Self-configuring and autonomous public lighting system

Abstract

A public lighting system comprising at least one public lighting device (10) suitable for installation on public roads, comprising at least one video sensor (12) providing a video stream composed of a series of digital images, said images representing an environment (120) of said device; and at least one lighting element (11) configurable in terms of luminous intensity and lighting area (110) according to parameters; and at least one digital processing element (20) comprising a first predictive model (21) configured to determine environmental information from said series of images, and a second predictive model (22) to determine presence information within the environment and configured to issue a command to modify the parameters of the lighting element according to the environmental and presence information. Figure for the abstract: Fig. 1

Description

Title of the invention: Self-configuring and autonomous public lighting device

[0001] FIELD OF THE INVENTION

[0002] The present invention relates to the field of public lighting. It relates in particular to a system comprising one or more public lighting devices suitable for installation on the roadway, for example in the form of street lamps.

[0003] CONTEXT OF THE INVENTION

[0004] It has been known for a long time to have lighting devices, or luminaires, on the roadway in order to provide lighting that provides safety and comfort to its users when natural lighting is insufficient, particularly at night.

[0005] Luminaires are proposed for different types of environments: urban, residential, road, etc., each of which may have specific characteristics. Different lighting qualities may be provided depending on the type of environment, in terms, in particular, of light intensity, lighting surface, lighting orientation, etc.

[0006] In addition, for a given type of luminaire, it may be interesting to configure certain luminaires in a specific way depending on their location on the roadway.

[0007] Another constraint lies in good energy management which imposes a balance between the safety and comfort provided by public lighting and the energy expenditure generated.

[0008] Recently, LED (Light Emitting Diode) technology has become established in public lighting due to its luminous efficiency but also for its control possibilities allowing the lighting to be adapted in time and space. This now mature technology is used in many recent luminaires and also allows an energy transition by converting old luminaires mainly equipped with discharge lamps.

[0009] Recent luminaires can include sensors to enable the lighting to be dynamically adapted according to certain measurements of their environment such as the presence of movement, traffic density or a degree of ambient lighting.

[0010] We can for example cite patent applications IN201941039625 (“A System and method for automatic Street lamp lighting and energy saving control”), CN109152185 (“An intelligent control System for multi-source sensing street lamps ), CN110913542 (“Street lamp controller, Street lamp control System and Street lamp control method”)...

[0011] Furthermore, recent state-of-the-art luminaires are highly configurable and can give, depending on the parameters chosen, different results. These parameters can affect the luminous flux, the light distribution, the colorimetry, the glare, etc.

[0012] It is therefore possible to provide general luminaires and to configure them differently depending on the location of their implementation. Such a method involves collecting and analyzing data from the scene to be illuminated (road, street in an urban environment, intersections, etc.) characterized by the type of roadway, the type and density of users, the dimensions, the light reflection properties of the illuminated surfaces, etc.

[0013] The collection and analysis of the project management's requirements (lighting rules adopted, consideration of lighting standards, specific requirements on color, lighting levels, uniformities, etc.)

[0014] A calculation of the predicted performances fed by the performances of all the luminaire configurations which will determine the best set of parameters allowing to provide the expected results.

[0015] Furthermore, the parameters thus determined must be taken into account by the entire chain of project stakeholders, from its design to the installation of each luminaire on the road. Furthermore, any modification requires the intervention of an operator on site in order to modify the parameters of the luminaire concerned, even if remote management mechanisms have been deployed.

[0016] SUMMARY OF THE INVENTION

[0017] The invention aims to improve the situation.

[0018] To this end, according to a first aspect, the present invention can be implemented by a system for public lighting comprising

[0019] - at least one public lighting device suitable for installation on the road, comprising at least one video sensor providing a video stream composed of a series of digital images, said images representing an environment of said device; and at least one lighting element configurable in light intensity and lighting zone according to parameters;

[0020] - at least one digital processing element comprising a first model predictive model configured to determine environmental information from said series of digital images, and a second predictive model to determine presence information within said environment from said series of digital images, and configured to issue at least one command to modify said parameters of said lighting element based on at least said environmental information and said presence information

[0021] According to preferred embodiments, the invention comprises one or more of the following features which can be used separately or in partial combination with each other or in total combination with each other:

[0022] - said digital processing element is embedded in each of said at least a public lighting device;

[0023] - said digital processing element is embedded in a lighting device public said master among said at least one public lighting device, the others of said at least one public lighting device being said slaves; said master public lighting device and said slave public lighting devices comprising communication elements configured for the transmission of said digital images and said at least one command;

[0024] - said digital processing element is deployed on a processing platform remote information, and wherein said at least one digital processing device and said public lighting devices comprise communication elements configured for the transmission of said digital images and said at least one command;

[0025] - said second predictive model is configured to determine said information of presence according to real-time constraints;

[0026] - said second predictive model is a multi-layer neural network of the type “ Yolo » ;

[0027] - said first predictive model is configured to determine at least one region in said environment, said environmental information comprising a respective class associated with said at least one region;

[0028] - said lighting element is configured to adopt light intensities variables within said lighting zone and said control is adapted to modify parameters of said lighting element so that said light intensity is adapted according to the respective class associated with said at least one region;

[0029] - said first predictive model is intended to detect a particular element at within said environment, and wherein said control is provided to modify parameters of said lighting element so as to cause a particular lighting quality around said particular element;

[0030] - said digital processing element is configured to determine a particular event and trigger the transmission of an alert message to a third-party server via a telecommunications network.

[0031] The invention therefore proposes a system for public lighting which is self-configuring.

[0032] It thus makes it possible to avoid the human effort of determining the parameters according to its location. The latter is in fact able to perceive its environment and automatically determine the parameters, without human intervention, in order to function correctly.

[0033] It also makes it possible to take into account possible changes in this environment without requiring human intervention on site.

[0034] Other characteristics and advantages of the invention will appear on reading the following description of a preferred embodiment of the invention, given by way of example and with reference to the appended drawings.

[0035] BRIEF DESCRIPTION OF THE FIGURES

[0036] The attached drawings illustrate the invention:

[0037] [Fig. 1a] illustrates a public lighting device according to one embodiment of the invention.

[0038] [Fig. 1b] illustrates a digital processing element according to one embodiment of the invention.

[0039] Figures 2a, 2b, 2c illustrate three possible embodiments for a system comprising at least one public lighting device and at least one digital processing element.

[0040] [Fig.3] illustrates a segmented digital image according to one embodiment of the invention.

[0041] DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

[0042] A public lighting device is a device intended and adapted to be placed on the roadway in order to provide lighting for the safety or comfort of users.

[0043] A public lighting device can be deployed in different types of locations, or environments. For example, it can be a road in a rural area, a highway, a road intersection, a street in an urban or peri-urban area, a parking lot, a cycle path, a pedestrian path, a tunnel, etc. Each of these locations, or environments, can be impacted by various characteristics, such as average traffic, natural lighting, the proximity of other types of artificial lighting, etc.

[0044] Depending on the environment, different types of users may be considered: cars, cyclists, pedestrians, etc.

[0045] A public lighting device can take different material forms, in particular depending on the environment.

[0046] Typically, a public lighting device may take the form of a lamppost, consisting of a vertical base at the top of which is arranged a downward-facing lighting element.

[0047] Other forms can however exist: wall lights, catenaries...

[0048] [Fig. 1a] illustrates a public lighting device 10, in accordance with one embodiment of the invention and having the form of a lamppost comprising a vertical support and a head oriented substantially downwards.

[0049] This head comprises at least one lighting element 11 and at least one video sensor 12.

[0050] The lighting element is configurable at least in intensity and lighting zone. It can also be configurable according to other characteristics, such as colorimetry. The lighting element is therefore associated with a set of parameters which are modifiable and allow its configuration.

[0051] The lighting zone 110 may be characterized in particular by an orientation and a surface of the lighting on the ground. According to one embodiment, the lighting zone may comprise several sub-zones which may be joined or disjointed. It is then possible to assign a distinct light intensity for each sub-zone.

[0052] The nature of the parameters enabling the lighting element to be effectively configured to obtain such results depends in particular on the technology of this element (angles, solid angles, LED identifiers in a matrix, etc.)

[0053] According to one embodiment of the invention, the lighting element is based on LED-based technology. Among different LED technologies, mention may in particular be made of:

[0054] - “High power LEDs”,

[0055] - “mid power LEDs” (medium power LEDs),

[0056] - COBs (“Chip On Board”),

[0057] - OLEDs (“Organic LEDs”),

[0058] - CSPs (“Chip Scale Packaging” or chip encapsulation package)...

[0059] The video sensor 12 may be compliant with different technologies.

[0060] Several sensors may be provided. For example, sensors may have different orientations and / or arrangements in order to increase the captured surface area and thus allow a better understanding of the environment of the device by the digital processing element(s). Also, sensors of different types may be provided in order to offer better robustness to different environmental conditions (for example night / day, clear weather, rain / fog, etc.). Thus, it is possible in particular to provide an infrared camera in order to better capture the night environment when the lighting element 11 is off or outside the lighting zone thereof.

[0061] Generally speaking, the video sensor(s) may be adapted to cover an environment 120 which encompasses the lighting zone 110.

[0062] The video sensors provide a video stream composed of a series of digital images. These digital images represent the environment 120.

[0063] The public lighting system according to the invention also comprises at least one digital processing element, one embodiment of which is illustrated in [Fig.lb].

[0064] A digital processing element 20 comprises at least two predictive models 21, 22.

[0065] A first predictive model 21 is configured to determine environmental information 42 from a series of digital images 41 provided to the digital processing element 20.

[0066] A second predictive model 22 is configured to determine presence information 43 within the environment 120 from this series of digital images 41.

[0067] Furthermore, the digital processing element 20 is configured to issue at least one command 44 from the environmental information 42 and the presence information 43. These commands are intended to modify at least some of the parameters of the lighting element 10.

[0068] Means 23 may be provided for determining this command from the information 42, 43 and transmitting it to the lighting element 11. These means may be, for example, a software module.

[0069] This software module as well as the first and second predictive models can be conventionally implemented on an information processing platform comprising a combination of electrical circuits such as microprocessors, memories and other specific circuits, and software modules (processor firmware, operating system, etc.)

[0070] The environmental information 42 concerns a static state of the environment 120 perceived by the video sensors 12.

[0071] This static state groups together characteristic data of the environment which are slightly variable over time. In particular, this static state is independent of the presence or absence of a user in the environment at a given time. In other words, this environmental information 42 is distinguished from the presence information 43 which is dynamic and dependent on the presence of a user at a given time.

[0072] The environmental information may include data relating to users but with the aim of deducing statistics therefrom: several digital images are taken into account in order to determine statistical data on traffic and the types of users frequenting the environment 120 (pedestrians, vehicles, etc.).

[0073] The environmental information may relate to the type of location in which the lighting device 10 is deployed, as well as various static characteristics of this location.

[0074] Examples of the type of locations to be detected include parking lots, cycle paths, bus shelters, electric vehicle charging stations, pedestrian crossings, roundabouts, etc.

[0075] This environmental information can therefore be determined from a digital image or several digital images (for example by taking an average of values ​​extracted from each digital image).

[0076] Once determined, the environmental information may be stored and used until it is updated. The update period may be a configurable parameter of the system 10, 20.

[0077] According to one embodiment, it can be considered that the environment is immutable and therefore that the environmental information is totally static and therefore does not need to be updated.

[0078] According to another embodiment, however, it can be considered that the environment is subject to changes (roadworks, changes in the destination of a neighborhood or a road, changes in user habits, opening or closing of businesses in the neighborhood, etc.). In such a case, the static nature of the environmental information must be understood as including fluctuations over the long term (i.e. on a time scale much greater than that relevant for presence information), for example beyond the day.

[0079] The presence information 43, on the contrary, has a dynamic character. It aims to represent the presence or absence of users in the environment 120 captured by the sensors 12. The presence information quickly loses its relevance since the users can move. It must therefore be regularly determined again and also be determined quickly, with a real-time constraint, in order to reflect reality at the time of availability of the presence information.

[0080] Figures 2a, 2b and 2c illustrate three possible implementations for a system for public lighting.

[0081] According to a first implementation illustrated by [Fig.2a], a digital processing element 201, 202, ... 20n is embedded in each of the public lighting devices, respectively 101, 102, ... 10n, where n is a number of public lighting devices.

[0082] Each public lighting device can then be autonomous, in the sense that the information collected (environmental 42, and presence 43) can be locally processed by the on-board digital processing element in order to provide a command to modify the parameters of the lighting element of this public lighting device.

[0083] According to a second implementation illustrated by [Fig.2b], a digital processing element 201 is embedded in a public lighting device 101 called master among. The other public lighting devices 102, ... are called slaves.

[0084] The public lighting devices comprise communication elements configured for the transmission of digital images and commands. The slave devices 102, ... 10n therefore comprise communication elements, respectively 302, ... 30n. The communication element of the master device is not shown (according to one embodiment, it may be a sub-element of the digital processing element 201).

[0085] The arrows represent the transmission of digital images from the slave devices to the master device.

[0086] Thus, the calculations of the slave devices are delegated to a master device. The system itself remains autonomous, in the sense that it does not require human intervention, the different slave devices exchanging the information they need in order to achieve the goal of the correct configuration of the lighting of each of the public lighting devices.

[0087] Obviously, a system may comprise several subsystems, each comprising a master device and a set of slave devices. In such a system, each slave device must be configured so as to be associated with a master device.

[0088] According to a third implementation illustrated by [Fig.2c], a digital processing element 20 is deployed on a remote information processing platform 200

[0089] This digital processing device 20 and the public lighting devices 101, 102, ... 10n comprise communication elements 301, 302, ... 30n configured for the transmission of said digital images and said at least one command. The remote information processing platform communication element 200 is not shown (according to one embodiment, it may be a sub-element of the digital processing element 20).

[0090] The arrows represent the transmission of digital images from the public lighting devices to the remote information processing platform 200.

[0091] According to one embodiment, this remote information processing platform may be a server accessible to public lighting devices via a telecommunications network, which may be wired or wireless. This server may, for example, be hosted and administered by the manager of the lighting device fleet.

[0092] According to another embodiment, the remote information processing platform may be a server accessible via a long-distance network such as the Internet. It may be a set of servers, for example organized as a farm. or "cluster." It can also be an abstract platform deployed on a shared cloud-type infrastructure, or "cloud."

[0093] The predictive models 21, 22 may be models based on machine learning (or “machine learning” according to the English terminology).

[0094] The first and second predictive models are distinct. They may be based on the same technology, but have different states. For example, in an embodiment in which the predictive models are learning-based models, training on distinct training sets may be sufficient for the resulting models to be distinct.

[0095] As described previously, the second predictive model 22 is provided to determine presence information within the environment 120 from the series of digital images provided by the video sensors 12.

[0096] According to a preferred embodiment, provision is made to determine this presence information according to real-time constraints, that is to say so as to minimize the time between a date of capture of a digital image and a date of provision of the presence information. As mentioned previously, it is important that at the time of provision of the presence information, this information is still relevant.

[0097] According to one embodiment of the invention, this second predictive model 22 is provided to determine, in a binary manner, whether a user is present or not in a digital image. According to another embodiment, it can be provided to determine a number and / or a type of user: for example, it can determine whether a user present is a pedestrian, a cyclist, a vehicle, etc. (and possibly, for each type of user, determine their number).

[0098] According to one embodiment, the second predictive model is a neural network. It may in particular be a convolutional multi-layer neural network, for example of the “Yolo” type.

[0099] In the case of a neural network, the state of the predictive model can be reflected by the values ​​of the synaptic weights. The training process aims to assign values ​​to the synaptic weights so that the model is able to predict a satisfactory output value from a new input. In the present case, once properly trained, the second predictive model will be able to predict the presence information from a submitted digital image (which did not belong to the training set).

[0100] To do this, a large number of digital images are associated with labels that indicate corresponding presence information. These digital images can be generated in a context similar to that of the operation of the public lighting device. This means in particular that they are generated from sensors identical and similar and from the same point of view: thus, the video sensors being classically arranged high up at the top of a lamppost and oriented downwards, the learning set must therefore be made up of digital images captured under these same conditions. Also, the digital images used for learning can be generated by digital simulation.

[0101] This learning set must preferably include a large diversity of images in order to obtain, in prediction (or inference), robustness and precision in determining presence information, whatever the shooting conditions (illumination conditions), the type of users, the clothing and the physical characteristics of pedestrians, cyclists, etc. (in particular the wearing of hats, umbrellas, etc.) ...

[0102] According to this embodiment, therefore, by learning, this second predictive model is configured to determine presence information from the digital images submitted to it.

[0103] The Yolo neural network was described in the article "You Only Look Once: Unified Real-Thread Object Detection" by Joseph Redmon, Santosh Divvala, Ross Girschick, and Ali Garhadi, 2016.

[0104] According to one embodiment, the Yolo v4 version can be used. This was proposed in the article “YOLOv4: Optimal Speed ​​and Accuracy of Object Detection” by Alexey Bochkovskiy, Chien-Yao Wang, Hong-Yuan Mark Liao, arXiv:2004.10934

[0105] Yolo is a convolutional neural network for object detection, built on the “Darknet53” architecture which is an image classifier consisting of 106 layers including 53 convolution layers.

[0106] One of the characteristics of Yolo is to propose a prediction of a region of interest (corresponding to a detected user) with several different resolutions: after a set of sub-sampling layers, the next layer provides a first detection at the lowest resolution. The following layers are over-sampling layers, and typically 2 layers among these allow new detections at increasing resolutions.

[0107] In version 4 of Yolo, two new concepts are introduced: “Bag of freebies” (BOF) and “Bag of Specials (BoS).

[0108] The BoF (which could be translated as "bag of freebies") impacts the learning strategy of the Yolo neural network. This can, for example, involve data augmentation methods, which automatically increase the diversity of input data in order to improve the robustness of the resulting predictive model.

[0109] The "bag of specialties" (BoS) concerns post-processing and methods that can increase the inference cost but aim to improve object detection.

[0110] The aforementioned article provides lists of examples of such "bags" (sections 2.2 and 2.3), as well as the choices made for the architecture of YOLOv4 (section 3.4).

[0111] One of the advantages of this YOLO model is that it allows for real-time detection: we can therefore submit digital images of the flow generated by the video sensors, so that the presence of a user is automatically detected.

[0112] The first model aims to determine environmental information. As mentioned previously, this environmental information evolves much less quickly, so that a real-time constraint does not weigh on this first predictive model. It is therefore possible to prioritize the quality of detection.

[0113] Here we can use a neural network of the HARD-Net type (Hardness AwaRe Discrimination Network). This type of network is notably described in the founding article by Tianjiao Li, Jun Liu, Wei Zhang, and Lingyu Duan, “HARD-Net (Hardness AwaRe Discrimination Network for 3D Early Activity Prediction” in Conference: European Conference on Computer Vision, November 2020 (DOI: 10.1007 / 978-3-030-58621-8_25)

[0114] According to one embodiment, this first predictive model is configured to determine at least one region in the perceived environment. The environmental information then comprises a class associated with each of these regions.

[0115] This step of determining regions in the perceived environment can be an image segmentation.

[0116] [Fig.3] illustrates a segmented digital image 100.

[0117] In this example, a first class 101 represents a road or a street, class 102 represents a sidewalk, class 103 represents a vegetated area.

[0118] According to one embodiment, this environmental information can make it possible to determine the lighting to be carried out.

[0119] In particular, the lighting element may be configured to adopt variable light intensities within the lighting zone 110. The command generated by the digital processing element 20 may be adapted to modify parameters of the lighting element 11 so that this light intensity is adapted according to the respective class associated with the detected regions.

[0120] Indeed, based on this segmentation, it is possible to have a sufficient understanding of the environment of the device and therefore to deduce a lighting to be achieved.

[0121] Thus, in this example, the class region 101 corresponds to an area whose users are usually vehicles. Since these have their own lighting systems, the device may only have to provide moderate lighting. On the other hand, since the region 102 corresponds to an area where the usual users are pedestrians, the lighting may have to be more intense. Finally, the region 103 is a priori an area where there are no users and therefore dimmer lighting can be provided, or even no lighting at all.

[0122] Depending on its training, the predictive model 21 can detect different classes of region within the digital images submitted to it during deployment. Examples include:

[0123] - Vehicle lanes (streets, roads, etc.)

[0124] - Pedestrian paths,

[0125] - Cycle paths or tracks

[0126] - Parking lots

[0127] - Parks and similar areas, etc.

[0128] According to one embodiment, this determination of distinct lighting per region can be automatically carried out by the predictive model 21, by carrying out training with a learning set in which labels indicating lighting to be achieved are associated with digital images.

[0129] According to one embodiment of the invention, the environmental information determines a maximum illumination to be achieved (for the entire lighting zone 110 or for regions of this zone, depending on the embodiments).

[0130] The presence information 22 can be used to modulate this lighting according to the presence.

[0131] According to one embodiment, the presence information can also be correlated with the detected regions. In other words, the predictive model 21 determines the location of the users in order to match them with the regions obtained by the environmental information.

[0132] Different regions are likely to receive different levels of illumination and uniformity. These levels can, for example, be determined in France by the NF EN 13201 standard, which indicates the requirements according to the class of public roadway.

[0133] The presence information can make it possible to activate the level required for the region concerned, whereas in the absence of a user, the lighting level can be zero or residual.

[0134] For example, the appearance of a vehicle in the “road” region 101 will cause moderate lighting associated with this region. The appearance of a pedestrian in the “sidewalk” region 102 will cause the strong lighting associated with this region.

[0135] It should be noted that the invention makes it possible to have both environmental information and presence information and that from these two types of information, it is possible to determine different possible lighting functionalities and configurations.

[0136] For example, according to one embodiment, the digital processing element may issue a command intended to modify a colorimetric parameter of the lighting element. It is thus possible to provide distinct lighting colors depending on the class of regions, or, more generally, environmental information.

[0137] For example, one can generate cool white lighting for a region dedicated to vehicles and warmer white lighting for a region dedicated to pedestrians.

[0138] According to one embodiment of the invention, this first predictive model 21 is designed to detect a particular element within its environment 120. It is then possible to generate a command designed to modify parameters of the lighting element so as to cause a particular lighting quality around this particular element.

[0139] The particular element may, for example, be an element of street furniture: bus shelter, public bench, etc. A particularly interesting example is a charging station for an electric vehicle.

[0140] The particular lighting quality may consist of light intensity (which may be independent of presence information), colorimetry, etc.

[0141] According to one embodiment, public lighting can thus specifically illuminate certain elements of the roadway, automatically.

[0142] For example, certain elements may be illuminated even in the absence of users, so that, at night, they can be seen from a distance (for example, a bus shelter). Similarly, specific colorimetry may allow these elements to be better readable for a distant user (bus shelter, taxi area, etc.)

[0143] In [Fig.3], region 105 may correspond to a public bench located within the vegetated region 103.

[0144] Furthermore, according to one embodiment of the invention, the environmental information comprises characteristics linked to the reflection properties of the detected environment.

[0145] In particular, different characteristics may be determined by regions of the environment. In particular also, these characteristics may represent the ground luminance of the regions (or of the region).

[0146] The luminance of the roadway depends on its reflection properties. These are defined, and taken into account in the calculations in a standardized manner, by matrices established in the 1960s. They are no longer suitable for current surfaces and therefore the calculation error is significant between this model and reality.

[0147] According to one embodiment of the invention, the image of the roadway taken by the camera can be analyzed to determine the actual luminance level.

[0148] The performance of the lighting system will therefore be adapted to the reality of the road surface and even to its evolution over time. Work on measurements of road surfaces, notably carried out in France by CEREMA, shows that the clarity of the surfaces has increased. This will therefore lead to a reduction in power levels to achieve the expected result (and therefore to energy savings).

[0149] The parameters of the lighting element which can be modified by the command issued by the digital processing element can be of different natures and depend strongly on the technology and the possibilities of the lighting element.

[0150] As possible examples, which can be considered alone or in combination:

[0151] - the lighting element can integrate several different optical subsystems each controlled by an intensity. Optimization software will adjust the linear combination of the intensities of the different systems. The modified parameters can thus be these intensities whose values ​​make it possible to modulate the particular lighting of each subsystem and therefore of each zone associated with these subsystems.

[0152] - the lighting element comprises a movable LED source. A modifiable parameter may be its displacement relative to a fixed optic, thus allowing the light distribution to be modified.

[0153] - the lighting element may comprise a matrix of small optical systems, Each of these systems is capable of illuminating an area of ​​the space. Depending on the area to be illuminated, each of these mini-systems is activated, or not, or partially.

[0154] - the lighting element may comprise flexible elements. The parameters can then concern the deformation to be imposed on them by mechanical action (pressure, extension, etc.) in order to modify the distribution of light intensity.

[0155] - the parameters may relate to optical properties of the optical system of the lighting element, by electrical control,

[0156] etc.

[0157] According to one embodiment, the digital processing element 20 is configured to determine a particular event within the digital images.

[0158] Such an event may be a modification of the regions resulting from the segmentation of the digital images, for example, or a change in the texture of a region, etc. It is thus possible, in particular, to detect deterioration of the road (change in texture, encroachment of a grassy area onto the roadway, etc.)

[0159] If such an event is detected, an alert message can be transmitted to a third-party server via a telecommunications network. In this way, road services or any other organization can be informed and act as quickly as possible.

[0160] The device of the invention can thus make it possible to resolve additional problems automatically, without human intervention, linked to the perceived environment 120.

[0161] The system according to the invention can thus be seen as a platform offering basic services on which new application functions can be deployed. These new functions can be deployed by simple downloading and / or updating of the software means implemented by the digital processing element (whether it is embedded within the public lighting device itself or remote).

[0162] Of course, the present invention is not limited to the examples and the embodiment described and shown, but is defined by the claims. It is in particular susceptible of numerous variants accessible to those skilled in the art and is defined solely by the content of the claims.

Claims

1. Claims System for public lighting comprising at least one public lighting device (10) adapted for installation on the roadway, comprising at least one video sensor (12) providing a video stream composed of a series of digital images, said images representing an environment (120) of said device; and at least one lighting element (11) configurable in light intensity and in lighting zone (110) according to parameters; at least one digital processing element (20) comprising a first predictive model (21) configured to determine environmental information from said series of digital images, and a second predictive model (22) to determine presence information within said environment from said series of digital images, said presence information being intended to represent the presence or absence of at least one user in the environment at a given time, said at least one first predictive model being further configured to: - determining at least one region in said environment, said environmental information comprising a respective class associated with said at least one region, and, - when at least one user has been detected in the environment, determining a location of said at least one user and correlating it with said at least one detected region, and said at least one digital processing element (20) being configured to issue at least one command to modify said parameters of said lighting element as a function of at least said environmental information and said presence information, said lighting element being configured to adopt variable light intensities within said lighting zone, the environmental information being used to determine a required lighting level for said at least one detected region in said lighting zone based on said class associated with said at least one region, and the presence information being used to modulate said lighting level in said at least one region, said lighting level required being activated when at least one user has been detected in said at least one region.

2. The system of claim 1, wherein said digital processing element is embedded in each of said at least one public lighting device.

3. System according to claim 1, wherein said digital processing element is embedded in a public lighting device called master among said at least one public lighting device, the others of said at least one public lighting device being called slaves; said master public lighting device and said slave public lighting devices comprising communication elements configured for the transmission of said digital images and said at least one command.

4. The system of claim 1, wherein said digital processing element is deployed on a remote information processing platform, and wherein said at least one digital processing element and said public lighting devices comprise communication elements configured for the transmission of said digital images and said at least one command.

5. System according to one of the preceding claims, wherein said second predictive model is configured to determine said presence information according to real-time constraints.

6. [System according to the preceding claim, wherein said second predictive model is a multi-layer neural network of the “Yolo” type.

7. The system of claim 1, wherein said first predictive model is adapted to detect a particular element within said environment, and wherein said control is adapted to modify parameters of said lighting element so as to cause a particular lighting quality around said particular element.

8. System according to one of the preceding claims, in which said digital processing element is configured to determine a particular event and trigger the transmission of an alert message to a third-party server via a telecommunications network.