Search

The efficient and accurate tracking of a target in complex scenes has always been one of the challenges to tackle. At present, the most effective tracking algorithms are basically neural network models based on deep learning. Although such algorithms have high tracking accuracy, the huge number of parameters and computations in the network models makes it difficult for such algorithms to meet the real-time requirements under limited hardware conditions, such as embedded platforms with small size, low power consumption and limited computing power. Tracking algorithms based on a kernel correlation filter are well-known and widely applied because of their high performance and speed, but when the target is in a complex background, it still can not adapt to the target scale change and occlusion, which will lead to template drift. In this paper, a fast multi-scale kernel correlation filter tracker based on adaptive template updating is proposed for common rigid targets. We introduce a simple scale pyramid on the basis of Kernel Correlation Filtering (KCF), which can adapt to the change in target size while ensuring the speed of operation. We propose an adaptive template updater based on the Mean of Cumulative Maximum Response Values (MCMRV) to alleviate the problem of template drift effectively when occlusion occurs. Extensive experiments have demonstrated the effectiveness of our method on various datasets and significantly outperformed other state-of-the-art methods based on a kernel correlation filter.

One of the mayor goals in computer vision is object representation. Object representation aims to determine a set of features that best represents a specific object in an image, for example interest points, edges, color and texture. However, objects are generally composed of several regions containing different information which is more or less convenient to be represented by one of these features. Furthermore, each of these regions could be static or moving with respect to each other. In this sense, this paper presents an object representation based on fuzzy color blobs and spatial relationships among them. This approach of object representation is used to track rigid and articulated objects.En visión por computadora una de las metas principales es la representación de objetos. Dada una imagen, la representación de objetos intenta definir un conjunto de características que mejor los representa a partir de la información visual. Por ejemplo, se han utilizado puntos de interés, bordes, color y textura. Sin embargo, los objetos están formados por diferentes regiones con información variada que es más o menos conveniente a ser representada por alguna de las características antes mencionadas. Además, cada una de estas regiones puede estar estática o en movimiento respecto al resto de las regiones que componen el objeto. En este sentido, este trabajo presenta una representación de objetos basadas en regiones de color difuso y sus relaciones espaciales. Este enfoque es utilizado para el seguimiento de objetos rígidos y articulados.