System of detection

A system of detection is a system making it possible the user automatically to observe an event by the means of an electronic device or mechanics. One can distinguish several types of systems of detection, which fall under many fields. First of all personal safety (System of detection of intrusion), or data processing (System of detection of intrusion of network). Then, because of growing computerization, the men sought to identify objects in an automatic way so, of contôler their territory (Radar) or to index forms (ex: system of detection of road panels ).

System of detection of the forms

The step is simple it has as a principle of comparing a target with a model present in memory what utilizes two stages:

Modeling of the form to be known.
Decision making on the membership of this form to a known Class.

synoptic Diagram

stage of acquisition of the data;
stage of reduction in the quantity of information to be handled which is very important afn to preserve only the useful part. It is first of all necessary to extract information in the form of characteristics (ex: vectors representatives);
stage of decision where one puts in correspondence, using rules suitably chosen, the preceding observations and classes;
stage of performance evaluation of classification.

It should be noted that the stage of extraction of information is paramount. Indeed information especially in the case of images, contains an impressive quantity of often redundant information for the operations of comparison. The following elements are thus sought:

to thus obtain a weak dimensionality of the characteristics of the image a reduced number of parameters;
to keep only relevant information allowing to distinguish two objects in two different classes but being able to be close;
to ensure a certain geometrical coherence of this space, i.e. a minimal distance is equivalent to a maximum resemblance.

System of detection of the road panels

the principle is always the same one starting from an image or from a series of image it will be necessary that the system analyzes information in three demensions in an automatic way and makes decisions in the case of an artificial intelligence. That will make it possible to make sure of the conformity of the road panels set up, to manage their inventories and to be more effective with regard to the replacement and the maintenance of the road panels. Thus one first of all will estimate their forms and their colors, thanks to a numeric camera and a image processing, in order to determine their positions geographical compared to this one. For that one can base oneself on two methods

Image by segmentations analyzes

As one can notice it, the panels have all the same forms, it acts of triangles, rounds or of roughly of the same rectangles cuts. Their main feature is that it all consist of homogeneous area. Their detection will be all the more easy as one places oneself in space RVB. For that we will use two entities of segmentation:

a Segmentation based on the discontinuity of brightness within the image in order to determine the existence of any panel by analysis of contour.

second is based on a segmentation of texture and color and thus makes it possible to determine the nature of the panel.

; Segmentation by detection of contours

When an abrupt variation of Luminance of the image is detected this one been the subject of a circumference characterizing a form or an object. In the literature one recorded four types of the shapes of panels (triangle equilateral, round, square, right-angled). That makes it possible to characterize a discontinuity by the study of variation of brightness. Indeed the uneven one of brightness of the image is generally represented by a more or less brutal discontinuity. One thus will be able to treat the image received in order to detect the road panels there.

; Segmentation of the colors

one will carry out an analysis by the color of the panels and according to their frequential answers, one will be able to determine if the object which one detected beforehand is well a panel. This method is very advantageous owing to the fact that it can occur in real-time. Moreover the analysis by segmentation allows to classify by advance a certain category of panel. Despite everything, it can be held only during the day because the method of segmentation of the panels is based on the phenomenon of Absorption and reflection of the white light (coming from the sun).

Use of a correlator with crystals liquid ferroelectric

The recent development of the space modulators of ferroelectric light to Liquid crystals, and the installation of algorithms of filtering in optical correlation, developed in the field of optics the capacity to recognize geometrical forms in real-time.

Principles of operation

It is a solution more global than that of the method by segmentation, which seeks to compare an image with all the relocated versions of another image. It consisted a long time of the search for a peak of Autocorrélation (by method of filtering) but was the subject of new significant developments these last years.

Features

The correlator with liquid crystals is visible at the laboratory of optics of the 3Ecole Nationale Sup3erieure of Telecommunications of Brittany. L makes approximately 45 cm X 30 cm X 20 cm.

It functions with an external laser source.

There is another the more compact and used for other comparable applications of dimension: 30 cm X 15 cm X 8cm.

The major advantage of the method of correlation is in the technique which makes it possible to reduce information considerably by removing all that is not necessary.

Cost The cost of a liquid correlator with crystals rises between 10.000 and 15.000€ that depends mainly on the components of posting.

Bonds

theses on the image processings in order to extract from the characteristics essential with the localization and the recognition of the road panels.

Study and ferroelectric realization of a correlator with liquid crystals. Application to the detection of road panels.

en.wikipedia.org/wiki/Network_intrusion-detection_system

thesis of Gilles Keryer: Study of optical correlators with mono or multichannel united correlation: application to the pattern recognition

Random links: