[Batpuff]

here you go tribe



Satellites [indent]Satellites literally provide their owners an eye in the sky and when used for surveillance purposes are more commonly referred to as spy satellites. One major class of spy satellite are Imagery Intelligence Satellites (IMINT).



Imagery Intelligence Satellites orbit at several hundred kilometres and use film and electronic cameras or radars to produce high resolution images of objects on the ground at ranges of up to one thousand kilometres.

A certain type of IMINT satellite known as 'Keyhole Class' satellite performs area surveillance. This type of satellite returns images to Earth via an electronic link. The most advanced of these satellites have a resolution of around 10 - 15 cm which means they can distinguish an object that small, but no smaller. As an example, this type of satellite is not capable of reading license plates of vehicles but they can tell if a vehicle has one, and they cannot provide real-time images however other 'assets' such as can.

A space based imaging radar can see through clouds and can potentially provide images with a resolution that approaches that of photographic reconnaissance satellites. A project to develop such a satellite by NASA is code named Lacrosse, first launched in 1988.

[Image: http://www.iis.ee.ic.ac.uk/~frank/surp99...crosse.jpg]

National Reconnaissance Office (NRO) de-classified photo of LacrosseLacrosse's main attribute, like most spy satellites, is its image sensor. Lacrosse beams microwave energy to the ground and reads the weak return signals reflected into space. This allows the satellite to 'see' objects on Earth that would otherwise be obscured by cloud cover and darkness. In order to send out these signals Lacrosse has very high power needs. It meets these needs with very large solar panels, larger than would be found on most satellites its size. Lacrosse uses a rectangular antenna, 48 feet long and 12 feet wide, that is very different from the standard mechanical antenna. This antenna is covered by rows and columns of small transmitting and receiving elements that help Lacrosse pick up the faint return signals bouncing back from the Earth.

[/indent]

[Image: http://www.iis.ee.ic.ac.uk/~frank/surp99...7/back.gif]


CCTV [indent]

Though modern satellites are capable of very high resolutions, Closed Circuit Television (CCTV) is a much cheaper solution to monitor streets, and is also used to monitor activity in buildings such as shops. In recent years, the use of Closed Circuit Television (CCTV) in the UK has grown to unprecedented levels. Between 200 and 300 million pounds per year is now spent on a surveillance industry involving an estimated 300,000 cameras. These camera systems use sophisticated military technology utilising technologies such as infra red night vision, automatic tracking, remote control, audio channels, and a zoom so powerful that it can identify facial characteristics in full colour at two hundred yards.

[/indent]

[Image: http://www.iis.ee.ic.ac.uk/~frank/surp99...7/back.gif]

Pattern Recognition [indent]One technique used in both Satellites and CCTV is Pattern Recognition. Pattern Recognition is the research area that studies the operation and design of systems that recognise patterns in data. Because pattern recognition encloses many sub-deciplines like discriminant analysis, feature extraction, cluster analysis (together sometimes called statistical pattern recognition) and parsing (sometimes called syntactical pattern recognition) this do***ent will focus on statistical pattern recognition as it is commonly used in spy satellites and CCTV systems with computerised facial recognition (CFR).



An image serves as input to various segmentation algorithms that partition the image into components. These components are further identified as possible candidates that represent objects of interest. The process is listed below:

• The system extracts a set of measurements from each of the component sub-images. For example, it might compute the centroid of the largest and shortest axes. These measurements are referred to as the features of an object in the context of pattern recognition. The result of the processing is an n-dimensional vector of features (i.e. each detected sub-image is characterised by one of these feature vectors). The number of features and what they represent varies with each problem.
  
• The feature vector is input into the pattern-recognition system. The system’s goal is to determine whether the feature vector belongs to one of several classifications that the system knows about.
  

Now the pattern recognition challenge consists of partitioning a vector space F (feature vectors) into various object types which can be identified. Two main pattern recognition approaches address this issue.

• Cluster approach. Here you have a set of feature vectors that characterise some objects, and your goal is to find out if some partition of the space F groups these objects into any meaningful sets.
  
• Classification approach. You are given a partition of the space F into a set of classes and a new vector. This approach determines to which of the given classes a new feature vector belongs.
  

CFR - Pattern Recognition Applied to CCTV

By incorporating CFR into CCTV systems, it is possible to gather intelligence covertly using hidden cameras while being able to identify specific people of interest. The main advantage of using CFR as opposed to other identification technologies is that it is passive and non-intrusive. It is far easier to obtain an image of a suspect than using other biometric technologies such as fingerprints, retinal scans or hand geometry systems.

Previous attempts at incorporating real-time CFR required powerful and expensive computers, which were often slow and produced inaccurate results. A person’s new hairstyle or spectacles could confuse and defeat many systems.

Identifying an individual by analysing a face is a complex process which usually requires sophisticated artificial intelligence and machine learning techniques. Artificial intelligence is needed to simulate human interpretation of faces. People do change over time. Facial hair, glasses and the position of the head can affect the way CFR can match one face with another. Machine learning is important to adapt to these changes and accurately compare new samples with previously recorded templates. The illustration below shows how CFR works in conjunction with a CCTV system. [indent]Process of facial recognition

In CFR, computers perform three distinct but related tasks:

[/indent][/indent][Image: http://www.iis.ee.ic.ac.uk/~frank/surp99...97/cfr.gif][indent][indent]Capture
The face must be located within the image. This can be simple (e.g., finding two circles that are assumed to be eyes), or it can consist of complex mini-recognition algorithms that divide the entire image into smaller sub-images and attempt to recognise a face in each subimage. Once a face has been found a number of points on the face is mapped out. For example, the position of the eyes, mouth and nostrils may be plotted so that a unique template is built. Alternatively a three – dimensional map of the face may be created from the captured image.[/indent][/indent][indent]

[Image: http://www.iis.ee.ic.ac.uk/~frank/surp99.../faces.jpg]

A three dimensional facial image using CFR[indent]Extraction
The facial image is converted into a pattern and then a unique mathematical code. This is stored as a template in a database.[/indent][/indent][indent][indent]Comparison
For recognition the target template is compared to all templates in the database for matches. Verification is considered a much simpler task than recognition, because only a single comparison is necessary.

[Image: http://www.iis.ee.ic.ac.uk/~frank/surp99...lcompa.GIF] Comparison of facial images

[/indent][/indent][indent][indent]Limitations of CFR

The complex nature of facial recognition still poses many difficulties even now with the advent of advanced Neural Networks and DSP's. The precise position of the user’s face and the surrounding lighting conditions may affect the systems performance. Human beings are inconsistent and physical characteristics can change slightly over time, This is why modern CCTV systems using CFR must allow for these subtle changes, so a threshold is set. This can take the form of an accuracy score. Here, comparison between the template and new sample must exceed the systems threshold before a match is recorded. The fact that CFR can be used in a wide variety of applications and not just in the context of surveillance will mean that there is a great incentive for scientists and engineers to overcome these difficulties. It will be only a matter of time before we see CFR being used in many aspects of our lives.

[/indent]


[/indent]