Intelligent Surveillance System

NB: We do not resell your papers. Upon ordering, we do an original paper exclusively for you.

NB: All your data is kept safe from the public.

Introduction

To monitor transient objects in any given environment in real-time requires an intelligent surveillance system. The main purpose to have a surveillance system is to understand and predict the ramifications and interactions in any one given scene of objects under observation through automatic interpretations effectively done by the intelligent surveillance system. There are several steps involved in any intelligent surveillance system, they include, detecting and recognizing mobile objects, the ability to track them, analyzing their pattern managing data as well as the artificial intelligence Sergio (2006).

The creation, as well as deployment of automated or intelligent surveillance systems, is due to the fact that there is increased importance associated with closer scrutiny as far as the military, public as well as commercial or private purposes are concerned. As far as the public is concerned, the surveillance systems can assist in keeping a close eye on the public, the transport industry in particular, and record anything of interest for multitudes or individuals that involve the public. Any intelligent surveillance system involves several processes including social ethical studies, management, telecommunications as well as signal processing.

Intelligent surveillance system operates on the following aspects

Visual surveillance through computer vision algorithms or image as well as video processing.
Building a complete surveillance system based on a wide range of technologies capable of bringing together different types of vision algorithms.
System design, communication as well as distribution, and how all these can be incorporated into bigger systems so as to meet the ever-increasing demand for CCTVs for the near future Sergio (2006).

Intelligent Surveillance Systems

To be able to deploy a relatively efficient surveillance system, the technique is required though its not necessarily enough; therefore, a need to process signals as well as data from strategically positioned cameras is equally essential. Most of the surveillance systems are largely though not restricted to public transport or parking lot domains. There are several different types of intelligent surveillance systems, they include DETEC, Gotcha, and DETER among others Bohdan T (2007).

These surveillance systems largely rely on motion detectors with the ability to digitally record and store events that appear on the scene. All the data collected is relayed and stored in a central database available on demand for any workstation on the network. DETEC surveillance system operates on customized hardware with a provision of up to 12 cameras connected to a single workstation. (Bovik)

DETER, an acronym for Detection of Events for Threat Evaluation and Recognition is mainly meant for outdoor applications. Its main purpose is to detect and report unique or irregular movement patterns of both vehicles as well as pedestrians outdoors, especially in car parks. DETER system consists of two modules, a Computer vision module that handles the detection, recognition as well as tracking of objects across cameras and the threat assessment or alarm management module that consists of feature assembly or high-level semantic recognition, the offline training as well as the online threat classifier.

Since DETER is a cost-sensitive application, it uses fewer cameras with very efficient threat assessment abilities despite the absence of a feedback loop that could have made it even better. (Bovik)

Video Object Tracking Methods

The main objective for video object tracking is to get the objects motion information. The results thus obtained can be used to detect the position or the orientation of the object being tracked. The main advantage of video object tracking is the fact that it significantly reduces the search area of an object to within a single frame. There are several techniques used for video object tracking. Some of the most efficient include:

Inertia trackers are gadgets made up of gyroscopes and accelerators that are capable of measuring the rate of change of the translation velocity as well as the angular velocity of an object.
Mechanical trackers, they could be either serial or parallel, are based on a kinematic structure with sensorized joints interconnected with links.
Radio and microwave trackers, determine the range of an object, whereby the time of flight of the corresponding type of wave from a stationary transmitter to a moving receiver on the object of interest can be determined.
Ultrasonic trackers involve the use of transmission and sensing of ultrasonic waves. To identify the objects position, the time taken for a brief ultrasonic pulse to travel from a stationary receiver placed in the environment to a receiver attached to a moving object can be determined and thus track the object.
Magnetic trackers, they are gadgets that employ the magnetic field produced by a stationary transmitter to measure the real-time position of a receiver placed on the moving object. They are noncontact devices and thus very convenient (By Sergio A. Velastin).

MATLAB Video Image Processing

MATLAB is a type of software that interprets commands that shortens programming time through the elimination of the compilation process which otherwise is time-consuming. As a vectorized language, MATLAB is able to perform several operations on grouped numbers as vectors or matrices. Therefore MATLAB is more efficient, compact, and parallelizable compared to other software that involves more processes that take longer to process information. MATLAB provides a java-based graphical user interface (GUI) used as an interpreter for MATLAB programming language, 2D as well as 3D plotting functions. The best aspect of this software is the fact that its able to build customized GUIs. Bohdan T (2007)

MATLAB offers faster computations as it functions through the MEX utility. MATLAB provides application specific functions through its toolboxes available as add-ons that greatly extend its capabilities to a great extend whereby its able to carry out functions such image processing as well as signal processing. The toolbox for image processing represents an image as a matrix that provides image file input or output for JPEG, TIFF as well as other standard formats. It also provides FIR filter design in two dimensions, morphological operations, DCTS, colour space manipulation as well as conversions. The image processing tool box also offers canny edge detector. The strength of MATLABs image processing and developing signal abilities lies in its powerful functionality as well as its ease of use. (Bovik)

MATLAB Tools

Data visualization abilities are one of the aspects that make MATLAB tools superior compared to other tools. MATLAB tools offer users a wide range of tools for graphing as well as plotting. These are the most vital tools responsible for plotting time-based sequences such as the duration for response of dynamic systems to various inputs or initial conditions.

To determine the relative stability of closed-loop systems, MATLAB control toolbox offers powerful numerical tools specifically for this purpose. A case in point is its function margin capable of computing the phase and gain margin of a system in a unity feedback configuration. Apart from that, its able to give a graphical illustration in an annotated Bode diagram form. The loci of closed-loop poles can also be drawn automatically through the rlocus function given the transfer function of an open-loop system. Finding the open-loop gain that corresponds to a given location of closed-loop poles is fast and easy due to the interactive plot it generates (Bohdan 2006)

Image and Video Processing Techniques

There are several different types of algorithms for both image as well as video processing purposes. The processing techniques may include, resizing, contrast, cropping, brightness, improvement and sharpening as far as images are concerned. In the case of video algorithms, they include video segmentation, effect of camera motions on the motion vector field and how to get video segmentation through the exploitation of the effect of camera motions on the motion vector field Sergio (2006).

There are package tools for both image as well as video processing. The most common for image processing include:

Microsoft PhotoDraw which uses windows as its operating system to provide processed images offers great graphics as well as manipulation tools. Its also enables the layering of images, graphics as well as text. Its also quite user friendly especially for those who are not well versed with image editing.
Adobe Photoshop, it provides manipulation tools, great graphics, image processing and also allows text, graphics and image layers. It also capable of several graphic drawing as well as painting tools enhancing the users creativity and uses windows as it operating system.
Macromedi Freehand, Supports many bitmaps, vector formats and FHC (shockwave Freehand), provides text and web-graphics editing tools. Unlike others, it uses both windows as well as Unix as its operating system.

A few selected video processing tools include available commercially or even free include:

Adobe Premiere happens to be an excellent tool for professional digital video editing. It has a whole array of tools for editing. It supports Illustrator artwork, native Photoshop files and Photoshop layers. It uses windows as the operating system.
Video Wave 5.0, its an excellent Video editing tool with high quality multimedia bundles that include high quality multi bundles with clear and intuitive layout that can make an armature shot video look great. It operates on windows 2000/XP as its operating system.
Video studio 6.0, its a video editing software capable of trimming video, add soundtrack, creating compelling titles as well as drop in exciting effects with a very easy to learn interface. It uses Windows98/2000/ME/XP as operating system (Bovik 2005)

Algorithms That Detect and Track Objects

Based on multiple hypothesis tracking (MHT) and joint probability data association filter (JPDA), there are a variety of methods used for tracking multiple objects where algorithm plays a very vital role. To detect and track objects with straight line trajectories, a dynamic programming approach (DAP) had been proposed with a truncated sequential probability ratio test. But in the long run, this method proved to be not only costly, but also only dealt with straight line trajectories only limiting its functions. However, a better, more versatile and effective as well as more affordable method was developed that uses modified pipeline algorithm as well as filter bank approach. In the absence of any prior information about the objects dynamics, the filter bank approach is able to not only track manoeuvring objects, but non-manoeuvring objects as well with a lot of ease. Because of this, the multiple hypotheses tracking were outdone as it could only track objects with straight trajectory Sergio (2006).

The Filter bank approach detects and characterizes the dynamical behaviour of objects in a scene, in which wavelet transform for temporal filtering is used. To indicate whether there is temporal change or not in an object, temporal multistage decomposition facilitates the construction of intensity change maps. To authenticate the temporal applied. Then post processing is incorporated to make the detection scheme robust to clutter as well as noise. (Bovik 2005)

In a highly evolving background, its usually very difficult detecting small objects especially those with low contrast. To overcome the problem of small objects with low contrast detection, Gradient based detection of small objects algorithm is used which makes use of spatial gradient thresholding as well as region merging. To register motion, this algorithm employs image-differencing technique where the object motion or any displacement is captured in the difference image once the regions corresponding to any variation between frames take place. Through the application of Sobel operator which identifies the edges in the difference image, the spatial gradient of the difference image is then obtained. (Bovik 2005)

Methods to Detect Suspicious Objects on the Scene

The most vital task in any video surveillance applications is event description and recognition. Developing efficient recognition algorithms to deal with description and general event description is essential. There are two key methods for event recognition; they include stochastic inference and deterministic inference. Using hidden Markov models, recurrent Bayesian network, Stochastic Parsing among others is what refers to stochastic inference. Through this method, simple video events with fixed structures can be learned and recognized by use of stochastic models. (Bovik 2005).

On the other hand inference rules are applied as far as deterministic inference is concerned. Domains such as A1 or symbolic logic are some of the complex problematic areas where deterministic inference is applied. To describe actions such as scene states, events or anything else in the scene, a declarative model is used. This is achieved by mapping temporal relationship to a triple of PNF networks in order to get fast detection of actions as well as sub-actions that characterize actions as conditions of objects in the scene after which events in the video are recognized by uses of conventional algorithms. Surveillance video events can be recognized by modelling video events by Fuzzy Petri Nets based on FPN model of video events by defining VE_FPN models. (Bovik)

Fuzzy Color Histogram (FCH)

The other method proposed to detect suspicious objects on the scene is the Fuzzy colour histogram (FCH). The method is quite effective as it is only sensitive to noisy interference such as illumination changes and quantization errors unlike the conventional colour histogram (CCH) that needs vast computation on histogram comparison. FCH is further made use of in the application of image indexing and retrieval. Results have proven that FCH yields better retrieval results than the conventional colour histogram (CCH). This methodology is quite useful for image retrieval over large image databases. This makes it quite useful as far as intelligent surveillance systems are concerned. (Ma)

References

Bovik, Alan Conrad. Handbook of Image and Video Processing. Austin: Academic Press, 2005.

By Sergio A. Velastin, Paolo Remagnino, Institution of Electrical Engineers. Intelligent Distributed Video Surveillance Systems. Rome: IET, 2006.

By Bohdan T. Kulakowski, John F. Gardner, J. Lowen Shearer. Dynamic Modeling and Control of Engineering Systems. Pennsylvania: Cambridge University Press, 2007.

Ma, Ju Han Kai-Kuang. Fuzzy color histogram and its use in color image retrieval. Fuzzy color histogram and its use in color image retrieval (2002).

Click Here To Order Now!