Method for determining the position of a person in a room, thermal imaging device and home detector

Description

Domaine technique

[0001] The present disclosure relates to a method for obtaining the position of a person in a space of a building. The present disclosure also relates to a thermal imager and a home automation detector configured to implement the method for obtaining the position of the person. Technique antérieure

[0002] In building management, it is known to use home automation detectors to detect the presence of people in one or more spaces, whether indoor or outdoor, of the building. The space is, for example, a room or an area of a room.

[0003] People detection allows you to adjust energy consumption in the corresponding space, for example for lighting, heating, ventilation or air conditioning of this space. People detection also allows you to implement video surveillance functions in the space. It also allows you to understand the actual use of spaces in a building, for example the actual utilization rate of meeting rooms, the number of free or occupied workstations in shared offices, etc. People detection also allows you to determine where it is most appropriate to install fire protection devices in the building, among other things.

[0004] Person detection is achieved by mounting a thermal imager in the space concerned, capable of capturing the infrared radiation emitted by bodies present in the space and creating thermal images of the space from this captured radiation. A thermal image can be represented either by grayscale or by areas of different colors depending on the amount of infrared energy emitted by each body in the space. This amount varies depending on the temperature of each body. Also, it is possible to distinguish on the thermal image people, whose temperature is in a temperature range of approximately 20°C to 40°C depending on the clothing of the people, from other bodies present in the space, and whose temperature deviates from the temperature range of the people.

[0005] In some cases, the thermal imager uses a wide-angle lens, for example, with a field angle greater than or equal to 120°. Such a lens therefore has a wide field of view, which allows capturing a larger area of space than a conventional lens. However, visual distortions appear in the resulting image, which in particular has a convex appearance instead of a flat one. Also, these known thermal imagers do not allow to know precisely the position of people in the monitored space. Optimal management of the building thus becomes difficult.

[0006] Mandel C. et al: "People Tracking in Ambient Assisted Living Environments Using Low-Cost Thermal Image Cameras". In: Chang, C., Chiari, L., Cao, Y., Jin, H., Mokhtari, M., Aloulou, H. (eds) Inclusive Smart Cities and Digital Health. ICOST 2016. Lecture Notes in Computer Science, vol 9677. Springer, Cham. DOI:10.1007 / 978-3-319-39601-9_2, describes the use of a thermal imager for people tracking in assisted living environments. Background subtraction and segmentation are applied to low-resolution thermal images. Then, people tracking is performed using Monte Carlo particle filtering.

[0007] Swamidoss IN et al: "Systematic approach for thermal imaging camera calibration for machine vision applications", Optik, , vol. 247, September 29, 2021, ISSN: 0030-4026, DOI: 10.1016 / J.IJLEO.2021.168039, describes a technique for calibrating a thermal imager based on the design of a calibration card intended to calibrate mid- and long-wavelength infrared (MWIR and LWIR) thermal cameras. Résumé

[0008] This disclosure improves the situation.

[0009] For this purpose, a method is proposed for obtaining the position of a person in a space of a building by a thermal imager located high up in said space and comprising a wide-angle lens, the method comprising the steps of: a) acquisition by the thermal imager, through its wide-angle lens, of a distorted thermal image of a scene in space comprising the person, the thermal image being optically distorted by the wide-angle lens, and the distorted thermal image comprising a hot zone; b) extraction of a region of interest corresponding to the person from the distorted thermal image using a hot zone detection algorithm; c) determination of the position of the region of interest in the distorted thermal image by identifying the polar coordinates distance r and angle θ of the center of the region of interest in a polar reference frame whose origin is located at the center of the distorted thermal image; d) obtaining the actual position of the person in space by applying a wide-angle correction function to the distance r to compensate for the distortion of the thermal image due to the wide-angle lens.

[0010] By applying the wide-angle correction function to the distance r, it is possible to obtain the actual position of each person in the scene of the monitored space of the building. This makes building management, for example in terms of real knowledge of the use of the building, the uses and occupancy rate of meeting rooms, energy consumption, video surveillance or the installation of fire protection devices, simpler and more accurate. Access to the actual position of people also allows for more precise monitoring of the occupancy of workstations in shared spaces. It also helps avoid overcounting in the case of adjacent thermal imagers with field overlap. Finally, it allows for more precise monitoring of queue management.

[0011] Furthermore, applying the correction function only to the distance r from the center of the region of interest representative of the person's position, and not to the entire distorted thermal image, makes it possible to obtain the position of a person with low computational time and load. Thus, the method according to the present disclosure can be implemented even by a low-power thermal imager.

[0012] The features set out in the following paragraphs may, optionally, be implemented, independently of each other or in combination with each other: the wide-angle correction function applied in step d) comprises a normalization factor Fn of the distance coordinate r from the center of the region of interest and a metric factor Fm taking into account a field angle of the wide-angle lens and an installation height of the thermal imager in the space of the building relative to a floor of said space, the wide-angle correction function comprising in particular the product of the normalization factor Fn and the metric factor Fm; the normalization factor is defined as: Fn = r β 1 − α r 2 where β is a scale coefficient and α is a distortion coefficient, the scale coefficient β and the distortion coefficient α having predetermined values; the metric factor is defined as: Fm = h − h 0 tan γ 2 where h is the height of the thermal imager relative to the ground of the space, h 0 is a value representative of the height of the center of the region of interest relative to the ground and γ is the field angle of the wide-angle lens; the wide-angle correction function applied in step d) further comprises a preferably linear angular correction factor Fa: the wide-angle correction function comprises the product of the normalization factor Fn times the metric factor Fm times the angular correction factor Fa; a value of the angular correction factor Fa is a function of a value of the angle coordinate θ, the value of the angular correction factor Fa following a sawtooth-shaped distribution of which a maximum value is associated with the values 45°, 135°, 225° and 315° of the angle coordinate θ.

[0013] According to another aspect, there is provided a thermal imager comprising a wide angle lens, the thermal imager being configured to implement the method described above.

[0014] The features set out in the following paragraphs may, optionally, be implemented, independently of each other or in combination with each other: the lens is a hypergonal lens; the imager includes a long-wave infrared imaging sensor for acquiring thermal images.

[0015] According to another aspect, there is provided a home automation detector comprising a thermal imager as described above. Brève description des dessins

[0016] Other features, details and advantages will become apparent upon reading the detailed description below, and upon analyzing the attached drawings, in which: Fig. 1 [ Fig. 1 ] shows a flowchart of a method for obtaining the position of a person in a space of a building according to an exemplary embodiment of the invention. Fig. 2 [ Fig. 2 ] shows a schematic view of an exemplary embodiment of a thermal imager installed in the space of a building for the implementation of the method of the figure 1 . Fig. 3 [ Fig. 3 ] shows an example of a thermal image obtained from the thermal imager of the figure 2 when implementing a step of the process of the figure 1 . Fig. 4A [ Fig. 4A ] shows a typical distortion of a checkerboard-type image obtained by an imager equipped with a wide-angle lens. Fig. 4B [ Fig. 4B ] shows a corrected distribution of the checkerboard-like image of the figure 4A . Fig. 5 [ Fig. 5 ] shows the thermal image of the figure 3 after application of a subsequent step of the process of the figure 1 . Fig. 6 [ Fig. 6 ] is a schematic representation showing the center of a region of interest of a thermal image on a polar coordinate system. Fig. 7 [ Fig. 7 ] is an example of a graph showing the values of a normalization factor applied during a step of the process of the figure 1 as a function of a distance in pixels from a center of a region of interest of the thermal image of the figure 3 in the center of this image. Fig. 8 [ Fig. 8 ] is an example of a graph showing the values of the radial coordinate of the position of a person detected in the building space as a function of the distance in pixels from the center of the region of interest of the thermal image of the figure 3 in the center of this image, as well as depending on various installation heights of the thermal imager of the figure 2 . Fig. 9 [ Fig. 9 ] shows a schematic representation of the installation of a home automation detector including the thermal imager of the figure 2 in space E of a building. Fig. 10A [ Fig. 10A ] shows an example of a curve showing the evolution of an angular correction factor used during one of the steps of the process of the figure 1 . Fig. 10B [ Fig. 10B ] shows another example of a curve showing the evolution of the angular correction factor used during one of the steps of the process of the figure 1 . Fig. 11 [ Fig. 11 ] shows an example of density mapping obtained based on the method of the figure 1 . Fig. 12 [ Fig. 12 ] shows an example of a filter applied to sort the positions of people obtained from the method of the figure 1 . Description des modes de réalisation

[0017] There figure 1 represents a flowchart of a method 100 according to the invention for obtaining the position of one or more person(s) in a space of a building. The method comprises steps 110 to 140 which will be described subsequently.

[0018] A "building" means any immovable construction, such as a building, an apartment, an office, etc. The building space may be an interior part of the building, for example, a room or an area of a room, or an exterior part of the building, for example, a balcony, a courtyard, or a garden.

[0019] As will be detailed later in this text, the "position of a person" means the coordinates of the point in space where the person is located relative to the origin of a predetermined reference frame for the space of the corresponding building. These coordinates can be Cartesian or polar. They are expressed in units of length, for example in meters, and, in the case of polar coordinates, in angular units, for example in degrees.

[0020] Now steps 110 to 140 of method 100 will be described.

[0021] Step 110 includes acquiring a thermal image of a scene including one or more people. The scene may include all of the selected space of the building or a portion of that space.

[0022] The acquisition 110 of the thermal image is for example made by a thermal imager 10, visible on the figure 2 . For explanatory purposes, in this figure, the thermal imager 10 is oversized relative to the size of the scene, referenced S, and of the space, referenced E.

[0023] As visible in particular on the figure 9 , the thermal imager 10 is advantageously arranged at a high point above the ground W of the space E, in particular at a height h which allows it to be above any person P present in the space E. For example, the height h is greater than or equal to 2.2 m. In the case of an interior space, the thermal imager 10 is for example installed on the ceiling. In the case of an exterior space, the thermal imager 10 can be installed against a wall delimiting the space E or on a high point of a high element included in the space E, for example a lamppost. Also, all the people present in the scene S during the step 110 appear on the thermal image obtained, regardless of their size.

[0024] As will be detailed, the thermal imager 10 is configured to obtain the thermal image from the absorption of the infrared radiation 11 emitted by the different bodies present in the scene S. By “body” is meant any living being or any object present in the scene S. Thus, the body can be a person, an animal, a plant, an electronic device, a piece of furniture or a wall included in the scene S, among others.

[0025] The thermal image obtained is formed by pixels distributed on a plane comprising a first direction and a second direction perpendicular to each other. The number of pixels in the thermal image is determined by the resolution of the thermal imager 10. For example, for an imager with a resolution of 80x80 pixels, the thermal image obtained comprises a row of 80 pixels in the first direction, and a column of 80 pixels in the second direction.

[0026] The thermal imager 10 comprises a computing device 12, a set of thermal elementary sensors 14, also called infrared sensors, and a lens 16.

[0027] On the example of the figure 2 , the computing device 12 is integrated into the thermal imager 10, but it could be an external device connected to the thermal imager 10. The computing device 12 comprises a processor 13. The processor 13 can intervene in the implementation of steps 110 to 140 of the method 100 as will be detailed.

[0028] The infrared sensors of the set 14 are distributed along the first direction and along the second direction. In particular, the number of infrared sensors along each of the first and second directions is equal to the number of pixels on the thermal image along each of these directions.

[0029] Each infrared sensor is configured to absorb the infrared radiation 11 emitted by the bodies present in the scene S. The infrared radiation 11 absorbed by the sensor 14 causes the variation of a physical quantity at the level of each thermal elementary sensor which is a function of the quantity of infrared radiation absorbed. In a known manner, the higher the temperature of a body, the more this body emits infrared radiation, which increases the quantity of infrared radiation 11 absorbed by the sensor 14. The infrared sensor 14 thus generates different signals depending on the temperature of each body present in the scene S.

[0030] Depending on how the signals to create the thermal image are generated, the infrared sensor 14 may be a bolometer (or microbolometer), a pyroelectric sensor or a thermoelectric sensor, among others.

[0031] In a bolometer or microbolometer, the signals for obtaining the thermal image are generated by the sensor 14 from a variation in its electrical resistance. In particular, the heating of the sensor 14 caused by the absorption of infrared radiation causes the electrical resistance of the sensor 14 to vary. The value of the variation in electrical resistance is associated with a variation in temperature.

[0032] In the case of a pyroelectric sensor, the heating of the sensor 14 caused by the absorption of infrared radiation generates a variation in electrical polarization in the sensor 14 which is linked to the increase in temperature.

[0033] In the case of a thermoelectric sensor, the absorbed infrared radiation generates a variation in electrical voltage in the sensor 14 associated with a variation in temperature.

[0034] The signals generated by the infrared sensor 14 therefore depend on the temperature of each body in the scene S.

[0035] Advantageously, the infrared sensor 14 operates at room temperature. The thermal imager 10 is therefore an uncooled thermal imager.

[0036] As previously indicated, the signals generated by the infrared sensor 14 are transmitted to the computing device 12. The processor 13 processes these signals which correspond to the thermal image associated with the scene S.

[0037] As stated above, the thermal image comprises a plurality of pixels. Each pixel represents a specific temperature point of a body present in the scene S. Depending on the temperature of the respective point, each pixel of the thermal image can be represented in grayscale. Also, as is clear from the figure 3 , the thermal image includes 15 light areas and 17 dark areas.

[0038] The light areas 15 and the dark areas 17 help to identify the different heat sources present in the scene. In the image of the figure 3 , the light zones 15 correspond to bodies whose temperature is within a given temperature range, called the reference temperature range, while the dark zones 17 correspond to bodies whose temperature is outside the reference temperature range. The reference temperature range extends for example from 20°C to 40°C. For the sake of clarity, in the following the light zones 15 will also be called “hot zones”, and the dark zones 17 will also be called “cold zones”. It should be noted, however, that this correspondence between hot zones / light zones and cold zones / dark zones is not limiting. It should also be noted that the temperature of the cold zones 17 is generally lower than the temperature of the hot zones 15, but the opposite is also possible.In other words, the temperature of cold zones can be either below the lower limit of the reference temperature range or above the upper limit of the reference temperature range.

[0039] The lens 16 is a wide-angle lens. The wide-angle lens 16 has a short focal length, which gives it a wide field of view, in particular wider than the field of view of the human eye. In particular, a field angle γ of the wide-angle lens is preferably greater than or equal to 120°. As can be seen from the figure 2 , in the present text the field angle γ is understood to be the opening angle of the conical field of vision 18. By “opening angle” is understood the maximum angle formed by two generating lines 18A, 18B of the cone 18. In other words, the opening angle of the cone 18 corresponds to twice the angle formed between the axis of revolution A of the cone 18 and one of its generating lines.

[0040] According to a non-limiting example, the wide-angle lens 16 is a hypergone lens, also called a “fisheye lens”. The hypergone lens is a type of wide-angle lens having a field angle γ which can reach 180°.

[0041] Since the wide-angle lens 16 has a wide field of view, it is possible to capture larger scenes S than with a lens whose field of view is similar to that of the human eye. However, the thermal image obtained from the thermal imager 10 provided with the wide-angle lens 16 has convex distortion.

[0042] To better understand this phenomenon, we now refer to the figures 4A And 4B . There figure 4A , shows a checkerboard-like reference image acquired through a hypergonal lens. As clearly shown, the reference image has a convex distortion. The figure 4B shows the shape of the undistorted checkerboard-type reference image. Unlike the figure 4A , in this case the image has a flat appearance.

[0043] An orthonormal reference frame comprising a first axis U1 and a second axis U2 is used for each of these figures. Point O corresponds to the origin of the orthonormal reference frame. Origin O corresponds to the center of the image of the scene obtained, as well as to the center of this scene. The center of the image obtained corresponds to the optical center of the wide-angle lens used.

[0044] The first axis U1 indicates the distance in pixels along a direction parallel to the U1 axis occupied by a given pixel of the image relative to the origin O. The second axis U2 indicates the distance in pixels along a direction parallel to the U2 axis of a given pixel of the image relative to the origin O. The origin O corresponding to the center of the image of the scene S obtained, the number of pixels arranged on either side of the U1 axis is the same. Similarly, the number of pixels arranged on either side of the U2 axis is the same. Thus, for an imager with a resolution of 80x80 pixels, 40 pixels are arranged on each side of the U1 axis, and 40 pixels are arranged on each side of the U2 axis.

[0045] As clearly visible on the figure 4A , the further we move away from the center O of the image obtained by the imager, the more the image is distorted compared to the image of the figure 4B .

[0046] As will be detailed later, the closer one is to any direction forming an angle of 45° with the U1 and U2 axes, the greater the distortion. Thus, areas located on directions forming an angle equal to 45°, 135°, 225° or 315° with the U1 axis are the areas most subject to distortion related to the use of the wide-angle lens.

[0047] Step 120 of method 100 will now be described. Step 120 comprises extracting a region of interest corresponding to a person visible in the distorted thermal image.

[0048] As indicated previously, in the distorted thermal image obtained during step 110, the hot zones 15 correspond to the bodies in the space whose temperature is in the reference temperature range. Such a temperature range corresponds to a temperature range that could potentially be the temperature of the people present in the scene S depending on their clothing. Any person present in the scene S therefore appears in the distorted thermal image as a hot zone 15. In the scene S there may be other bodies whose temperature is in the reference temperature range, for example an animal or an electronic device in use. Consequently, such bodies also appear as hot zones 15 in the distorted thermal image obtained during step 110.

[0049] In order to distinguish the hot zones 15 corresponding to people from the other hot zones 15, the method 100 uses a person detection algorithm allowing a more detailed analysis of these hot zones 15. The algorithm is for example included in software executed by the computing device 12 of the thermal imager. This software can be implemented by the processor 13.

[0050] By applying this algorithm, the computing device 12 is capable of identifying among the hot zones of the distorted thermal image, regions of interest 21. On the figure 5 , the regions of interest 21 are the hot zones 15 on which circles are arranged. The regions of interest 21 correspond to people present in the scene S.

[0051] According to a non-limiting example, the person detection algorithm can be configured to detect certain characteristics specific to people, for example from the analysis of the surface of each hot zone 15 or the capacity of each of these hot zones 15 to move in the scene S.

[0052] It is noted that the identification of the regions of interest 21 on the deformed thermal image does not make it possible to directly deduce the real position of the people relative to the center of the scene S. Indeed, the thermal image being deformed, the position of a region of interest 21 is deviated relative to the position of the corresponding person in the scene S. To correct this deviation, the method 100 comprises the steps 130 and 140 which will be described below.

[0053] Step 130 of the method 100 comprises determining the polar coordinates of the center of each region of interest 21 in a polar reference frame whose origin is located at the center O of the distorted thermal image.

[0054] By "center of each region of interest" is meant the center of gravity of the points forming the region of interest 21. The center of each region of interest 21 corresponds to the central point of each circle positioned on each of the regions of interest 21 on the figure 5 .

[0055] On the figure 6 , the polar reference frame corresponds to the orthonormal reference frame comprising the U1 and U2 axes previously used on the figures 3 à 5 . As already indicated, the center O of the distorted thermal image corresponds to the optical center of the wide-angle lens 18.

[0056] We consider that point M is the center of a region of interest 21. Point M has coordinates (x, y) in the orthonormal frame used. This means that point M corresponds to the point located at a distance of x pixels from the origin O following the direction parallel to the axis U1 and located at a distance of y pixels from the origin O following the direction parallel to the axis U2.

[0057] The determination of the polar coordinates of point M includes, on the one hand, the determination of the distance r separating point M from the center O of the deformed thermal image. This distance r corresponds to the modulus of a vector OM connecting the center O and the point M. The center O of the thermal image having coordinates (0, 0), the distance r is defined as: r = x 2 + y 2

[0058] The distance r is a decimal number without units.

[0059] It is noted here that the identification of the regions of interest 21 is limited to the surface included in a circle 23 (visible on the figures 3 And 5 ) whose diameter in pixels is equal to the total number of pixels on one of the directions parallel to one of the axes U1 or U2. As already explained, this total number of pixels depends on the resolution of the thermal imager. In the example presented of the imager with a resolution of 80x80 pixels, the diameter of the circle 23 is therefore equal to 80 pixels. This implies that the maximum value r max of distance r which is obtained during the present step 130 of the method 100 is equal to the distance in pixels separating the last pixel located on the axis U1 or U2 from the origin O. In other words, the maximum value r max of distance r which can be obtained is equal to the radius in pixels of the circle 23. In the case of the imager with a resolution of 80x80 pixels, the maximum value r max of distance r is therefore equal to 40. In the example of the figure 6 , the maximum value r max of distance r is therefore obtained for point N.

[0060] On the other hand, the determination of the polar coordinates of the point M includes the determination of the polar angle or azimuth θ. As is clear from the figure 6 , the azimuth θ corresponds to the angle formed between the axis U1 and the vector OM. The azimuth θ is therefore defined as: θ = tan − 1 y x

[0061] The polar coordinates (r, θ) of the center of a region of interest 21 allow this center to be located unambiguously in the distorted thermal image.

[0062] The center of a region of interest 21 is considered here as corresponding to the “center of gravity” of the person identified for the region of interest 21 on the deformed thermal image. The center of gravity of a seated or standing person being typically located at a height of between 0.7 m and 1.3 m from the ground, it is considered here that the polar coordinates are given at a height h 0 from the ground which is for example equal to 1 m.

[0063] Step 140 of the method 100 comprises obtaining the actual position of the person by applying a wide-angle correction function to the distance r to compensate for the distortion of the thermal image. This step 140 is in particular applied to the distance r obtained for each region of interest identified during step 120.

[0064] The wide-angle correction function includes a first factor, called the normalization factor.

[0065] The normalization factor aims to normalize the distance coordinate r obtained for the center of each region of interest 21 identified during step 120. By "normalize" is meant to associate with each value of the distance coordinate r, a value between 0 and 1. In particular, the normalization factor is defined so that its value for the center of a region of interest 21 coinciding with the center O of the deformed thermal image is equal to 0, and so that its value for the center of a region of interest 21 whose coordinate r takes its maximum value r max is equal to 1. Thus, for a region of interest 21 whose center does not coincide with the center O of the thermal image but which is located at a radial distance from the center O less than the maximum value r max of the distance r, the normalization factor takes a value greater than 0 and less than 1.More precisely, the normalization factor is defined so that for any value of distance r between 0 and r max , the value of the normalization factor follows an increasing law as the distance r approaches its maximum value r max , as in the graph of the . figure 7 .

[0066] According to a non-limiting example, the normalization factor Fn is defined as: Fn = r β 1 − α r 2

[0067] Applying this definition of the normalization factor, we obtain a distribution of values of the normalization factor as a function of the distance r similar to that of the graph of the figure 7 .

[0068] The coefficient α is a distortion coefficient that determines the curvature of the distribution obtained in the graph of the figure 7 Changing the value of the coefficient α allows this curvature to be modified, which makes it possible to adjust the growth rate of the normalization factor between the values 0 and 1.

[0069] The coefficient β is a scaling coefficient which ensures, whatever the value chosen for the distortion coefficient α, that the normalization factor takes the value 1 when the coordinate r of the center of a region of interest 21 is equal to the maximum value r max.

[0070] The values of the coefficients α and β are therefore predetermined so as to guarantee that the normalization factor effectively makes it possible to associate each distance coordinate value r with a value between 0 and 1. These predetermined values are obtained empirically.

[0071] For example, in a case where r max is equal to 40, α can be equal to 0.00032 and β can be equal to 83.

[0072] As will be detailed, thanks to the normalization factor, it is possible to obtain the real position of each person in space E in a simple way. It is sufficient to know only the field angle γ of the wide-angle lens 16 and the installation height h of the thermal imager 10. The field angle γ of the wide-angle lens 16 is usually indicated in its technical data sheet, while the installation height h of the thermal imager 10 is generally easily measurable

[0073] For this purpose, the wide-angle correction function includes a second factor, called the metric factor, which takes into account the field angle γ and the installation height h.

[0074] More precisely, the metric factor has a value which depends on the field angle γ of the wide-angle lens 16, the installation height h of the thermal imager 10 and the height h 0 admitted as reference height relative to the ground W. In particular, according to an example, the metric factor Fm is defined as: Fm = h − h 0 tan γ 2

[0075] The metric factor is therefore a factor that is constant as long as the field angle γ, the height h and the height h 0 are determined. In other words, this factor does not depend on the coordinates of the center of the identified regions of interest 21.

[0076] As can be clearly deduced from the figure 9 , the value of the metric factor obtained from [Math. 4] corresponds to the distance R separating the center C of the scene S from an extreme point D located at the edge of the scene S. This extreme point D is found on the thermal image obtained during step 110 at the edge of the latter with a distance coordinate r equal to the maximum value r max .

[0077] The metric factor is expressed in meters or any other commonly used unit of length.

[0078] The wide-angle correction function Fc is defined as the product of the normalization factor Fn and the metric factor Fm, that is: Fc = Fn ⋅ Fm

[0079] Thus, when the normalization and metric factors are defined as proposed by [Math. 3] and [Math. 4], the wide-angle correction function Fc is: Fc = r β 1 − α r 2 h − h 0 tan γ 2

[0080] Since the normalization factor is dimensionless, the result of the wide-angle correction function is given in meters (or any other unit of length) using the metric factor. The result of the wide-angle correction function corresponds in particular to the distance separating the associated person from the center C of the scene S. Thus, the result of the wide-angle correction function and the azimuth θ obtained during step 130 give the real position of the person in space E with a satisfactory degree of precision.

[0081] A graphical representation of the Fc function is shown in the figure 8 . This graph represents three curves of evolution of the wide-angle correction function Fc as a function of the distance r, each curve being associated with a different installation height h of the thermal imager 10. On this graph, the height h1 is lower than the height h2, and the height h2 is lower than the height h3. As can be seen from this figure, the installation height h of the thermal imager 10 slightly influences the evolution of the curve. In particular, the higher the installation height h of the thermal imager 10, the more the distance separating the identified person from the center C of the scene S increases with the increase in the distance r.

[0082] In order to further increase the accuracy of the actual position of the person obtained from the wide-angle correction function, a third factor Fa, called the angular correction factor, can be included in the wide-angle correction function Fc. The angular correction factor is dimensionless.

[0083] Indeed, as previously indicated, the closer a region of interest 21 of the deformed thermal image is to a direction forming an angle of 45° with the U1 and U2 axes, the greater its deformation. Thus, the result of the wide-angle correction function as defined by [Math. 6] turns out to be less precise for such a case.

[0084] The angular correction factor compensates for this lack of precision. In particular, when the wide-angle correction function includes the angular correction factor Fa, it is defined as: Fc = Fn ⋅ Fm ⋅ Fa

[0085] The value of the angular correction factor is a function of the value of the azimuth θ. figures 10A et 10B show two examples of curves defining the value of the angular correction factor as a function of the azimuth θ of the center of the region of interest. As seen in these figures, the value of the angular correction factor is between 1 and a maximum value Fa max . For example, the maximum value Fa max of the angular correction factor can be between 1.10 and 1.30.

[0086] In the case of the figure 10A , the value of the angular correction factor Fa follows a sawtooth-shaped distribution whose maximum value Fa max is associated with the values 45°, 135°, 225° and 315° of the azimuth θ. This makes it possible to compensate for the greater deformation specific to the regions of interest located on one of these directions. The angular correction factor Fa is equal to 1 when the azimuth is equal to 0°, 90°, 180° or 270°. This implies that when the center of the region of interest 21 is located on one of the axes U1, U2 of the polar reference frame presented above, the wide-angle correction function including the angular correction factor gives the same result as the wide-angle correction function when it does not include the angular correction factor. For azimuth values θ other than 0°, 45°, 90°, 135°, 180°, 225°, 270° and 315°, the value of the angular correction factor is greater than 1 but less than the maximum value Fa max .

[0087] In the case of the figure 10B , the value of the angular correction factor Fa follows a sawtooth-shaped distribution but between which plateaus are interspersed. In this case, the maximum value Fa max is always associated with the values 45°, 135°, 225° and 315°. However, unlike the distribution of the figure 10A , the value 1 of the angular correction factor is associated with value ranges centered around the values 0°, 90°, 180° or 270°. For example, these value ranges are from 0° to 20°, from 70° to 110°, from 160° to 200° and from 250° to 290°.

[0088] The method 100 explained above therefore makes it possible to obtain the actual position of the people identified in the space E of the building in a precise manner. In addition, since the wide-angle correction is only applied to the distance coordinate r, the implementation of the method does not require heavy calculations.

[0089] It is noted that the wide-angle correction function and the definition of its various factors presented above are not limiting, other wide-angle correction functions being able to be used during step 140. For example, the wide-angle correction function could be a polynomial function.

[0090] The method 100 can find several applications.

[0091] For example, method 100 may be used to create a traffic map of space E, an example of which is visible in the figure 11 . The attendance map is obtained after repeating the method 100 as described above for an extended period of time, for example two weeks. In particular, during this period of time, the method 100 may be repeated at regular or irregular intervals. Thus, several thermal images are generated to obtain the actual positions of the people by applying the wide-angle correction function. By superimposing the positions obtained for the people after each repetition of the method 100, it is possible to determine in which areas of the space E the people are usually present, in which areas of the space the people are occasionally identified and in which areas of the space no person is ever identified. Different colors are assigned to each of these areas on the attendance map. For example, on the figure 11 , the light areas, referenced 30, correspond to circulation areas, such as a corridor, in which people are occasionally identified. The dark areas, referenced 32, correspond to areas where no person is identified. This is the case for areas occupied by furniture or walls. Finally, the areas 34, whose color is intermediate between the light areas 30 and the dark areas 32, correspond to areas where people are usually identified. These areas 34 may correspond, for example, to workstations.

[0092] The application of the method 100 described above for producing attendance maps makes it possible, thanks to the use of the wide-angle lens, to study the presence of people in a large space with a single thermal imager. Furthermore, thanks to the application of the wide-angle correction function, the position of the people identified in the space is obtained with a high level of precision, so that the different zones of the attendance map produced are delimited precisely.

[0093] The method 100 can also be used in counting people. In particular, the identification of the regions of interest makes it possible to easily count the number of people in the space E. In addition, since the position of these people in the space E is determined precisely, the method 100 is of major interest for counting people when several thermal imagers are distributed across a large room. In such a case, the lack of precision of the methods for obtaining the position of a person of the prior art leads to errors in counting people in the room. These errors include in particular the multiple counting of the same person who is located at a position which is covered by the field angle of several thermal imagers.Indeed, since the position of the persons obtained by these prior art methods is deviated from the true position of the person, each of the imagers can identify the same person in different positions. The same person therefore risks being counted several times.

[0094] As visible on the figure 12 , in order to facilitate the counting of people, a filter 40 can be applied to delimit the identification and counting of people to a surface of the space E of a size smaller than the surface of this space E that the field angle of the thermal imager allows to be covered. In particular, the identification and / or counting of people is limited to the surface included in the filter 40. On the figure 12 , the filter 40 has a square shape without this being limiting. The filter 40 could also have, for example, a rectangle, circle, or any other shape defined by a set of points which define a closed perimeter.

[0095] The filter 40 can simplify the method 100 described above when it is known that a hot body which is not a person, for example a radiator, is present in the scene covered by the thermal imager. In particular, the filter can be configured so as to exclude the identification and / or counting of people in the part of the scene occupied by such a hot body.

[0096] Advantageously, when several thermal imagers are arranged in a room, the filters 40 applied by the imagers are adjacent to each other, which makes it possible to guarantee the identification and / or counting of people over the entire surface covered by the field angles of the thermal imagers, while avoiding the simultaneous identification and / or counting of the same person by several imagers.

[0097] The present disclosure also relates to the thermal imager 10 described above.

[0098] Finally, the present disclosure relates to a home automation detector 50, visible on the figure 9. The home automation detector 50 comprises the thermal imager 10. The thermal imager 10 is integrated into the home automation detector 50. The home automation detector 50 is configured to implement the method 100 for obtaining the position of a person in the space E as described above. It is noted that the home automation detector 50 may further comprise other devices giving it other functionalities. For example, the home automation detector 50 may comprise a brightness sensor, a sound sensor, a motion sensor, a temperature sensor, a smoke sensor, an ambient air pollution sensor, etc.

Claims

1. Method (100) for obtaining the position of a person in a space (E) of a building via a thermal imager (10) located at height in said space (E) and comprising a wide-angle lens (16), the method comprising steps of: a) acquiring (110), with the thermal imager (10), through its wide-angle lens (16), a deformed thermal image of a scene (S) in the space (E) comprising the person, the thermal image being optically deformed by the wide-angle lens (16), and the deformed thermal image comprising a hot zone (15); b) extracting (120) a region of interest (21) corresponding to the person from the deformed thermal image using an algorithm for detecting hot zones (15); c) determining (130) the position of the region of interest (21) in the deformed thermal image by identifying the polar coordinates distance r and angle θ of the centre of the region of interest (21) in a polar coordinate system the origin of which is located at the centre (O) of the deformed thermal image; d) obtaining (140) the actual position of the person in the space (E) by applying a wide-angle correction function to the distance r so as to compensate for the deformation of the thermal image due to the wide-angle lens (16).

2. Method (100) according to Claim 1, wherein the wide-angle correction function applied in step d) comprises a normalization factor Fn for normalizing the distance coordinate r of the centre of the region of interest (21) and a metric factor Fm accounting for a field angle of the wide-angle lens (16) and for a height of installation of the thermal imager (10) in the space (E) of the building with respect to a floor (W) of said space (E), the wide-angle correction function especially comprising the product of the normalization factor Fn and of the metric factor Fm.

3. Method (100) according to the preceding claim, wherein the normalization factor is defined as: Fn = r β 1 − α r 2 where β is a scale coefficient and α is a distortion coefficient, the scale coefficient β and the distortion coefficient α having predetermined values.

4. Method (100) according to Claim 2 or Claim 3, wherein the metric factor is defined as: Fm = h − h 0 tan γ 2 where h is the height of the thermal imager (10) with respect to the floor (W) of the space, h0 is a value representative of the height of the centre of the region of interest (21) with respect to the floor (W) and γ is the field angle of the wide-angle lens.

5. Method (100) according to one of Claims 2 to 4, wherein the wide-angle correction function applied in step d) further comprises an angular correction factor Fa, and wherein a value of the angular correction factor Fa is dependent on a value of the angle coordinate θ, the value of the angular correction factor Fa following a sawtooth-shaped distribution a maximum value of which is associated with the values 45°, 135°, 225° and 315° of the angle coordinate θ.

6. Method (100) according to Claim 5, wherein the wide-angle correction function comprises the product of the normalization factor Fn times the metric factor Fm times the angular correction factor Fa.

7. Thermal imager (10) comprising a wide-angle lens (16), the thermal imager (10) being configured to implement the method (100) according to any one of Claims 1 to 6.

8. Thermal imager (10) according to Claim 7, wherein the lens (16) is a fisheye lens.

9. Thermal imager (10) according to Claim 7 or 8, comprising a long-wave-infrared imaging sensor (14) for acquiring the thermal images.

10. Home automation sensor (50) comprising a thermal imager (10) according to one of Claims 7 to 9.