Dynamic masking of private areas in a PTZ camera.

Abstract

The invention relates to a method for statically masking areas of interest in images of a surveillance area, this method comprising the following steps: - during a static detection phase: - scanning the surveillance area and acquiring a plurality of images by means of a PTZ type surveillance camera, - sending the plurality of images to a processing unit, - automatic detection, within the processing unit, of areas of interest visible in each image of the plurality of images, - during a static grouping phase, automatic determination of one or more masks, each mask grouping at least one area of ​​interest from among all the areas of interest detected on the plurality of images, - sending the mask(s) to the surveillance camera, - during an exploitation phase, when at least one area of ​​interest is visible on an image intended to be transmitted to a display screen,application within the camera of at least one mask covering this area of ​​interest so that this area of ​​interest is blurred in the image transmitted to the display screen. . Figure for the abstract: Fig. 1,

Description

Title of the invention: Dynamic masking of private areas in a PTZ camera.

[0001] The present invention relates to a method for static or on-the-fly masking of areas of interest in images of a surveillance area.

[0002] It finds a particularly interesting application in the context of video surveillance in public places. Video surveillance is an essential tool for detecting any abnormal situation or behavior, but it often raises the question of violation of privacy. The need has arisen to find a compromise between surveillance to ensure the safety of people and property and respect for privacy.

[0003] Generally speaking, in Europe for example, image capture in public places is subject to the European General Data Protection Regulation (GDPR). Its objective is to protect the personal data of European citizens.

[0004] To do this, it is known to apply masks to images from a PTZ camera in order to blur parts of the image that would be private. The use of PTZ (Pan-Tilt-Zoom) cameras equipped with the masking functionality requires tedious work on the part of the operators (video patrollers) to manually enter the masks on the areas that they consider private. Thus, when the camera zooms in on one of these parts, the camera sends a blurred stream. This system, however, has several drawbacks:

[0005] - the number of masks is limited (generally to 32 masks); this limitation imposing either the grouping of several private areas, or the non-masking of certain areas;

[0006] - PTZ type cameras are generally under heavy load, which causes exhaustion of their mechanical components; thus, the correspondence between the position of the camera and the mask fixed by the operator becomes erroneous over time; this phenomenon being known as “mask slippage” and requiring the repetition of the tedious phase of deploying the mask;

[0007] - operator intervention to manipulate masks or correct slippages diverts it from its main function, which is to remove doubts about alarms sent by an intelligent video surveillance system.

[0008] The present invention relates to a new dynamic masking method allowing the operator to concentrate on his main function.

[0009] Another object of the invention is to propose a new masking method dynamic with precisely determined masks.

[0010] Another object of the invention is to limit the time taken to create the masks.

[0011] The invention also relates to an effective north setting to combat the slipping of masks.

[0012] At least one of the aforementioned objectives is achieved with a method of statically masking areas of interest in images of a surveillance area, this method comprising the following steps:

[0013] - during a static detection phase:

[0014] - scanning the surveillance area and acquiring a plurality of images at using a PTZ type surveillance camera,

[0015] - sending the plurality of images to a processing unit,

[0016] - automatic detection, within the processing unit, of visible areas of interest in each image of the plurality of images,

[0017] - during a static grouping phase, automatic determination of one or several masks, each mask grouping at least one area of ​​interest among all the areas of interest detected on the plurality of images,

[0018] - sending the mask(s) to the surveillance camera,

[0019] - during an exploitation phase, when at least one area of ​​interest is visible on an image intended to be transmitted to a display screen, application within the camera of at least one mask covering this area of ​​interest so that this area of ​​interest is blurred in the image transmitted to the display screen.

[0020] With the method according to the invention, a preliminary processing of automatic detection of areas of interest is carried out and these areas of interest are saved in a database.

[0021] Each area of ​​interest may, for example, correspond to a window. Detection consists of defining a frame around this window, each frame then being saved in a database internal or external to the processing unit. Each frame can be identified by its dimensions and coordinates in each image where the corresponding window is visible.

[0022] Masks are previously determined and then transmitted to the camera for application in real time during monitoring. By mask is meant a rectangular area [Xm, Ym, Wm, Hm] of the image, defined on a particular P(pan), T(tilt), Z(zoom). Generally speaking, a mask is a frame defining a part of an image.

[0023] The present invention makes it possible to automatically mask private areas in images from the surveillance camera. The notion of private area is to be defined by the user, the automatic detection algorithm then being configured to detect the type of area defined by the user.

[0024] Depending on the type of camera used and the user's wishes, during the grouping phase, the number of masks to be defined can be predetermined, for example less than or equal to the maximum number of masks authorized by the camera. Generally, it is possible to use a processing unit for several cameras, the communication between the processing unit and the cameras being able to be done by different types of protocols and means of communication, wired or wireless, via the Internet or within a private network, in an encrypted manner or not.

[0025] The present invention proposes a new masking solution which can comprise three phases: an automatic offline detection phase for the construction of a database of detections, a phase of formation of the masks from the detections of the areas of interest and an exploitation phase which consists of sending the masks to the PTZ camera to apply them in real time to images intended to be viewed on a display screen.

[0026] The subject of the present patent application relates to alternative solutions to a particular problem of automatic masking of images from a PTZ camera. In addition to the static masking solution defined above, a method for on-the-fly masking of areas of interest in images of a surveillance area is also proposed, this method comprising the following steps carried out during an operating phase:

[0027] - acquisition of an image of the surveillance zone by means of a surveillance camera PTZ type surveillance, this acquired image being intended to be transmitted to a display screen,

[0028] - sending the acquired image to a processing unit,

[0029] - within the processing unit:

[0030] - projection onto the acquired image of areas of interest previously detected for the same camera positioning, i.e. identical P(pan), T(tilt) and Z(zoom) coordinates, or

[0031] - automatic detection within the processing unit of areas of interest in said at least one image,

[0032] - during an on-the-fly grouping phase, automatic determination of one or several masks, each grouping at least one area of ​​interest of the acquired image,

[0033] - sending the mask(s) to the surveillance camera,

[0034] - application within the camera of the determined mask(s) so that any hidden part is blurred in the image transmitted to the viewing screen.

[0035] This “on-the-fly” embodiment makes it possible to create and apply the masks “on-the-fly”, i.e. in real time during monitoring. Ideally, the processing unit intervenes on an image acquired by the camera before transmission to the display screen. The processing unit thus performs the detection in real time areas of interest, determination of masks, transmission of masks to the camera for application to the current image in real time. The detection of areas of interest can correspond exactly to the determination of masks, that is to say that each detected area of ​​interest is identified as a mask.

[0036] The advantage of an on-the-fly mode is in particular the fact of determining a new set of masks almost for each image. That is to say that in the context of using a PTZ camera offering a maximum number of masks, this on-the-fly embodiment makes it possible to use this maximum number of masks each time. In the static masking method, the masks are determined beforehand for all possible images and their number is limited to this maximum number.

[0037] According to one embodiment of the invention, the on-the-fly masking method may comprise a preliminary step during which said previously detected areas of interest are obtained during a static detection phase comprising the following steps:

[0038] - scanning the surveillance area and acquiring a plurality of images at by means of the surveillance camera,

[0039] - sending the plurality of images to the processing unit,

[0040] - automatic detection, within the processing unit, of visible areas of interest in each image of the plurality of images.

[0041] In other words, for detection, provision is made, as an alternative or complementary method, for the possibility of using prior detection of the areas of interest. In this case, the coordinates of the areas of interest previously detected during the static detection phase, for example, are taken, then they are defined as masks to be transmitted to the camera.

[0042] It is possible to combine the static masking method with the on-the-fly masking method. Indeed, it is possible to implement the static masking method for a limited number of masks L and to reserve some masks Q for the implementation of the on-the-fly masking method, the sum of L and Q being equal to the maximum number of masks authorized by the PTZ camera used.

[0043] In other words, the present invention proposes several modes of operation:

[0044] - dynamic on-the-fly masking where masks are defined in real-time and sent to the camera to blur private areas,

[0045] - dynamic masking with a fixed number of masks in order to limit the use of the camera,

[0046] - a hybrid mode where part of the masks is fixed and another part is de completed on the fly during camera movement.

[0047] According to an advantageous characteristic of the invention, the detection step au- automation can be based on a deep convolutional neural network.

[0048] Artificial intelligence is thus used to learn to recognize the type of areas of interest to be masked. Image analysis and deep learning techniques are then used to identify areas of interest or private areas such as building windows, French windows, terraces, etc. Other techniques such as support vector machines can be used as learning techniques.

[0049] The video streams which are transmitted by the camera to the viewing screen are then automatically blurred at the level of these private areas.

[0050] Machine learning techniques can also be used, such as the clustering method for the grouping step, during which several areas of interest are grouped into a single mask.

[0051] The present invention thus makes it possible to automate and fully exploit the masking functionality of PTZ cameras based on artificial intelligence tools for analyzing images and grouping shapes to define masks.

[0052] According to one embodiment of the invention, during the static detection phase, for a first and a second different image having in common an area of ​​interest, if this area of ​​interest is detected in the first image and not in the second image, the processing unit can be configured to apply a geometric transformation also making it possible to locate this area of ​​interest in the second image.

[0053] In fact, a transfer function is applied between two images in order to exhaustively detect all the areas of interest visible on an image.

[0054] According to an advantageous characteristic of the invention, during the static grouping phase, the step of automatically determining one or more masks may comprise the following steps:

[0055] - formation of a panoramic view from a set of acquired images,

[0056] - projection onto this panoramic view of all the areas of interest detected on the plurality of images, and

[0057] - grouping the areas of interest into several groups and determining a mask per group.

[0058] The invention is particularly remarkable in that a panoramic view is produced onto which all detected areas of interest are projected. In this way, masks overlapping several images can be easily determined.

[0059] Preferably, the panoramic view comprises images acquired for all P(pan) and T(tilt) coordinates of the PTZ camera, but with a predefined fixed zoom. This The predefined fixed zoom can advantageously be the lowest zoom, ideally a zoom equal to 1. In other words, we choose a zoom that makes a maximum number of areas of interest visible.

[0060] According to an advantageous embodiment of the invention, the projection of the areas of interest onto the panoramic view can be carried out by applying a geometric transformation between the plurality of images acquired during the scanning and the images of the panoramic view.

[0061] According to the invention, the grouping phase may further comprise a post-processing step during which useful areas are detected in the plurality of acquired images and the determined mask(s) are modified so as not to cover these useful areas.

[0062] The useful areas can for example be the roads present in an outdoor surveillance zone. To do this, deep learning techniques can also be used to detect these useful areas. It is also possible to detect additional elements such as vehicles, poles, white stripes, etc. in order to deduce the presence of a road. Then, the masks determined during the grouping phase and covering useful areas are modified so as to reveal these useful areas. The modification can consist of changing the dimensions and / or the shape of a mask or splitting a mask into several masks, provided that the maximum number of masks authorized is respected.

[0063] According to one embodiment of the invention, the acquisition step may consist of acquiring all the images of the entire surveillance zone. This then involves a continuous or fixed-step scan of the entire surveillance zone.

[0064] According to an advantageous embodiment of the invention, the camera can be configured to apply a mask to an image only when at least one area of ​​interest included in this mask has a predetermined size. This predetermined size can be defined such that the height of the area of ​​interest is greater than or equal to 1 / 3 of the height of the image. A mask is thus applied only to areas of interest such as windows for which viewing may result in an obstruction to local privacy protection legislation. Preferably, this size criterion is a parameter that the user can modify.

[0065] It is also envisaged to set up a northing or recalibration process, which for a given image, the processing unit is configured to:

[0066] - carry out automatic detection of areas of interest,

[0067] - compare the areas thus detected with the areas of interest previously detected for the same image,

[0068] - if there is a difference, setting up a northing process during from which the camera is realigned relative to predetermined fixed points.

[0069] This northing process helps combat mechanical drift in PTZ cameras. Automatic detection of areas of interest is used to recalibrate the camera and thus ensure that the masks completely hide private areas.

[0070] Indeed, PTZ type cameras are generally very stressed, which causes the exhaustion of their mechanical components. Thus, the correspondence between the position of the camera and the mask fixed by the operator becomes erroneous over time. This phenomenon is known as “mask slippage” and requires in the prior art the repetition of the tedious phase of deploying the masks.

[0071] The processing unit according to the invention is preferably a remote server capable of communicating and implementing the methods according to the invention with several cameras. It is also possible to envisage a processing unit integrated or associated with a camera.

[0072] According to the invention, the camera can be mobile. In the case of scanning, the movement of the camera can be carried out by means of a robot or a drone. The camera can also be arranged in a mobile manner on a rail. In this case, a scanning phase with the mobile camera makes it possible to record the coordinates of the area of ​​interest, the position of the camera on its mobile path and its position or its PTZ coordinates. The geometric transformation between two images must thus take into account the position of the camera on its mobile path; this position can be, for example, the position of the camera on its rail or the position of the robot or drone moving this camera.

[0073] In particular, for a mobile camera, the step of projecting areas of interest onto the panoramic view is carried out by applying a geometric transformation between the plurality of images acquired during scanning and the images of the panoramic view, this geometric transformation thus taking into account the position of the camera. This amounts to determining a transfer function between two images of two different reference frames. It is necessary to be able to project a window detected in PI, T1, Z1 with a camera position in XI, Y1, Z1 into an area in P2, T2, Z2 with a camera position in X2, Y2, Z2. For a fixed camera, X2=X1, Y2=Y1 and Z2=Z1.

[0074] According to another aspect of the invention, there is provided a system for masking areas of interest in images of a surveillance area, this system comprising:

[0075] - a PTZ camera for acquiring images and applying masks to acquired images,

[0076] - a processing unit configured to carry out the following steps:

[0077] - automatic detection of areas of interest visible in each acquired image,

[0078] - automatic determination of one or more masks, each mask grouping at least one area of ​​interest detected,

[0079] - sending the mask(s) to the surveillance camera.

[0080] This system is particularly suitable for implementing the methods according to the invention.

[0081] A computer program product is also provided comprising instructions which, when the program is executed by the processing unit, cause the latter to implement the steps of the method according to the invention.

[0082] Other characteristics and advantages of the invention will appear on reading the detailed description of implementations and embodiments which are in no way limiting, with regard to the appended figures in which:

[0083] [Fig-1]: [Fig.l] is a schematic view of a system according to the invention,

[0084] [Fig.2]: [Fig.2] is one of the views of the surveillance area,

[0085] [Fig.3]: [Fig.3] is the photo of [Fig.2] on which frames are represented identifying areas of interest, such as windows, detected according to the invention,

[0086] [Fig.4]: [Fig.4] is another view of the camera in a position other than that of [Fig.3],

[0087] [Fig.5]: [Fig.5] is a panoramic view of the surveillance area for the projection of areas of interest,

[0088] [Fig.6]: [Fig.6] is an image displayed on a viewing screen on which a masked area is blurred.

[0089] The embodiments which will be described below are in no way limiting; it will be possible in particular to implement variants of the invention comprising only a selection of characteristics described below isolated from the other characteristics described, if this selection of characteristics is sufficient to confer a technical advantage or to differentiate the invention compared to the state of the prior art. This selection comprises at least one preferably functional characteristic without structural details, or with only a part of the structural details if this part only is sufficient to confer a technical advantage or to differentiate the invention compared to the state of the prior art.

[0090] In particular, all the variants and all the embodiments described are intended to be combined with each other in all combinations where there is nothing technically opposed to this.

[0091] The various embodiments of the present invention comprise various steps. These steps can be implemented by instructions of a machine executable by means of a microprocessor for example.

[0092] Alternatively, these steps may be performed by specific integrated circuits comprising hard-wired logic to perform the steps, or by any combination of programmable components and custom components.

[0093] The present invention may also be provided in the form of a product computer program which may comprise a non-transitory computer storage medium containing instructions executable on a computing machine, which instructions may be used to program a computer (or other electronic device) to perform the method.

[0094] Although the invention is not limited thereto, the methods according to the invention will be described for masking areas of interest which are windows of buildings visible in an exterior surveillance zone.

[0095] [Fig.l] is a schematic overall view of a system for implementing the methods according to the invention. A processing unit 7 is shown, connected via a private or internet-type network 8 to several PTZ cameras 9, 10 arranged in surveillance zones. For the following, only camera 10 will be described, camera 9 being able to be configured in the same way or differently.

[0096] There is also a computer 11 connected to the internet network 8 and linked to a display screen 12 to display images from one or more cameras 9, 10. Ideally, an operator (not shown) is installed in front of the display screen 12 and monitors one or more areas scanned by the cameras.

[0097] The processing unit 7 may be in the form of a remote central server integrating software and hardware means necessary for implementing the methods according to the invention. It integrates a software application for communication with the cameras with a graphical interface for configuring these cameras.

[0098] The camera 10 is a PTZ camera for "pan", "tilt", "zoom", that is to say a camera capable of producing panoramic views, tilting and zooming. Its movements can be controlled remotely. Other types of cameras incorporating masking intelligence can be used. The base of the camera 10 can be fixed or mobile on a rail allowing it to scan a well-defined area.

[0099] The camera 10 is advantageously equipped with hardware and software means enabling it in particular to manage masks in situ. That is to say that when a mask must be applied to an image, the camera is configured to transmit the image stream, but the part corresponding to this mask appearing stuck on the image displayed on a display screen.

[0100] The computer 11 is configured to receive the images, sometimes bounced, from the camera 10.

[0101] Static mode.

[0102] For the operation of the system according to the invention, a static mode is provided, generally comprising a configuration phase and an operating phase. During the configuration phase, the masks are determined and stored in a database and then sent to the camera 10. During the operating phase, i.e. the use phase, the masks are applied to images transmitted to the video screen. sualization 12.

[0103] In [Fig.l], steps 1-4 correspond to the parameterization phase. Steps 5 and 6 correspond to the operating phase.

[0104] The configuration phase actually includes a detection phase (excluding exploitation) for the construction of the detection database, a phase of formation of the masks from the window detections.

[0105] Detection phase.

[0106] The detection phase consists of programming the PTZ camera to make an exhaustive path in P, T and Z to cover the entire area to be monitored. Thus in [Fig.l], in step 1, a plurality of images is sent to the processing unit 7 by the camera 10. This is for example a sampling with 3 axes: {Pn=n AP], {Tm = m AT], {Zi = 1AZ], with n=l..N, m=l..M, 1=1..L. The camera 10 therefore travels NxMxL different positions. For each position (Pn, Tm, ZL), the camera stabilizes for a few seconds and the processing unit automatically detects all the windows in the image. Each detected window is mathematically modeled by the quadruple (x,y,w,h) where (x,y) is the position of the top left corner of the window and (w,h) are the width and height of the window.

[0107] A database is thus constructed in real time within the processing unit during the camera scan: each line of the database contains the following seven elements: [xi, yi, wi, hi, P, T, Z] which means that at the position P, T, Z of the camera, a window has been detected at the position (xi, yi, wi, hi).

[0108] The same window can be detected in two different views or is detected only in one view and not in the other even if it is visible there. This depends on the efficiency of the detector which can easily detect a window with a frontal exposure and miss an oblique window. However, if a window is detected in a view (PI, Tl, Zl) with coordinates (xl,yl,wl,hl), its coordinates (x2,y2,w2,h2) can be geometrically deduced in another view (P2,T2,Z2) with a geometric transformation T defined from P2-P1, T2-T1 and Z2 / Z1:

[0109] (x2,y2,w2,h2) = T (xl,yl,wl,hl)

[0110] This detection phase (scanning with detection), which can last for hours, can be programmed automatically several times a year in order to update the detection database. Indeed, during a period of the year, trees hide windows. Relaunching this detection phase during the autumn, for example, allows this to be taken into account. The detection database can also evolve due to changes in the landscape: construction, demolition, change of facade, etc.

[0111] In [Fig.2], we see an example of a view of a part of the surveillance area. This is an image of a building with many windows.

[0112] In [Fig.3], we can see an image of the surveillance zone acquired for example at position P=0, T=0, Z=1. The numerous windows 13 have been detected and are then identified by at least one frame around each of them.

[0113] In [Fig.4], we can see another image acquired by the camera but in a position P=24, T=4, and Z=8. Three windows are detected then represented by three frames 14-16.

[0114] Grouping phase.

[0115] The grouping phase consists of grouping the windows in such a way as to define a mask per group of windows.

[0116] In the static mode, the static grouping phase takes into account the constraint of a maximum number of masks (for example 32) to be transmitted at once to the camera.

[0117] The first step of this static grouping phase consists of forming a panoramic view summarizing all the views covered by the PTZ camera during the first detection phase. In [Fig.5] such a view is shown. The panoramic view contains the views over the entire extent of the pans and tilts but for a single zoom (for example Z=1 corresponding to a maximum zoom out).

[0118] On this panoramic view, we project all the detections recorded in the database. This projection is based on the same principle as the T transformation. For example, if we have chosen to form a panoramic view on Z=1 (maximum zoom out) with tilts 0, we must browse the database and project each (xi,yi,wi,hi) (which was detected on Pi, Ti, Zi) onto the panoramic view:

[0119] Xpano? Ypano, Wpano, hp^o T(xi,yi,wi,hi)

[0120] xpano, ypano, wpano, hpano being the coordinates in the panoramic view.

[0121] The second step consists of grouping the N windows of the panoramic view into K groups, with K set by the user. In general, a PTZ camera has a maximum of 32 masks, the user can choose to use 30 masks and keep 2 masks for particular areas that he considers important to mask. From each group of windows, we form the mask which is a rectangle containing all the windows of the group. Each mask is therefore a rectangle (xm, ym, wm, hm). In [Fig.5] on the panoramic view, we clearly distinguish the masks which are frames grouping several detected windows. We can notably distinguish mask 17 grouping several windows of the image of [Fig.3].

[0122] The masks thus formed are then transmitted to the camera 10 during steps 3 and 4 in [Fig.l].

[0123] A post-processing step can be provided to correct masks to avoid masking areas of interest. Indeed, it happens that the grouping of windows hides part of the ground floor or part of the road. Intelligent processing makes it possible to modify the masks to avoid these situations. Additional elements as vehicles can be detected as seen in [Fig.3] and an algorithm can be applied to identify a road and thus review the grouping of windows into masks.

[0124] Operational phase.

[0125] If this has not been done at the end of the grouping phase, the start of the exploitation phase may include the transmission of the statically formed masks to the camera via a communication protocol with it. During this communication, the zoom level from which the mask is activated is calculated automatically from the proportion chosen by the user. For example, the user may wish the private area to be masked only if it exceeds 1 / 3 of the image. The zoom of the mask to be filled in is then calculated accordingly.

[0126] Thus in operation, when the user remotely controls the camera 10, blurred areas may appear depending on the position of the camera and the zoom level used. The masks have been automatically predetermined once and for all and stored in the camera.

[0127] According to the invention, a second masking mode can be provided, not static but on the fly and without any constraint on the maximum total number of masks.

[0128] In this on-the-fly mode, the objective is to form the masks on-the-fly, in real time during the exploitation phase. In [Fig.l], this amounts to performing steps 1-6 in real time on the fly.

[0129] Thus, when the camera is in a position (P, T, Z), the image is transmitted to the processing unit. In the on-the-fly mode, the windows can be pre-detected as below and saved in the database, but they can also be determined in real time.

[0130] If they have been predetermined, we project all the windows of the database onto this view with the same function T defined above.

[0131] If they have not been predetermined, they are detected in real time for the current image. In this case, the detection database is no longer essential. However, this approach may present constraints, because in a position (P,T,Z) all the windows are not necessarily detected, whereas they can be in another position (P',T',Z') with a different pan or tilt or zoom.

[0132] Preferably, each projected or detected window is considered as a mask to be applied.

[0133] Furthermore, as the mask can only be activated if the window exceeds a certain proportion of the image (for example 1 / 3), a situation achievable from a certain zoom, the number of masks can be much lower than 32 (maximum capacity of the camera). Therefore, each window can be considered as a mask which will be transmitted in real time to the camera to be immediately applied.

[0134] When the camera changes PTZ position, the masks are erased and new ones are formed with the same principle. The advantage of this approach compared to the static approach is the fact that the grouping of windows is avoided but the disadvantage is the frequent solicitation of the camera (frequent sending / erasing of masks).

[0135] According to the invention, it is possible to envisage combining the static and on-the-fly modes as follows: 1. A number Kl is reserved for operation in static mode (grouping of detected windows into Kl groups). 2. A K2 number is reserved for on-the-fly operation. 3. We reserve a number K3 of masks to be used manually to mask areas that are not necessarily windows.

[0136] With the constraint of K=K1+K2+K3= maximum number of masks authorized by the PTZ camera.

[0137] The interest of the hybrid approach is above all the correction of the static approach. Indeed, during the post-processing of the masks in order to let useful areas appear, windows can be removed and therefore not masked. The on-the-fly mode will make it possible to mask these windows.

[0138] In [Fig.6] we see an image displayed on the viewing screen with a blurred part corresponding to mask 17.

[0139] Of course, the invention is not limited to the examples which have just been described and numerous adjustments can be made to these examples without departing from the scope of the invention.

Claims

Claims

1. Method for statically masking areas of interest in images of a surveillance area, this method comprising the following steps: - during a static detection phase: - scanning the surveillance area and acquiring a plurality of images using a PTZ surveillance camera, - sending the plurality of images to a processing unit, - automatic detection, within the processing unit, of areas of interest visible in each image of the plurality of images, - during a static grouping phase: - automatic determination of one or more masks by: - formation of a panoramic view from a set of acquired images, - projection onto this panoramic view of all the areas of interest detected on the plurality of images, and - grouping of areas of interest into several groups taking into account a constraint of a maximum number of masks to be transmitted at once to the camera, and determination of one mask per group, - sending the mask(s) to the surveillance camera, - during an operating phase, when at least one area of ​​interest is visible on an image intended to be transmitted to a display screen, application within the camera of at least one mask covering this area of ​​interest so that this area of ​​interest is blurred in the image transmitted to the display screen.

2. Method for on-the-fly masking of areas of interest in images of a surveillance area, this method comprising the following steps carried out during an operating phase: - acquisition of an image of the surveillance area using a PTZ type surveillance camera, this acquired image being intended to be transmitted to a display screen, - sending the acquired image to a processing unit, - within the processing unit: - projection onto the acquired image of areas of interest previously detected for the same positioning of the camera, or - automatic detection within the processing unit of areas of interest in said at least one image, - during an on-the-fly grouping phase, grouping of areas of interest and automatic determination of one or more masks, each being defined by group of areas of interest, - sending of the mask(s) to the surveillance camera, - application within the camera of the determined mask(s) so that any masked part is blurred in the image transmitted to the display screen.

3. Method according to claim 2, characterized in that it comprises a preliminary step during which said previously detected areas of interest are obtained during a static detection phase comprising the following steps: - scanning of the surveillance area and acquisition of a plurality of images by means of the surveillance camera, - sending the plurality of images to the processing unit, - automatic detection, within the processing unit, of the areas of interest visible in each image of the plurality of images.

4. Method according to any one of the preceding claims, characterized in that the automatic detection step is based on a deep convolutional neural network.

5. Method according to claim 1 or 3, characterized in that during the static detection phase, for a first and a second different image having in common an area of ​​interest, if this area of ​​interest is detected in the first image and not in the second image, the processing unit is configured to apply a geometric transformation also making it possible to locate this area of ​​interest in the second image.

6. Method according to claim 1, characterized in that the panoramic view comprises images acquired for all P and T coordinates of the PTZ camera, but with a predefined fixed zoom.

7. Method according to claim 6, characterized in that the predefined fixed zoom is the lowest zoom, ideally a zoom equal to 1.

8. Method according to any one of claims 1, 6 or 7, characterized in that the projection of the areas of interest onto the panoramic view is carried out by applying a geometric transformation between the plurality of images acquired during the scanning and the images of the panoramic view.

9. Method according to any one of the preceding claims, characterized in that the grouping phase further comprises a post-processing step during which useful areas are detected in the plurality of acquired images and the determined mask(s) are modified so as not to cover these useful areas.

10. Method according to any one of the preceding claims, characterized in that the acquisition step consists of acquiring all the images of the entire surveillance zone.

11. A method according to any preceding claim, characterized in that a mask is a frame defining a portion of an image.

12. Method according to any one of the preceding claims, characterized in that the camera is configured to apply a mask to an image only when at least one area of ​​interest included in this mask has a predetermined size.

13. Method according to claim 12, characterized in that the predetermined size is defined such that the height of the area of ​​interest is greater than or equal to 1 / 3 of the height of the image.

14. Method according to any one of the preceding claims, characterized in that the areas of interest comprise elements among building windows, French windows, terraces.

15. Method according to any one of the preceding claims, characterized in that for a given image, the processing unit is configured to - carry out automatic detection of the areas of interest, - compare the areas thus detected with the areas of interest previously detected for the same image, - if there is a difference, setting up a northing process during which the camera is recalibrated relative to predetermined fixed points.

16. Method according to any one of the preceding claims, characterized in that the camera is mobile.

17. Method according to claim 16, characterized in that in the case of scanning, the movement of the camera is carried out by means of a robot or a drone.

18. Method according to claim 1 and any one of claims 6 to 17, characterized in that for a mobile camera, the step of projecting areas of interest onto the panoramic view is carried out by applying a geometric transformation between the plurality of images acquired during the scan and the images of the panoramic view, this geometric transformation taking into account the position of the camera.

19. System for masking areas of interest in images of a surveillance area, this system comprising: - a PTZ camera for image acquisition and application of masks on acquired images, - a processing unit configured to carry out a method according to any one of the preceding claims.

Images