Research Paper Example on Object Tracking in Real-Time and Field-Tracking Intelligence

Abstract-In this paper, there is the examination of the object surveillance through real-time tracking where there is the analysis of the various techniques which help in the analysis of the methods that are employed in the modern society for the delivery of the right goals. The paper uses a method to help in the tracking and recognizing an object in the surveillance area, much focus in this process is the pixel-approach which helps in arriving at the solution for the problem. The camera system is also used as sensor for the purposes of tracking the object used for the study in the surveillance area. Additionally, there is the use of edge detection in the analysis, and this is an image segmentation process which helps in having a better understanding on the real edges of real time video. There is also the use of background separation algorithm which provides a clear knowledge about the foreground and the background. There are various types of video preprocessing, and some of them are thresholding, histogram equalization, frame separation, binary operation, and edge detection of traffic video. Such are used for the purposes of tracking multiple objects and recognizing them. The use of the stepper motor also proves to be of great importance in that it helps in orienting the camera to any position for tracking and recognizing the object under surveillance. Finally, the use of contourlet transform is used to extract features and for the recognition of objects under surveillance area, pattern matching can also be used for recognizing different objects in a video.

Is your time best spent reading someone else’s essay? Get a 100% original essay FROM A CERTIFIED WRITER!

Order now

Keywords: Object tracking, features selection, surveillance and tracking, object recognition

INTRODUCTION

Tracking of objects in video sequences has been of great importance in various ways in the modern society and the same skill is employed in various fields and areas among them being surveillance, human-computer interaction, smart vehicles, interactive TV, and augmented reality applications. The process of tracking objects involves two main steps, and which are the detection and tracking. In the first step of detection of the object, there is the use of a common approach, and detection is used in the first frame, tracking then proceeds in the rest of the video. However, such an approach overlooks spatial information. There is a better approach available and it is the pursuance of a continuous integration of spatial and temporal information which combines the methods of detection and the tracking approach. The other method that can be used for tracking objects of taxonomy and it employs the use of three classes; Point tracking, Kernel tracking, and Silhouette tracking. Point tracking is where the objects that have been detected in consecutive frames are represented by points, Kernel tracking deals with the representation of objects by points such as a rectangular or elliptical shape. In the last method, Silhouette tracking, the objects are represented through the use of contour or the region which is inside a contour, in this method, there is a need to employ the use of a mechanism that can help in the detection of objects in each frame. The second class best fits rigid objects and employed for intensive real-time applications, the third class, on the other hand, best fits non-rigid and complex objects. The Kernel tracking methods employs the use of two models; the first one is used to gather information from the most recent observation and the second model is where the different views of the object can be learned offline and employed for the tracking processes.

PROBLEM IDENTIFICATION

There is a lack of novel methods for object tracking, and which meets the computational performance required for the interactive TV programs. Additionally, there is the challenge of tracking some objects as hot spots for user interactions among them being character selection or as regions for the real-time compositing. An example of the latter is either illuminating a face or attaching a text to a character. There is the difficulty of achieving real-time detection and tracking objects and there are two main reasons for the same. The first reason is because there is more than one object the is to be detected in each video frame and this compromises the performance of algorithms. Secondly, the high resolution videos that are used affects the processing of time because a higher resolution requires that a larger area be searched for every object. The project therefore seeks to provide solutions to the two main problems in video tracking.

RELATED WORK

The study of object tracking in real-time has attracted various forms of surveys which have proven to be of great importance to the modern society. Various authors have carried studies on the same topic with the aim of developing a robust system for real-time tracking. There have also been identified various methods of object tracking that are based on particle filters. In such methods, the target model of the particle filters is defined by the use of colour information of the objects that are being tracked. The processes of tracking objects are expensive to compute, and there have been developed methods that lead to cheaper computational costs among them being parametric models motion, optical flow, and matching blocks.

Moreover, there has been proposed a reliable algorithm which can detect objects in images in real time. The algorithm developed by Johnsen and Tews is fifteen times faster than the previously proposed techniques.

Another proposal by Roth is a technique of an algorithm which is based on a set of rotated Haar-like features which enriches the previous works on object tracking in real-time.

Moreover, there is a proposal by Ray, Dutta and Chakraborty which is a real-time facial feature detection on mobile devices based on integral images. The advantage of the proposed method of Johnsen and Tews is that the algorithm is based on features and not the pixels thereby leading to higher performance. The study by Roth is an extension of the study of the algorithm by Johnsen and Tews to the motion and domain. The only difference is that the study focusses more on the low resolution videos of human figures under difficult situations. Additionally, the frame rate is too slow.

THE PROPOSED METHOD

SEARCH AREA REDUCTION

In the proposed method by Johnsen and Tews, finding an object in an image requires a new search over the image to be started afresh. Such a search goes all over the image by moving a window with a varying size. The window to be used begins with the smallest possible size of an object in the image such as 25*25 or 35*40. Every time the window completes the process of searching the whole image, its size has to be increased by a factor l, then e new search is the started. The process ascertains that an object in the image is detected regardless of its size. The algorithm below is a summary of the work by Johnsen and Tew.

In this case, the whole image is tested to determine whether it contains the desired object. However, the changes in the consecutive frames occur only in the small regions, the search area can then be minimized through checking only the regions that have changed. Accomplishing this task requires the use of a non-statistical approach for the background segmentation by an adaptive mean. The advantages of a subtraction technique through adaptive mean are recursive thereby proving no need to maintain a buffer memory for the storage of a background model. Finally, the technique conforms to the changes in lighting and physics and has a low computational cost.

OBJECT REAL-TIME TRACKING TECHNIQUES

In the recent decades, the computers have shown improved power, functionality, ubiquity which hen coupled with progress in the internet discovery and use have transformed human lives (Bodor, Jackson, & Papanikolopoulos, 2003). The focus of scientists has been directed towards using the unique features of technology to facilitate overreaching changes in critical societal aspects such as security. In fact, pundits now look at the world as continuously undergoing saturation with unique innovations in the form of powerful computer chips that are continually becoming part of the human life (Johnsen & Tews, 2009). Cameras with high resolution are have become essential requirements for a successful system used for real-time object tracking. Multiple or single cameras can be fitted on moving platforms such as robots, motorbikes facing any direction; forward, rear or sideways (Fransen et al., 2009). The choice of the number of cameras to use in any case depends on the cost imperatives. Multiple single camera systems installed on moving platforms such as robots or automobiles and inclined at various angles and distance that capture every detail about the object of focus (Roth, 2010).

The automatic reconfiguration of an objects shape requires the system to overcome three initial problems (Kragic, Miller, & Allen, 2001). These challenges include; the movement of the robotic platforms on which the sensors and camera used to monitor an object are mounted, a wireless system of communication that provides a means through which the gathered information from the sensors is remitted to a stationary computer surveillance and; altering of the real-time projection of an object on the desired surface.

Object real-time tracking techniques currently form an integral part of the practicing security engineering. It has the potential of solving a whole range of security problems that would otherwise become intractable. Due to the difficulty in comprehending and predicting the possibility of a crime occurring, it is impossible to rely solely on human intelligence for tracking safety alerts (Kragic, Miller, & Allen, 2001). Therefore, technology progressively undergoes harnessing to aid information gathering and dissemination, which has been critical in dismantling criminal networks and dangerous weapons. Object and individual tracking applications have been integrated with robotics technology to facilitate safety risk detection (Ray, Dutta, & Chakraborty, 2017). Nonetheless, one first necessity for the operations of tracking technology is that it must know the elements within a scene and its changes over time before being able to interact efficiently with the environment.

The tracking of objects using artificial mechanism takes different forms including video monitoring, surveillance systems, and robotic platforms. These three fields have been researched proactively in the past (Ray, Dutta, & Chakraborty, 2017). In most cases, tracking devices are installed stationary at strategic points to gather diverse information that is critical in ensuring societal security (Zhao, Zhang, & Shibata, 2012). However, remarkable progress continues to be made towards different forms of mobile intelligence gathering technologies such as drone fitted with high sensitive sensors that detect and transmit real-time information from the field to computer systems located in safe places.

In stationary video tracking systems, various requirements must be met to ensure effectiveness. One important aspect of monitoring using immobile technologies is that the desired object must remain within the range of surveillance within the period of information collection (Sobh & International Conference on Systems, Computing Sciences and Software Engineering, 2008). Therefore, if the device or object goes outside the range, it becomes difficult to track. From a technological perspective, once an object extends beyond the surveillance range covered by the intelligence gathering equipment then it becomes intractable (Kragic, Miller, & Allen, 2...