Article Preview
Top1. Introduction
Act of locating and tracking an object via an online tracking mechanism in a video finds its usefulness in wide-ranged applications such as human-computer interaction, surveillance, activity recognition, motion analysis(Henriques, Caseiro et al. 2015). On tracing a position of an object in a video initially, it is trailed with an ease in upcoming frames too(Yu, Li et al. 2019). Hence, some worthy information is acquired out of tracking the object regarding its activity as well as the identity criterion of the object concerned(Choi, Lee et al. 2016)(Zhang, Yao et al. 2013, Gündüz and Acarman 2019). The vital criterion to be kept in line with the concept of object tracking is resistance towards the illumination alterations in the foreground of an image leaving behind the background region suppressed(Gündüz and Acarman 2019). So as to accomplish such a standard many useful algorithms such as Robust Fragment-based tracking (FragTrack), Incremental Visual Tracking (IVT), Multiple Instance Learning (MIL) and Graph based Discriminative Learning (GBDL) are employed that incurred a huge lot of resources and complicated notions to resist with the altering environments from where the object is tracked(Liu, Zhang et al. 2016)(Hu, Li et al. 2015). Robust optimization algorithm or multifaceted appearance model is necessitated for confining the object from the entire search space.
On providing some inventive tracking strategy by means of augmenting few particularized structure information, the object can be tracked but with some other shortcomings such as insufficient pixel values, distribution of colors and other texture descriptors. Hence, the total number of trackers is highly essential for tracking an object in a successful manner(Lee, Kim et al. 2018). This sort of strategy consequently surges the overall complexity of the system. Though there is a limitation in realizing the innermost structure of an imagery with respect to the existence of occlusion and structure deformation in an abstracted image. In the online tracking of an object, issues arise with segregation of background from the foreground in a frequently altering environment.(Fu, Angelini et al. 2019) Confining the interruption of background from the foreground that is the targeted image is completely subjected to a regular subjugation of tracking strategy getting deviated towards other objects. Conventionally utilizing bounding box approach does not suffice the need for segregating the background in presence of an occlusion.
As a means of resolving such issues, a sparse representation based pattern recognition methodology is introduced. It is capable of recognizing objects by revealing an appropriate patch that ultimately associates with the object concerned but this strategy is capable of controlocclusion only to a limited extent and it still exists. Hence, the overall quality of the tracked object gets affected in terms of quality.
The novel technical contributions of proposed work are listed as follows:
- •
A novel background normalization technique deployed on the basis of textural pattern analysis of the targeted object termed as Laplacian Chime Pattern (LCP) technique illuminates and enhances the visual quality of an image even in poor illumination effects
- •
The Environmental Succession Prediction suppresses the background region and thereby enhances the foreground image region by means of articulating a binary image through masking procedure.
- •
Processing masked image by applyingProbability based Gradient Pattern (PGP)technique for offering utmost accuracy in tracking an object irrespective of alterations in illumination effects.
This paper is organized as follows: Section II describes the related works on visual object tracking and their limitation inferred. Section III discusses the proposed Laplacian Chime Pattern (LCP) for preprocessing the image and enhancing it and further segregation of foreground / background region through pattern formation and blob detection for tracking an object through ESP-PGP methodology. Section IV illustrates the performance analysis of proposed algorithm over the prevailing object tracking techniques. Finally, section V presents the conclusion.