METHOD AND SYSTEM FOR MONITORING AN AREA
This invention relates to a method and a system for monitoring an area, especially for the detection of gas and/or smoke. In recent times different system have become usual for detecting gas or smoke in factories or in ordinary homes as smoke detectors .
Gas detection is often done using optical systems in which the absorption spectre in the interesting range is measured, preferably by emitting a light beam with a known spectre, e.g. a laser beam, and measuring changes in the spectre to detect different types of gases. When using a laser the spectre is often scanned over a chosen range to provide absorption measurements at different wavelengths. The known optical systems have a number of disadvantages such as complicated and sensitive arrangements. Also, they will usually be aimed at the detection of certain gases, so that unexpected leaks or gas occurrences in factory environments will not necessarily bed detected. In addition, only the relatively limited volume through which the light beam passes is monitored.
Detection of smoke has traditionally been done by detecting particles, e.g. in ordinary fire detectors with optical or ion based detectors. These point detectors have the disadvantage that the smoke has to reach the position of the detector before it is detected. Thus a gas leak, fire or similar may have developed far before it is discovered.
None of these detectors may discover development of heat, which thus has to be done using suitable heat sensors. Thus it is an object of this invention to provide a method for passive monitoring of a relatively large area for measuring changes in a volume of air within the area . The patent application GB 2.269.506 describes the use of an image sensor, comprising a sensor matrix for surveying a chosen area, by comparing a number of subsequent images.
The comparing is performed by integrating the values in the image plane by adding the values from each sensor in the matrix. Changes between the subsequent images may then be detected as changes in the integrated values .
The system described in the GB application may detect changes over time but is not able to analyse and categorize the different types of changes in the image, such as heat, gas leaks and human activities, as only the quantity of the total change in the image is observed. Thus the practical value of the system described in the cited document is limited.
According to the present invention a system and a method is provided making it possible to detect and analyse the different types of changes that may occur in the image area. This is obtained using a method and a system according to the independent claims .
In addition to the obtained measuring of changes in the refractivity and diffusion in the abovementioned air volume an additional effect is obtained in that other changes, such as heat evolvement, in the monitored area may be detected.
The invention will be described more in detail below, with reference to the accompanying drawings, illustrating the principle of the present invention. Figure 1 illustrates a volume a gas and two types of detectors. Figure 2 illustrates the measuring principle according to the present invention. Figure 3 shows an example of a mapped measuring area . Figure 1 illustrates an air volume 1 containing a volume of gas 2. Neither the point detector 3 nor line detection, e.g. based on a light beam 4 positioned another place in the air volume 1, will discover the gas volume 2 before it has diffused or moved sufficiently to come in contact with the light beam or the point detector.
The present invention is based on the fact that different gases, as well as the same gas with different temperatures, has different refraction indexes. If the gas volume moves into an area with a different gas, this may be detected in that a point on the opposite side of the gas volume, relative to an observer, will move.
Figure 2 illustrates how a point B, relative to an observer in point A, will apparently move on a surface to a new position B' when a volume of gas 2 having a different refraction index than the surrounding medium, e.g. air is
introduced between the points A and B. Of illustration purposes the gas 2 is introduced from a gas container 5.
Figure 3 shows an example of a mapped measuring area with a number of chosen areas 6 to be analysed. As shown the chosen areas comprises well defined lines with high contrast, so that the measured changes are evident. In addition to line points, as illustrated in figure 2, or areas having well defined shapes and contrast, may be chosen. In addition to the movements of given points smoke and similar may give a diffusing effect, which will be detected as reductions in the contrast. Other changes that may be detected are, for example human activities within the area. The system according to the invention is based on the use of a detector matrix with a related imaging system for making an image of a chosen area . Within the chosen area a number of reference points, lines or areas with known positions and/or outlines are chosen having a number of characteristics making a reference image. A new picture of the area is made after a certain time, and is combined with the reference image. The resulting difference image provides a set of characteristic values showing the possible differences between the images.
Differences between the images may be classified in analysing the size and the degree of change.
A displacement of a point resulting from a gas leak will be small and may be detected as a weak and directional change without any change in the size of the point . - Human activities or similar events may be detected as larger changes in the picture, for example in that reference areas disappear from the image.
Diffusion because of smoke may be seen as a soft, primarily centralised change and a decrease in the contrast. A larger concentration of smoke may conceal the reference areas completely in the same way as human activities.
The analysis of the information in the images may be based on a number of different calculations, among them statistical methods of higher orders.
By calculating the fourth momentum over the medium of the images the kurtosis C4 is found with the expression:
C 4 [£
(*)]
4-3|-∑ [ D ( n ) ]
2
NneΩ
v in which Ω
x is the area of interest, n is the specific pixel, N is the number of pixels in the area and D(n) is the difference image given by the values of each pixel.
The kurtosis expression may be used to find how much the value of an area of pixels have changed between the two images constituting the difference image. Thus this expression may be used to separate between changes resulting from smoke or human activities, from changes resulting from gas.
As the odd numbered moments, e.g. the third, of the mean is equal to zero for symmetrical distributions this may be used to find a directional change or shift in the image areas. This may for example be used to distinguish between an occurrence of a diffusing medium and other events.
Perturbations in the image made by gas leaks or similar events give larger pixel variations than noise. Detection of phenomena like human activities will give even larger changes. Since the standard deviation indicate the spread in the measured data larger changes will give larger standard deviation. The standard deviation is given by:
in which Xi is the sampled values and x is the mean value.
Human activities are usually easy to see. Therefore a difference image will give large values for the standard deviation. This feature will thus make this type of events in the picture easily distinguishable from others. It will, however, be difficult to distinguish these events from observations of smoke if the standard deviation is used as the only indication, as smoke also will give relatively large changes in the image . Apparent changes in the images because of changes in the refractive index will give small changes in the picture, and thus be difficult to detect with the standard deviation alone.
The sum of the values of all the pixels may also indicate changes, as an unchanged image will result in a difference image with the value zero in each pixel, as described in the abovementioned patent application GB 2.269.506. Again the changes cause by smoke and human activities will be larger than changes due to gas leaks and similar events. This method may be used in addition to the method according to the invention.
This also applies to the method of finding the largest change detected at one pixel. This method is, however, sensitive to noise, and will not be able to distinguish between human activities and apparent movements because of gas .
Yet another detectable change in the difference image is the number of pixels changing their values over a chosen limit, the limit being chosen according to the expected noise in the system. This may also give relatively small values for apparent changes in positions because of gas or heat, but larger changes for smoke an human activities. As mentioned above local changes in the content of a medium, which results in changes in the refractive index, gives apparent changes in the position of elements in a picture. This leads to a changes in the position of an edge or a point which preferably has a clearly defined contrast, will change on the detector matrix. The gradient of the change in intensity along the x-axis may be expressed as:
Iχ ( x , y) = I ( x , y) -I(x+l,y)
and the altered intensity distribution of the apparent shift along the x-axis is a will be:
Iχr nβ x , y) = Iχr ief { x +a , y)
Since the gradient is calculated as described above it will also provide a value for the contrast between two adjacent pixels. The square of the normalized gradient will thus give a measure of the contrast along a chosen line or edge in the measuring area.
Smoke may, as mentioned above, completely or partially,
obscure a line or an edge, so that the measured value for the gradient becomes small . Thus the value for the square of the normalized gradient may be used to distinguish between images belong to the smoke or gas category. The measure will, however, not be able to distinguish between a reference image and an image being deformed by an apparent movement, as this will not alter the contrast.
The difference between the reference and distorted images may be measured by calculating the time derived It, given by
ιt l χ , y) = — ( K χ , y) -R ( χ , y) +i(χ,y+l) -R ( x , y+1 ) )
in which R is the reference, I is the present image and x and y are pixel coordinates. The value of It will be different from zero if corresponding pixels in the reference image and the present image have different values . As this measure gives the difference in the grey scale between two pictures it will usually not distinguish between images indicating the gas or smoke categories. A feature which may be used to distinguish the observations from these two categories and the reference may, however be obtained by combining It and Ix.
It and Ix may be combined in a number of different ways. Preferably, however, a method should be used providing large values when the images are different, and the edge or line also does not disappear in the present image. This combination may be performed using Ix τ It for each line. This provides a number for each line in the image area and the standard deviation may for example be given as a value for the region. By comparing these values with some of the values discussed above, e.g. the kurtosis for the measurements in the chosen areas, detections of gas/heat, smoke or human activities may be distinguished from each other, when referring to a reference image. In the specification above a number of different values are described providing a possibility to classify the different changes in an image relative to a reference image. These may of course be combined in a number of different
ways, all being within the scope of this invention.
Even if the invention here has primarily been discussed in relation to relatively limited area in a picture being used to detect changes, later developments in computing power will make it possible to use and analyse the complete image, without separating certain parts. The chosen area will then comprise the whole measured area. In these cases it will be interesting to find other features than the ones mentioned above, e.g. calculations of geometrical momentum.