and pointed out the advantages and disadvantages and application of the algorithm.The new algorithm is a good segmentation and tracking of the target, when the target and the background is similar, but also can effectively target and background separation, so that the tracking results more accurate. Through experiments conducted experiments on the new algorithm analysis to prove the new algorithm is similar to the target and background, or object movement faster, can be accurate segmentation and tracking.