FR2832799A1 - Optical gas detection system for monitoring industrial sites with widely varying ground temperatures, where temperature effects are eliminated by calculating flow difference ratios for two filters - Google Patents

Abstract

A measurement scale independent of temperature is determined in the calculating window from the ratio of differences in flow through the two filters. The gas concentration can be calculated from this scale by weighting the ratio using the mean flow in the calculating window or as a function of the difference between a preset temperature and the ground temperatures during the calculating window Optical gas detection system by observation at a distance of an area using a measurement filter with a transmission band comprising a specific absorption ray for the gas being sought and a reference filter with a transmission band corresponding to that of the measurement filter but not including the specific absorption ray for the gas. The process involves detecting the presence of the gas by determining the difference in flow through the gas (22) and coming from points at different temperatures in a calculation window, and of the ratio of the flow differences seen through the measurement filter and through the reference filter. The scale is equal to ( eta . phi p)/K, where eta is the ratio, phi the mean flow in the calculating window, p a parameter depending on the measuring and reference filters and K a normalization parameter equal to phi p for a preset temperature in the calculating window, e.g. 20 deg C. The values of the derivative with respect to the temperature of the spectral luminescence of a black body at a given mean temperature are calculated for difference ground temperatures to deduce the derivative with respect to the temperature of a ratio for the same mean temperature in the absence of gas. The values are recorded and the measured ratios weighted as a function of the difference between the ground temperatures and the preset mean temperature. For the detection of several gases, a filter assembly (F1, F2, F3) is used, with the transmission bands determined relative to each other as a function of the absorption rays of the gases (G1, G2, G3). Thus, a filter (F2) can be used as a reference filter for a gas (G1) and as a measuring filter for another gas (G2,G3), or vice versa, connecting the filters in pairs, each pair detecting one or more gases. An identification matrix is formed where the rows correspond to the pairs of filters and the columns to the gases to be detected. The matrix is formed by calibrating the various filters and is used to detect the presence and concentration of the gases. A series of filters with stepped transmission bands are used.

Description

<Desc / Clms Page number 1>

 METHOD FOR REMOTE OPTICAL GAS DETECTION.

 The invention relates to a method for the optical detection of gases, in particular polluting gases, and is particularly applicable to the monitoring of industrial sites such as factories, refineries, gas storage installations, etc.

 In European patent 0 544 962, a gas detection method has already been proposed based on the selective absorption of infrared radiation from the landscape by a gas present in the field of observation.

 This known method uses an infrared imager associated with a measurement filter which is sensitive to the presence of the desired gas (the transmission band of this filter comprises at least one specific absorption line for the gas) and a reference filter which is insensitive. the presence of gas (the transmission band of this filter is comparable to that of the measurement filter but does not include the specific absorption line (s) of the gas). Contrasts are determined in the image of the observed area seen through the measurement filter and in the image of this area seen through the reference filter, the contrasts being provided by the differences in flux coming from points at different temperatures. of the area observed and passing through the desired gas. The ratio of the contrasts in the measurement image (seen through the measurement filter) and in the reference image (seen through the reference filter) makes it possible to determine by calculation the quantity of gas present on the optical path between the bottom of the observed area and the observer.

 This ratio of contrasts in the measurement and reference images makes it possible, under certain conditions and as a first approximation, to eliminate the proper emission of the desired gas and to get rid of the values of the temperatures of the points of the area observed and of the deviations between these temperatures.

<Desc / Clms Page number 2>

 However, it has been observed that, when the range of background temperatures is relatively large, the measurements carried out do not make it possible to obtain exact values of concentration of the detected gas, and that the measurement error is moreover a function of the position in the spectrum of the absorption line of the desired gas.

 The present invention aims in particular to eliminate these drawbacks.

 It also aims to significantly lower the minimum concentration values from which gas detection is possible.

 It also aims to improve the aforementioned process, in particular to systematize and optimize the detection of several gases present in an area of observed space.

 To this end, it proposes an optical gas detection method by remote observation of a space area by means of a measurement filter, the transmission band of which comprises a specific absorption line of a desired gas and of '' a reference filter whose transmission band corresponds to that of the measurement filter but does not include a specific absorption line for this gas, this method consisting in detecting the presence of the gas by determining the differences in the flows passing through the gas and coming from points at different temperatures of a calculation window and from the ratio of the differences of the fluxes seen through the measurement filter and the differences of the fluxes seen through the reference filter, and being characterized in that it consists in determining from this report a measurement variable independent of the temperatures in the calculation window, and determining or calculating the gas concentration from this quantity, this this being determined by weighting the aforementioned ratio by the average flow in the calculation window or by

<Desc / Clms Page number 3>

 weighting of this ratio as a function of the difference between a predetermined temperature and the background temperatures detected in the calculation window.

l'invention, ladite grandeur de mesure est égale à (11. )/K, 11 étant le rapport précité, (D le flux moyen dans la fenêtre de calcul, p un paramètre dépendant des filtres de mesure et de référence, et K un paramètre de p normalisation égal à (Dp pour une température prédéterminée de fenêtre de calcul égale par exemple à 20 C. In a first embodiment of

the invention, said measurement quantity is equal to (11.) / K, 11 being the aforementioned ratio, (D the average flux in the calculation window, p a parameter depending on the measurement and reference filters, and K a parameter of p normalization equal to (Dp for a predetermined temperature of calculation window equal for example to 20 C.

 In an alternative embodiment of the invention, this method consists in determining beforehand, for different background temperatures, values of the derivative with respect to the temperature of the spectral luminance of a black body at a given average temperature. , to deduce therefrom the derivative with respect to the temperature of the aforementioned ratio for the same given average temperature and in the absence of gas, to record the values of this derivative in memory and to correct the values of the measured ratio according to the differences between the background temperatures and a predetermined average temperature.

 We can thus correct the effects of background temperatures on gas concentration measurements.

 According to another characteristic of the invention, this method also consists, for the detection of several gases, in using a set of filters whose transmission bands are determined with respect to each other as a function of the absorption lines of the gases to detecting so that a filter which can be used as a reference filter for detecting a gas can be used as a measurement filter for detecting another gas, or vice versa, associating the filters in pairs, each pair of

<Desc / Clms Page number 4>

 filters being intended for the detection of one or more gases, to make the measurements of the aforementioned flux differences for each filter, to calculate the values of the aforementioned ratio for each pair of filters and to deduce therefrom the presence and concentrations of the gases above in the observed area of space.

 This process makes it possible to systematize and optimize the process described in the aforementioned European patent, by carrying out a global measurement of the flows through each of the filters used and then determining by calculation which gases are present in the observed area. This method can then be applied to the detection of several gases whose specific absorption lines are distinct from each other or, on the contrary, partially overlap, which is a fairly common case in practice.

 In a preferred embodiment of the invention, this method consists in forming an identification matrix whose lines correspond to the pairs of filters and the columns to the gases to be detected, an element i, j of this matrix being formed by a value calibration obtained for a pair i of filters in the presence of a known quantity of a gas j to be detected, to form a measurement vector whose lines correspond to the pairs of filters and whose element i represents the measurements carried out on the pair i of filters (this element being for example constituted by an average of the values of the abovementioned ratio obtained at different points of the image of the observed area), to calculate coefficients of correlation of the columns of the identification matrix with the vector and to compare these correlation coefficients to determined thresholds to deduce the presence and concentrations of the aforementioned gases in the observed area.

 According to another aspect of the invention, this method also consists in using, to form one or

<Desc / Clms Page number 5>

 several pairs of measurement and reference filters, a series of filters having transmission bands, which are staggered over a band of observation wavelengths and which possibly overlap, to obtain an image of the area observed in each transmission band and summing the obtained images to reconstruct images seen through wideband measurement filters and images seen through wideband reference filters.

 This reconstruction of images on a wide band of wavelengths, from images obtained on narrower bands of wavelengths, is advantageous when it is desired to adapt the method according to the invention to instruments having several observation routes or when you want to detect the presence of a large number of gases in the same area.

 According to yet another characteristic of the invention, the reference filters are synthesis filters produced by stacking thin layers and having one or more narrow transmission blocking bands, corresponding to the specific absorption lines of the gas or gases to be detected .

 These synthesis filters can have extremely precise transmission blocking bands, which improves the sensitivity and the precision of the measurements.

 Alternatively, a continuously tunable filter can be used, the transmission band of which can be moved over a whole spectrum of wavelengths studied.

 According to yet another characteristic of the invention, this method consists in comparing the images obtained through the various filters with each other to detect a relative movement between the observer and the observed area and / or a movement in the observed area, the method then consisting in registering the images

<Desc / Clms Page number 6>

 relative to each other, before processing them, to eliminate the influence of the aforementioned relative movement and / or to eliminate from these images the parts corresponding to a movement in the observed area.

 This makes it possible, for example, to carry out measurements over large areas from an aerial observation (instrument on board an aircraft or a satellite) or carried out on board a vehicle or by scanning a scene by a camera from a fixed point.

 For this, it is possible to use an algorithm for detecting moving objects in the observed area, consisting in: - calculating the pixel-to-pixel difference of two measurement images obtained at time t-1 and at time t, respectively , - calculate a standard deviation of the difference image, which gives the overall spatial noise of the image, make a thresholding at (threshold. standard deviation) to obtain a binary image of suspect points, - label the points in regions of connected points, - make a thresholding on the number of points in these regions and retain as moving regions those containing more than (threshold2) points, - repeat these steps on the reference images obtained at t-1 and t.

 The application of a binary "AND" to the two maps of moving regions in the measurement images and the reference images gives a matrix of valid "shake" points which are applied to the mapping of the aforementioned difference report. flow at time t.

 According to a variant, the difference is used between an image at time t and a "cumulative backdrop" at time t. The calculation of the background is

<Desc / Clms Page number 7>

 performed using a recursive filter of the ARMA (Auto Regressive Moving Average) type. The background is updated at time t by injecting the measurement (or reference) image into it at time t-1, as well as the matrix of points moved at time t-1, to update the background only in places where there is no shake and get a real background with no moving object.

 The "temporal" memory used for the calculation of the backdrop can also be used for the calculation of the noise, the pixel to pixel difference of the measurement (or reference) image at time t and of the backdrop at l 'instant t being injected into another recursive filter of the same type as above, to obtain a cumulative noise. The noise accumulated at time t allows thresholding according to temporal noise criteria of the difference image (t)-background (t) in addition to thresholding according to spatial noise criteria, which reinforces the robustness of the algorithm.

 The invention will be better understood and other characteristics, details and advantages thereof will appear more clearly on reading the description which follows, given by way of example with reference to the appended drawings in which: - Figure 1 shows schematically means for implementing the invention; FIG. 2 is a graph representing the variations as a function of the wavelength of the derivative with respect to the temperature of the spectral luminance of a black body, for several predetermined average temperatures; - Figure 3 is a graph showing specific absorption lines of three gases to be detected and the transmission bands of three filters used for their detection;

<Desc / Clms Page number 8>

 - Figure 4 is a graph showing the wavelength scaling of a series of filters, usable as measurement filters and as reference filters.

 First, reference is made to FIG. 1, which shows schematically means for implementing the invention, comprising a camera or a thermal imager 10 associated with a system 12 for processing information by the via an analog-digital converter 14 and intended for the observation of a space area 16 which includes, for example, a background 18, buildings 20 and possibly at least one cloud 22 of one or more gases whose presence we want to detect and determine the concentration.

 The camera 10 used for the observation of the area 16 comprises an optical system 24, at least one filter 26 and an element or a set of photosensitive elements 28 on which the optical system 24 forms, through the filter 26, an image of the observed area 16.

 The camera 10 can be a television camera of the CCD type or the like for measurements in the visible spectrum or a thermography camera which is capable of directly providing measurements of the temperatures of the zones observed. By "thermal imager", we generally designate a device capable of providing an image of the temperatures of the area observed, that is to say an image whose colors or gray densities are a function of the temperatures of the points observed.

 As already indicated in the aforementioned European patent 0 544 962, it is necessary to use, for the detection of a given gas, two filters 26 which have broadly similar wavelength transmission bands, but one of which (that of the measurement filter) includes at least one specific absorption line for the gas to be detected

<Desc / Clms Page number 9>

 and the other (that of the reference filter) does not include this or these specific absorption lines.

The method described in the aforementioned patent consists in determining the contrasts in the reference image and in the measurement image, that is to say the differences in the heat fluxes coming from points at different temperatures from the observed and passing zone. through the cloud 22 of gas to be detected, the measurement image being that seen through the measurement filter and the reference image being that seen through the reference filter, the method then consisting in making the report of these contrasts or differences in flux in the two images to obtain a magnitude which is independent of the flux emitted by the cloud of gas 22, of the overall gain of the camera or of the thermal imager 10, of the temperatures of the points considered in the area observed, the difference in temperature of these points and the temperature of the gas to be detected.

22. This approximation is legitimate when the range of background temperatures is not very wide (a few degrees) and when the absorption line of the desired gas is at the wavelength of the maximum emission of the background. If this is not the case, especially in an industrial environment, it is necessary to take into account the background temperatures to precisely determine the concentration of the gas in the cloud.

22.

grandeur de mesure étant de la forme : \fi (Dp)/K où :. cD est le flux objet moyen dans la fenêtre de calcul, For this, it is possible, in a first embodiment of the invention, to use an invariant variable as a function of temperature, this

measurement quantity being of the form: \ fi (Dp) / K where:. cD is the average object flow in the calculation window,

<Desc / Clms Page number 10>

 . 11 is the aforementioned report of the contrasts or differences of flux in this window,. p is a parameter depending on the measurement filter and the reference filter, and which is calculable or measurable,. K is a normalization parameter, which is equal to (D for a predetermined temperature (of 200C for example).

 The integrated gas concentration in the line of sight by the relation is determined: ## C.dl = b (a - #) where a. l is the line of sight, a and b are parameters that can be measured by calibration (b expresses the sensitivity to a given gas and a depends on the filters chosen and the temperature of the scene observed).

températures allant de 25 à 35 C environ, est approximativement 7 fois plus faible que celle de 11. For example, it has been verified that the variation of the measurement variable T as a function of the average temperature of the calculation window, in a range of

temperatures ranging from about 25 to 35 C, is approximately 7 times lower than that of 11.

 It is also possible, in a variant of the invention, to proceed as indicated below.

The ratio of the contrasts or differences in flux in the measurement image and in the reference image can be written as follows:

<Desc / Clms Page number 11>

un. rdLÂ.l est la dérivée par rapport à la température de L dïj T la luminance spectrale d'un corps noir à la température T, 1c est la réponse spectrale relative de la caméra
10, Tn est le coefficient de transmission du nuage de gaz, T est la température moyenne de la scène observée AT est le contraste de température de la scène observée, 6Àg est la différence en longueur d'onde des bandes de transmission du filtre de mesure et du filtre de référence, 6Àref correspond à la bande de transmission en longueur d'onde du filtre de référence.

a. rdLÂ.l is the derivative with respect to the temperature of L dïj T the spectral luminance of a black body at temperature T, 1c is the relative spectral response of the camera
10, Tn is the transmission coefficient of the gas cloud, T is the average temperature of the observed scene AT is the temperature contrast of the observed scene, 6Ag is the difference in wavelength of the transmission bands of the measurement filter and the reference filter, 6Aref corresponds to the wavelength transmission band of the reference filter.

 Contrary to what can be assumed as a first approximation for certain transmission bands and when the temperature range of the observed scene is not very large, the derivative with respect to the temperature of the spectral luminance of a black body at a temperature given average is not constant but varies according to the wavelength on the one hand and the average temperature on the other hand, as we can see on the curves of figure 2 which represents the variations of this function over a range of wavelengths between 2 and 15 u. m approximately for three average temperatures of 320.300 and 2800K respectively.

 We see on these curves that the shape of this function changes with the average temperature and that its maximum moves towards the shorter wavelengths when the average temperature increases.

 The invention intends to take these variations into account, by proceeding as follows:

<Desc / Clms Page number 12>

"Oit C-VCtJ. UC''U'Jil Ct-LUJ-C CS Va-LC-UJthéoriques ilo du rapport il en l'absence de gaz pour différentes températures moyennes de fond de scène et pour différentes gammes de températures de fond de scène et on enregistre ces valeurs théoriques, - au moyen de la caméra 10, on relève les températures des points de fond de scène dans une fenêtre de calcul de l'image de la zone observée, - on détermine la température moyenne et la gamme de températures du fond de scène dans cette fenêtre, - on mesure le rapport des différences de flux précitées pour ce fond de scène, - et on corrige les valeurs du rapport précité à l'aide des valeurs théoriques enregistrées en mémoire.

"Oit C-VCtJ. UC''U'Jil Ct-LUJ-C CS Va-LC-UJ Theoretical ilo of the ratio it in the absence of gas for different average background temperatures and for different ranges of background temperatures scene and these theoretical values are recorded, - by means of the camera 10, the temperatures of the scene background points are recorded in a window for calculating the image of the area observed, - the average temperature and the range of temperatures of the background in this window, - the ratio of the aforementioned flux differences for this background is measured, - and the values of the aforementioned ratio are corrected using the theoretical values recorded in memory.

 The corrections of the values of the abovementioned ratio are "local": each measurement is made on a set of 16 to 25 pixels for example and it is corrected by taking into account the temperatures of the surrounding pixels.

 It is thus possible to correct measurement errors which would result from the presence in the backdrop of a very hot element such as for example a chimney in operation.

 We can evaluate as indicated below the gains resulting from the corrections according to the invention on the minimum concentration values from which gas detection is possible: - when the range of background temperatures is 1 K relative to the average temperature, the estimated gain is between 1.1 and 1.4, - when this range is 5 K, the evaluated gain is between 1.3 and 3.1, - when this range is 200K, the estimated gain is between 2.2 and 10, - when this range is 400K, the evaluated gain is between 3.4 and 18,

<Desc / Clms Page number 13>

 the minimum value of the gain being that obtained for a gas whose absorption line is at the maximum background emission, the maximum value being that obtained for a gas whose absorption line is at the limit of sensitivity of the camera or of the thermal imager (8 dd. m in IR III band), the correction making it possible to reduce the detection threshold by a factor of around 10.

 This assumes that the background temperatures are determined exactly.

 This determination is linked to the knowledge of the emissivity of the background. When the observed area is fixed, it is possible to know or evaluate the emissivities of the different parts of the background, which makes it possible to obtain an accuracy of a few tenths of a degree on the background temperatures.

 In the other cases, the non-knowledge of the emissivity induces an imprecision on the determination on the temperature. In infra-red band III (8-12) JLm), the majority of materials have emissivities between 0.8 and 1, i.e. a value of 0.9 0, 1 (with the exception of paints) and the precision on the determination of the background temperatures can then be 5.

 The gains resulting from taking the background temperatures into account on the minimum concentration values for detection are then as follows: - temperature range of 20 K = gain between 1.7 and 3.2, temperature range 40 K = gain between 2.6 and 5.8.

 Reference is now made to FIG. 3, which diagrammatically represents the variations, over a certain wavelength range, of the transmissions of three gases G1, G2 and G3 to be detected and which are

<Desc / Clms Page number 14>

 each characterized by a specific absorption line, the gas absorption lines G2 and G3 being very close to each other and partially overlapping. FIG. 3 also represents the transmission bands of three filters FI, F2 and F3 of the passhaut type and which, used in combination two by two, will make it possible to detect the presence or the absence of the three gases Gl, G2 and G3 in a observed area of space.

 As can be seen in FIG. 3, the transmission band of the filter FI covers the absorption lines of the three gases G1, G2 and G3, that of the filter F2 comprises the absorption line of the gas G3 and part of that of the gas G2, and the transmission band of the filter F3 does not include any of the absorption lines of the gases Gl, G2 and G3.

 The IF filter can therefore be used as a measurement filter and the F2 filter as a reference filter for detecting G1 and G2 gases.

 The F2 filter can be used as a measurement filter and the F3 filter as a reference filter for detecting G2 and G3 gases.

 Under these conditions, if a gas detection is carried out with the filter couple FI, F2 and no gas detection is carried out with the filter couple F2, F3, the gas present in the observed zone is Gl.

 If a gas detection is carried out with the two pairs of filters FI, F2 and F2, F3, the gas present in the observed zone is G2.

 If no gas detection is carried out with the filter couple FI, F2, and a gas detection is carried out with the filter couple F2, F3, the gas present in the observed zone is G3.

 It is understood that the group of filters FI, F2 and F3 whose transmission bands have been determined as a function of the specific absorption lines of the desired gases, makes it possible to detect the presence of one or more

<Desc / Clms Page number 15>

 several of these gases even if their specific absorption lines are very close or partially overlap.

We can generalize this process as follows:
A group of filters is determined for a group of gases to be detected, the transmission bands of which are defined with respect to the specific absorption lines of the gases, such that a filter which serves as a reference filter for detecting one or more gases can be used as a measurement filter for the detection of another or more other gases, or vice versa, these filters are combined in pairs so that each pair of filters is intended for the detection of one or more of several gases, the contrasts or differences in flux coming from the area observed through each of the aforementioned filters are measured, the values of the aforementioned contrast ratio are determined for each pair of filters, and the presences and concentrations are deduced from these values of different gases in the observed area of space.

This process can be implemented as follows:
A gas identification matrix is constructed, each column of which is associated with one of the gases to be detected and each row of which is associated with the aforementioned pair of filters, an element i, j of this matrix corresponding to the value of the ratio of contrasts in the measurement image and contrasts in the reference image for a known quantity of gas j, the pair of filters used being called Ci.

 The elements i, j of this identification matrix can be determined experimentally by calibration or theoretically by calculation.

<Desc / Clms Page number 16>

 A series of measurements consists in recording the responses of the various filters and, in the event of detection of one or more gases, in constructing a measurement vector whose element i is the value of the aforesaid ratio of the contrasts for the pair of filters This.

 Next, correlation coefficients between the different columns of the identification matrix and the measurement vector are calculated. Each correlation coefficient is determined using a known method (for example by calculating the ratio between, in the numerator, the covariance of the column of the identification matrix and the measurement vector and, in the denominator, the product of the variance of the column of the identification matrix and the variance of the measurement vector) and qualifies the correlation which exists between column j of the identification matrix (this column corresponding to one of the gases sought) and the vector of measurement which was constructed from measurements made on the different pairs of filters.

 For example, each element of this vector can correspond to an average of the values of the aforementioned ratio at different points of the image of the area observed.

 When one of these correlation coefficients exceeds a predetermined value threshold, the gas corresponding to this correlation coefficient is declared identified.

 It is of course possible to define several threshold levels which correspond to several levels of reliability in the identification of the gases sought.

 The gas concentrations in the observed area can be determined by resolution of the system constituted by the n measurements obtained from the n pairs of filters and by the concentrations of the p gases (the identification matrix comprising n rows and p columns).

<Desc / Clms Page number 17>

 This system can be solved as soon as the number n of measurements is greater than or equal to the number p of gases to be detected.

 The measurement and reference filters which are used for detecting gases in the method according to the invention are typically filters having a relatively wide transmission band compared to the absorption lines of the gases to be detected. These filters can be replaced by filters having narrower transmission bands which are staggered with respect to each other in wavelength as shown diagrammatically in FIG. 4, or by a tunable continuous filter which comprises a transmission band centered on a wavelength which can be varied continuously over the spectrum studied, for example from 8 to 12 p. m. This tunable filter can be a standard Fabry-Perot filter fitted with piezoelectric shims.

 In FIG. 4, each reference 1,2, 3,4, 5, .... denotes the transmission band of a different filter mounted for example on a different channel of an observation device. We see that these transmission bands are close to each other, that they can partially overlap (in the case of bands 1 and 2.2 and 3, and 4 and 5) or be separated from each other (in the case of bands 3 and 4), these bands can also be uniformly staggered over a range of wavelengths.

 The different filters provided in the different channels of the observation device make it possible to reconstruct measurement images and reference images with a wide band, by summing up several channels of the observation device.

 For example, if the gas to be detected has a specific absorption line included in the transmission band n 2, a measurement image can be obtained by summing the channels corresponding to the transmission bands.

<Desc / Clms Page number 18>

 transmission 1, 2, 3,4 and 5 and a reference image is obtained by summing the channels corresponding to the transmission bands 1,3, 4,5.

 When the transmission bands of the filters are contiguous, their combination makes it possible to reconstitute a wide transmission band. If they are discontinuous, summing them overcomes the presence of annoying gases by eliminating the wavelengths where these gases are absorbent.

 The filtering mode represented in FIG. 4 is advantageous when an apparatus with several observation channels is used, or also when a large number of gases is to be detected. An imaging spectrometer of low spectral resolution is thus produced while retaining a high spatial resolution.

-echnique être réalisés avec une grande précision par une technique d'empilement de couches minces, ce qui permet la synthèse de filtres adaptés aux gaz à détecter, avec une amélioration importante de la sensibilité de la détection. Filters with one or more transmission blocking bands, such as the one between transmission bands 3 and 4 in Figure 4, can

-technique be produced with great precision by a technique of stacking thin layers, which allows the synthesis of filters adapted to the gases to be detected, with a significant improvement in the sensitivity of detection.

 It is also possible in this way to produce synthesis filters having several transmission blocking bands, for the detection of a gas having several absorption lines. These filters are particularly suitable for the detection of gases having a spectrum with several absorption lines included in the sensitivity band of the camera or of the thermal imager, since the use of a synthesis filter faithfully reproducing the spectrum of the gas on the reference track allows you to work on the entire sensitivity band of the camera and to reduce by a factor of about 3 the minimum concentration value of the gas to

<Desc / Clms Page number 19>

 from which a detection is possible, compared to what would allow the use of a conventional reference filter whose transmission band in wavelength would necessarily be very narrow.

Claims (12)

1) Method for the optical detection of gas by remote observation of an area of space by means of a measurement filter, the transmission band of which comprises a line for specific absorption of a desired gas and of a filter reference whose transmission band corresponds to that of the measurement filter but does not include a specific absorption line for this gas, the process

 consisting in de + l consisting in detecting the presence of the gas by determining the differences in the flows passing through the gas (22) and coming from points at different temperatures in a calculation window, and in the ratio of the differences in the flows seen through the filter of measurement and of the differences in the fluxes seen through the reference filter, characterized in that it consists in determining from this ratio a measurement variable independent of the temperatures in the calculation window and in determining or calculating the concentration of the gas from this quantity, the latter being determined by weighting the aforementioned ratio by the average flux in the calculation window or by weighting this ratio as a function of the difference between a predetermined temperature and the background temperatures detected in the calculation window.  2) Method according to claim 1, characterized in that said measurement variable is equal

 at (tri.) / K, it being the aforementioned ratio, V the average flux in the calculation window, p a parameter depending on the measurement and reference filters, and K a parameter of

 p normalization equal to OP for a predetermined calculation window temperature, for example equal to 20 C.  3) Method according to claim 1, characterized in that it consists in determining, for different background temperatures, values of the derivative with respect to the temperature of the luminance <Desc / Clms Page number 21>  spectral of a black body at a given average temperature, to deduce the derivative with respect to the temperature of the aforementioned ratio for the same average temperature in the absence of gas, to record the values of this derivative in memory and to weight the measured values of the abovementioned ratio as a function of the differences between the background temperatures and a predetermined average temperature.  4) Method according to one of claims 1 to 3, characterized in that, for the detection of several gases, it consists in using a set of filters (FI, F2, F3) whose transmission bands are determined by compared to the others as a function of the absorption lines of the gases (Gl, G2, G3) to be detected so that a filter (F2) which can be used as a reference filter for detecting a gas (gel) can be used as measurement filter for the detection of another gas (G2, G3), or vice versa, to associate the filters in pairs, each pair of filters being intended for the detection of one or more gases, to measure the differences of the aforementioned fluxes for each filter, calculating the values of the aforementioned ratio for each pair of filters and deducing therefrom the presence and the concentrations of the aforementioned gases in the area of space observed.  5) Method according to claim 4, characterized in that it consists in determining beforehand calibration values of the aforementioned ratio for each pair of filters in the presence of a known quantity of each aforementioned gas, in calculating correlation coefficients between these calibration values and the values calculated from the aforementioned measurements, and to deduce therefrom the presence of a gas in the area observed when a correlation coefficient is greater than a predetermined threshold. <Desc / Clms Page number 22>  6) Method according to claim 5, characterized in that it consists in forming an identification matrix whose lines correspond to the pairs of filters and the columns to the gases to be detected, an element i, j of this matrix being formed by the calibration value obtained for a pair i of filters in the presence of a known quantity of a gas j to be detected, to form a measurement vector whose lines correspond to the pairs of filters and whose element i represents the measurements carried out on the pair i of filters (for example an average of the values of the abovementioned ratio for different points of the image of the observed area), to calculate the correlation coefficients of the columns of the identification matrix with the measurement vector and to compare these coefficients at thresholds determined to deduce the presence and concentrations of the aforementioned gases in the observed area.  7) Method according to one of the preceding claims, characterized in that it consists in using, to form one or more pairs of measurement and reference filters, a series of filters having transmission bands (1, 2,3 , 4,5) which are staggered on a band of observation wavelengths and which overlap if necessary, to obtain an image of the zone observed in each transmission band and to sum these images to reconstruct images seen through wideband measurement filters and images viewed through wideband reference filters.  8) Method according to claim 7, characterized in that the images in said narrow bands (1,2, 3,4, 5) are obtained simultaneously on parallel channels of an observation device.  9) Method according to one of the preceding claims, characterized in that it consists in comparing with each other the images obtained through the different filters to detect a relative movement between the <Desc / Clms Page number 23>  observation means (10) and the observed area (16) and / or a movement in the observed area (16), then to readjust the images in relation to each other, before processing them, to eliminate the influence of the relative movement mentioned above, and / or to eliminate from these images the parts corresponding to a movement in the observed area (16).  10) Method according to claim 9, characterized in that the movements are detected by difference of images at times t and (t-1), calculation of a standard deviation, comparison to a multiple threshold of the standard deviation, comparison of the number of points retained at a threshold to obtain a shake map, and application of a binary "AND" to shake maps at the time t of the measurement images and of the reference images.  11) Method according to claim 10, characterized in that a difference is calculated at time t between an image and an accumulated background scene updated by means of a recursive filter of the ARMA type from the image at time t-1 and a map of points moved at time t-1.  12) Method according to one of the preceding claims, characterized in that the reference filters are synthesis filters produced by stacking thin layers and having one or more narrow transmission blocking bands, corresponding to the absorption lines of the or gases to be detected.