PREDICTIVE ADAPTIVE LIGHTING CONTROL METHOD AND PREDICTIVE ADAPTIVE LIGHTING SYSTEM - Patent application
The predictive adaptive lighting system uses acoustic recognition AI to learn and predict traffic events, addressing inefficiencies in existing systems by reducing energy consumption and costs while enhancing illumination control.
Patent Information
- Application Number
- JP2022501142
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2019-07-10
- Filing Date
- 2020-07-03
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2040-07-03
AI Technical Summary
Existing lighting systems fail to efficiently adapt to traffic conditions due to the lack of accurate traffic data, leading to excessive energy consumption and inability to reduce lighting technology grades, and are costly and limited by visual recognition systems.
A predictive adaptive lighting system using acoustic recognition artificial intelligence to learn traffic data, segment and predict events, and control lighting settings based on acoustic and image sensing operations, reducing the need for costly cameras and improving predictability and responsiveness.
The system achieves energy savings by adaptively controlling lighting based on learned traffic data, overcoming visual limitations and reducing costs, enabling precise illumination adjustments and anticipation of vehicle movements and events.
Smart Images

Figure 0007593989000001 
Figure 0007593989000002 
Figure 0007593989000003
Abstract
Description
[Technical field]
[0001] The present invention relates to a method for predictive and adaptive lighting control, and to an associated lighting system. The invention has been made with particular reference to public lighting, and in particular road lighting, but does not exclude applicability to private lighting, for example in businesses, industries, shopping centres, homes, etc. The overall goal is energy efficiency. [Background technology]
[0002] In the patent literature, one of the most prominent documents in energy efficiency in lighting is EP2719258.
[0003] This document proposes modifying parameters of public lighting via current or forecasted weather data.
[0004] While this system is interesting, it does not take into account current methods and design regulations for lighting systems and does not appear to really aid in real energy savings, which should be based on the intensity of use of the area as well.
[0005] Current lighting engineering design criteria provide a first lighting sizing based on the shape of the area to be illuminated, thereby taking into account measurement parameters such as the length or width of the road.
[0006] Sizing is driven by varying based on risk variables, e.g. the presence of crossroads, corners, etc.
[0007] Finally, it is desirable to design taking into account real traffic variables, but data on these are not readily available.
[0008] In effect, a vigilance ratio is applied which tends to illuminate more than is actually needed.
[0009] Add to this the fact that in many countries there are road lighting technical classification laws and the systems light based on the associated lighting technical class. These classes are assigned in a conservative manner, often providing more lighting than necessary. Generally, rules for lowering the class are provided based on actual traffic conditions, but that data is difficult to obtain and often not available, so energy waste on roads to date has remained at unacceptable levels.
[0010] Another document in the art is US 2019 / 0008019, which describes a system for controlling the intensity of public lighting based on image recognition and image processing. US 2019 / 0008019 is aimed at a minimal communication system between lighting points. In particular, lighting points change their light intensity when motion is recognized in the field of view or when a change in the intensity of an adjacent street light is recognized. In this way, street lights communicate with adjacent street lights without the need for complex systems.
[0011] The system allows for lower levels of lighting to be kept when not needed, but cannot reduce the lighting technical rating originally assigned to the project because it cannot have traffic data.
[0012] The above systems have other drawbacks, in particular the cameras are costly, have limited detection zones, so that only a portion of the cameras are installed, and the cameras can be fooled if their view is obstructed, for example by trees that change foliage with the seasons, or road signs, etc. Furthermore, the described vision systems are only effective for detecting "motion" but cannot effectively detect stopping for conversations between people, which also requires lighting.
[0013] US 2019 / 0008019 also proposes an alternative system in which lighting is controlled via acoustic rather than visual detection. In this regard, the street lights are provided with microphones and the system has a control and communication unit that determines which street light is closest to the sound source and increases its intensity. This solution is therefore more complex than the previous one, but it also does not provide traffic data and therefore lacks predictability and the lighting technical grade initially assigned to the project cannot be reduced.
[0014] Finally, US Patent Application Publication No. 2019 / 0008019 proposes a hybrid between visual and acoustic solutions, where street lights are initially switched off and only acoustic detection is used to switch them on, activating the visual system, and then a decision is made to adjust or turn off the lighting units if other adjacent light sources are already providing enough light to the scene.
[0015] A more complex system, particularly one that is both adaptive and predictive, is described in US Patent Application Publication No. 2016 / 0050397.
[0016] The system is adaptive, as it aims to de-classify lighting technology, and predictive, as it adjusts the flow of light based on instantaneous predictions of events. To achieve these objectives, the system of U.S. Patent Application Publication No. 2016 / 0050397 utilizes visual recognition artificial intelligence, teachings that learn and classify traffic data based on images.
[0017] Overall, this system is very costly since the main component required is the camera. Moreover, its instantaneous predictive performance and reactivity based on images only is limited, because while the cameras are installed in a suitable way, the vision of the action field may not be optimal due to the directionality resulting in the overlap of obstacles or vehicles. Furthermore, as is known, the image is a light projection, which bounces off objects and only moves in a straight line. Therefore, it is not possible to recognize the image from an intersection, or behind a traffic light or a tree, with only one camera, and therefore it is not possible to predict, for example, the direction change of a vehicle. To eliminate this drawback, the number of cameras should be increased, which obviously increases the cost.
[0018] Hybridization with the system described in US Patent Application Publication No. 2019 / 0008019 would only result in artificial intelligence with all image-based learning, as in US Patent Application Publication No. 2016 / 0050397, where only audio is used to turn on the scene, allowing for vision at the scene. [Prior art documents] [Patent documents]
[0019] [Patent Document 1] US Patent Application Publication No. 2019 / 0008019 [Patent Document 2] US Patent Application Publication No. 2016 / 0050397 Summary of the Invention [Problem to be solved by the invention]
[0020] SUMMARY OF THE PRESENT EMBODIMENT It is an object of the present invention to overcome all or some of the drawbacks in the known art.
[0021] In particular, it is an overall goal of the present invention to increase energy savings in the field of lighting.
[0022] Another overall goal of the present invention is predictive adaptability, i.e. the ability to autonomously learn traffic data and use them to: -Adaptability objectives, such as the lighting technical rating assigned to the project, to lower lighting setting limitations; - predictability objectives, in order to predict local events and adapt the lighting to set limits; The present invention provides a process and illumination system that can be used for:
[0023] Yet another goal of the present invention is to improve instantaneous anticipation and responsiveness over known systems that use visual artificial intelligence.
[0024] Yet another goal of the present invention is to reduce the cost of predictive adaptive systems using visual artificial intelligence.
[0025] Yet another goal of the present invention is to allow adaptation of the lighting in anticipation of when the vehicle will turn or stop. [Means for solving the problem]
[0026] According to a first general aspect, the invention relates to a method for predictive adaptive lighting control, characterized in the following steps: - providing a controllable lighting system for a given area; -Providing acoustic recognition artificial intelligence; - carrying out an acoustic training step of an artificial intelligence with images, the artificial intelligence being taught to discriminate and classify traffic data from acoustic detection operations, characterized in that it learns by simultaneously making associations between acoustic detection operations and image detection operations in the same scene, preferably in at least one zone of said area, thereby generating a sample soundtrack (18); -After the step of acoustic training with images: a) a1) producing a historical traffic data record for the area based on acoustic detection activity compared to a sample soundtrack (18); a2) determining at least the lighting setting limits of the illuminated area based on historical data records; performing at least one adaptation step, comprising: b) b1) predicting local traffic events based at least on the sound detection performance compared to at least a sample sound track (18); b2) controlling the lighting system of the area to adapt to the predicted traffic event within set limits established by the adaptation step; performing at least one prediction step, comprising:
[0027] Advantageously, after training, the audio recognition artificial intelligence of the present invention allows detection operations in longer distances or in visually blocked areas. Moreover, this method has greater predictability, because it can distinguish various situations associated with the same image, thereby creating different sample sound tracks of the same image, which are distinguished from acoustic variability such as continuing straight or changing direction. This method also allows the collection of increased data for the visual artificial intelligence, because it can proceed with processing by comparison and association of audio detection operations after training, from which learning of situations that may elude image detection is obtained.
[0028] Advantageously, acoustic sensing artificial intelligence can be used to correlate vehicle traffic data with noise pollution levels and / or wear and tear on road substrates.
[0029] Yet another advantage is that it allows for automatic signaling of road-related abnormal events, such as, by way of non-limiting example, a car accident.
[0030] Preferably, the visual and acoustic detection zones are smaller than the above-mentioned areas. The visual and acoustic detection zones can be altered, for example moved, within the above-mentioned areas.
[0031] Preferably, the step of determining setting limits in the adaptation step comprises assigning a lighting technology class category to said area, and the step of controlling the lighting system in the prediction step comprises controlling said system based on said assigned class category.
[0032] According to a particular preferred embodiment, the acoustic training step is performed by simultaneously correlating localized acoustic and image sensing activity in one or more predefined sample zones 19. All of these sample zones 19 are preferably smaller than the area, while the acoustic sensing activity can be localized over the entire area in the step of creating the historical data record.
[0033] Advantageously, the overall cost of the system is particularly low because it uses a small number of image sensing means, e.g. cameras, and a large number of acoustic sensing means, e.g. microphones, the cost of the latter being significantly lower than the cost of the former.
[0034] "The entire zone is smaller than the above-mentioned area" means that the entire field of view of the image sensing means covers an area that is smaller than the entire illuminated area, or, as an additional alternative, that there are fewer image sensing means than acoustic sensing means.
[0035] Preferably, during the image acoustic training step, there are different sample zones, thus further reducing the number of image sensing means required and it is possible to use the same means for different zones, thus increasing the sample recording of the cases.
[0036] According to a particular preferred embodiment of the present invention, the image-based acoustic training step includes processing an image detection operation to identify and classify the objects represented, and transferring this classification to an acoustic detection operation to produce said soundtrack corresponding to said image detection operation.
[0037] For example, in general, the acoustic training step includes: - simultaneously detecting a film (96) and a spectrogram (98) of the scene captured on video; - performing, by visual identification software, visual identification and relative segmentation of at least a portion of the video; - creating segmented portions of the spectrogram, called sample soundtracks (18), by associating portions of the video with corresponding portions of the spectrogram and transferring this classification.
[0038] For example, this classification may identify the vehicle type and / or the type of activity the vehicle is performing.
[0039] Preferably, the correlation is made for those parts of the spectrogram that are above a predefined acoustic pressure threshold, called the silence threshold.
[0040] According to certain preferred embodiments, the method features the steps of predicting a traffic event (e.g. of a traffic path, such as continuing straight or changing direction) based on one or more sound detection actions compared to past records and / or a sample soundtrack, and controlling lighting of a zone in the area involved in the event and / or adjacent zones (e.g. of the expected traffic path) based on this prediction.
[0041] Preferably, the image acoustic training step includes processing the detection operation by correlating the acoustic and image data to obtain a sample soundtrack, characterized by at least one of the following information: - sample acoustic detection behavior characteristics of the passing of an object (object generally means a living being such as a vehicle or pedestrian); -Sample acoustic detection operating characteristics in the passage of two or more objects; - sample sound detection performance characteristics for a given number of objects (determined by learning how many objects belong to the detected sound); - sample acoustic detection behavior characteristics for a given type of object (e.g., car, truck, motorbike, or pedestrian); - Sample acoustic detection behavior characteristics at object stops (e.g. pedestrians having a conversation or a car stopping at a traffic light); - Sample acoustic detection behavior characteristics in a given direction that a given target is tracking (e.g., continuing in a straight line or changing direction).
[0042] According to certain preferred embodiments, generating the historical data record is based solely on acoustic sensing activity and includes at least the following self-learning steps: - cleaning the acoustic signal of ambient noise not belonging to vehicles or pedestrians; - determining the type of object being moved; - determining the speed of a moving object; - determining the direction of a moving object; - determining a change in direction of a moving object; - counting the moving objects in a given time interval.
[0043] Preferably, the act of counting moving objects includes more than one of the acts of clearing a traffic light, determining a type of object, determining a speed of an object, determining a direction of an object, and determining a change in direction of an object.
[0044] According to a particular preferred embodiment, the method includes any one of the preceding claims and is characterized in that it performs the following operations: - during an adaptation step, mapping the given area to be illuminated by dividing it into corresponding zones with lighting technical classes assigned based on at least one traffic characteristic learned by the artificial intelligence; - in the prediction step, an operation of correcting the initial lighting behavior of each zone based on the assigned lighting technical class, said correction including, for example, adjusting at least one of the following: attenuation, power, operating period, operating schedule, alternating operating periods, turning on or increasing brightness for prediction of a moving event, priority between zones, switch-off priority for controlled outages, demand in the presence of traffic and adjustment of light flow in real time.
[0045] Preferably, the method provides the determination by at least identifying vehicle movement from at least characteristic operating sounds, including sounds from engine noise, tire rolling noise, aerodynamic noise, artificially generated noise from hybrid and electric vehicles, or a combination of sounds, based on historical data and / or traffic forecasts.
[0046] Here, the characteristic motion sounds can be identified by using a frequency filter for separating frequencies between 500 and 4000 Hz, for example.
[0047] According to a second general aspect, the present invention relates to an adaptive and predictive lighting system comprising: - illuminated area (10, 12); - a plurality of lighting devices (15) distributed over the area and controllable to vary the illumination; -Sound recognition artificial intelligence (20) comprising the following a) and b): a) a plurality of intelligent devices (22) arranged in the area to be illuminated and each associated with one or several lighting devices which they control, each intelligent device (22) comprising, within predefined limits, a local neural learning network (21a) of acoustic-based traffic data for the associated lighting device, and acoustic detection means (25) interacting with the local neural network to predict local traffic events and to process subsequent local lighting commands; b) at least one general neural network (21b) separate from the intelligent devices (22) and in communication with each of them for receiving and processing the traffic data and determining relevant adaptive changes in the lighting setting limits of the lighting means (15), the artificial intelligence comprising means for transmitting said setting limits to the intelligent device (e.g. the general neural network determines the adaptation of the lighting technical class); The lighting system further comprises: - acoustic training means (32) of the artificial intelligence, preferably by temporary images, for example in a kit removable from the system, comprising image detection means (30) and sound detection means (25) at the same site and a neural training network (21c) cooperating therewith, for simultaneously correlating image and sound detection operations and generating a sample soundtrack (18).
[0048] Preferably the image sensing means is removable from the system.
[0049] Preferably, the image sensing means are arranged in a localized manner in one or more sample zones, the entirety of which is smaller than said region, for example the image sensing means can be moved in different zones of said region.
[0050] According to certain preferred embodiments, said adaptive changes in lighting setting limits include at least one of the following; - changing at least one lighting technical class pre-assigned to the area or its zone; and the local lighting command above is: - temporarily modifying the lighting of a zone of said area based on a prediction of a temporary traffic event predicted by the processing means by comparison of one or more detection operations with at least one of the following a) and b): a) the historical record of the acoustic detection behavior obtained after the training step; b) At least one sample soundtrack (18) obtained during the training step.
[0051] Also embodiments in which only properties are present to obtain adaptive or predictive functions are not excluded.
[0052] Further characteristics and advantages of the present invention will become more apparent from the following detailed description of preferred embodiments, given by way of indication and non-limiting example, with reference to the accompanying drawings, in which: [Brief description of the drawings]
[0053] [Figure 1] FIG. 1 illustrates a predictive adaptive lighting system according to the present invention. [Diagram 2] FIG. 2 is a diagram showing a histogram of the frequency composition of the sound emitted when a light vehicle passes by. [Diagram 3] FIG. 1 is a diagram showing a histogram of the frequency composition of sound emitted when a heavy vehicle passes by. [Figure 4] FIG. 1 shows successive frames of a video segmented based on the association of a sound track, called a spectrogram, with a visual track, specifically based on visual discrimination software. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0054] For the purposes of the present invention, the term "traffic" is conventionally considered as a comprehensive word, including the passage or stopping of any object requiring continuous illumination, such as living objects, such as animals, or inanimate objects, such as vehicles 85. Thus, "traffic" contemplates both outdoor movement, e.g., roads, and indoor movement, such as in a domestic environment. The illuminated areas that the system of the present invention concerns are therefore both public and private, both outdoors and indoors.
[0055] Referring to FIG. 1, an example and / or example of the use of the lighting system of the present invention is shown and generally designated by the reference numeral (1).
[0056] The system (1) comprises a surveillance area (5), in which for example various roads (10) and (12) exist.
[0057] The system (1) comprises a number of lighting devices (15), e.g. street lamps, distributed in an area, typically along a road, each controlled by a control means (16), e.g. varying adjustable light settings. Each control means (16) may be provided with any data network connection, e.g. a wireless network LPWAN, WIFI, 4G, 5G or equivalent cable network.
[0058] The system also includes an acoustic recognition artificial intelligence (20) for learning traffic data and determining associated lighting and / or lighting technology classification settings.
[0059] The artificial intelligence comprises a number of intelligent devices (22) arranged to cover a monitoring area (5). Each intelligent device comprises a local neural network (21a) for learning traffic data based on acoustics and for processing local lighting commands. The intelligent devices further comprise acoustic detection means, for example one or more microphones (25), interacting with the local neural network.
[0060] Preferably, the intelligent devices (22) are connected to the lighting devices, for example there is one intelligent device (22) for each lighting device (15). However, a less costly but in any case effective embodiment is possible, in which only a part of the lighting devices (15) is connected to the intelligent device (22) and thus provided with a microphone (25) which communicates with at least one local neural network (21a). In this last case, the remaining lighting devices (15) are preferably connected to a control means (16) to vary the lighting settings. They can be adjusted, for example, according to the instructions of at least one adjacent intelligent device (22).
[0061] The microphone, for example of the MEMS type or similar, is preferably integrated in the lighting device (15).
[0062] The artificial intelligence further comprises: - general purpose processing means, preferably separate from the intelligent device and realized for example via CLOUD computing means (32); -At least one general-purpose neural network (21b) associated with artificial intelligence.
[0063] Artificial intelligence also includes: - local processing means, preferably included in the intelligent device (22), for example realized by FOG / EDGE computing means (34), and associated to one or more local neural networks (21a).
[0064] In detail, each generic neural network (21b) makes it possible to process traffic data. The processed data comes from an intelligent device (22). It determines for example lighting technical classifications, or more generally lighting patterns, and transmits them to the intelligent device (22). The intelligent device can then determine local lighting settings. The generic neural network also receives training information for identifying the traffic data and transmits them to the intelligent device (22), as will be described below.
[0065] Through local processing means, the intelligent device determines local lighting settings within constraints set by the above information received from the general neural network.
[0066] For example, this communication may utilize a data network connection, for example a wireless network LPWAN, WIFI, 4G, 5G or an equivalent cable network, and the intelligent device may be preferably integrated into one of the control means of the lighting device, for example.
[0067] The artificial intelligence is also provided with training means (99), for example in a kit. Preferably the kit can be at least partially associated with and disassociated from the system (1) for temporary use.
[0068] In particular, the training means comprises a neural training network (21c), an acoustic detection means, and an image detection means (30), which cooperate with each other.
[0069] The sound detection means can be a microphone (25) permanently present in the system (1), such as a microphone of the intelligent device (22), or it can be a microphone belonging only to the training kit and therefore removable from the system and movable to different points.
[0070] The image sensing means (30) comprise, for example, one or several cameras, which are preferably movable to different points of the system and which can be associated and disassociated from the system, which belong solely to the training kit and which are preferably fewer in number than the microphones of the system.
[0071] As will be classified below, the image detection means are responsible for teaching the artificial intelligence to identify and classify traffic sounds. It is therefore sufficient for the image detection means to be set only at one or more representative points of the monitored area or to be moved between these points. At the end of the teaching process, they can also be removed from the system (1). The image detection means of the present invention therefore do not need to be spread over the entire network or to be permanently part of the network.
[0072] The acoustic recognition artificial intelligence (20) acquires acoustic data and transfers them into traffic data in the following way:
[0073] The monitored area is divided into zones, each corresponding to a road or a city block, for example, and is initially classified in the memory of the lighting system using the base attached lighting technology classification. This determines the initial lighting settings to be implemented, in particular the preferred settings to determine the schedule, duration, and the amount of applicable attenuation intended as a reduction in light flow. This initially assigned lighting technology class is also called the "project class".
[0074] In an initial learning step, the artificial intelligence (20) can be taught to identify vehicle types and / or traffic events from the acoustic spectrograms. This can be done by visual detection means, preferably by detecting images of moving vehicles only in one or several sample zones (19) in the area of the system. Various visual identification software exists on the market, which is able to process images to identify what they show and to segment them based on which vehicle or other vehicles are the subject of the image, for example by associating them with a vehicle. For the purposes of the present invention, any visual identification software and automatic segmentation of images can be used, including those already on the market. The images thus segmented are associated with the acoustic spectrum registered at the same time as the image, by transferring the image to an image segmentation, thereby generating a sample sound track (18).
[0075] Determining the sample soundtrack may be performed, for example, as shown in FIG.
[0076] A training means (99) simultaneously detects the film (96) and the spectrogram (98) of the videotaped scene.
[0077] The spectrogram in Figure 4 shows the frequency composition of a sound in Hz against time in seconds.
[0078] The training means then isolates the portion of the spectrogram whose frequency is above a predefined threshold (19), called the silence threshold. Above this threshold the system knows that something has occurred. The training means then associates this portion of the spectrogram with a corresponding portion of the video. This portion of the video is identified and segmented based on visual identification software, and the training means transfers this segmentation to the corresponding portion of the spectrogram, resulting in a sample soundtrack (18) that is a portion of the segmented spectrogram.
[0079] It is worth noting that visual and audio recording frequencies are different, whereby each frame of a video shoot corresponds to a range in the soundtrack.
[0080] The system can also recognize which vehicles are moving, on what frequency, etc., which can then generate statistical traffic data.
[0081] Once the system has been taught to classify sounds by associating classified images detected only in a given zone of the system, it is possible to proceed to acquire historical data using only sound detection operations in all areas of the neural network. For example, the image detection means 30 can be removed from the system and used in another similar system newly installed in a different area. They can in fact form a training kit and can be associated and disassociated from the system.
[0082] During the step of acquiring historical data, training data, which are sample tracks, are passed to the intelligent device, thereby enabling it to autonomously identify events based on acoustics and pass the associated data to a general purpose neural network (21b).
[0083] Once the first historical record of the traffic data detected by the acoustics trained with the segmented image is determined, the general neural network (21b) processes it and assigns the corrections to the designed lighting technology classification to each road, thereby determining the accurate zone mapping of the actual lighting needs according to the traffic. The artificial intelligence then sends the lighting parameters, such as attenuation and / or sound profile, to the intelligent device, which will be implemented based on the new classification.
[0084] Thanks to the received training data (18), the intelligent devices (22) are also able to autonomously generate instantaneous traffic behavior predictions and to adapt the performance of the lighting devices (15) within the lighting settings received from the artificial intelligence, in particular the lighting technical class assigned to those zones, according to the actual needs.
[0085] This system is more effective than systems with only visual recognition because, unlike light, sound can overcome obstacles. Indeed, even if you cannot see what is happening behind a corner, you can still recognize the noise. Furthermore, different noises can be associated with the same movement. For example, imagine a vehicle approaching an intersection, and the engine noise will be different if it continues going straight or if it changes direction. This generates discriminations that are not possible with visual recognition artificial intelligence alone, and generally leads to a pre-provided prediction of the vehicle's behavior continuity, thereby making it possible to control the lighting means on a route that includes around corners.
[0086] The system of the present invention can also recognize stationary noises, such as a running engine of a stopped vehicle or a conversation between two people, where the system controls appropriate lighting, which is not possible with visual recognition artificial intelligence alone.
[0087] The system can also be incrementally refined, as the neural network (21b) can continually acquire and process historical data, thereby correcting the lighting settings.
[0088] The training steps can also be improved, generating incremental updates of the training that are sent to the intelligent device (22).
[0089] The present invention is adapted to control any lighting variation, including intensity and color tone. For example, when a person is detected by footsteps or a voice, it may be desirable to change the color tone, for example switching from a cool light to a warm light.
[0090] FIG. 1 shows an example where a main road (10) intersects with a dead-end road (12).
[0091] These two roads initially have the same lighting technique classification, then the dead-end road (12) is reclassified and downgraded based on traffic sound detection performance, giving lighting priority to the main road (10).
[0092] From a practical operational point of view, the noise generated by a vehicle comes from various components, all of which can be identified by the system, in particular: 1. Combustion engine: points of compression, ignition and decompression. Sound emission levels depend on structural characteristics, rpm. 1a. Electric engine: A "whoosh" noise introduced by power control regulated by an actuator. Noises artificially generated by sound effect devices, adapted to announce the presence of vehicles below 20km / h. Related legislation, such as UN-ECE R138, has been introduced recently. 2. Rolling: noise caused by the trapping / releasing of air in the cavity between the tire tread and the asphalt. Depends on the tread design and the composition and porosity of the road surface. 3. Aerodynamics: Only relevant for speeds above 130km / h.
[0093] At low speeds, the noise generated by a motor vehicle engine is louder than the noise generated by the rolling of tires on asphalt. As speed increases, the rolling noise increases in intensity until it dominates the noise generated by the engine.
[0094] Therefore, measuring at the center around the acoustic contribution due to the rolling component of the sound allows the system to also accommodate the growing electric vehicle market and classify the road intended for any speed.
[0095] The centering of the frequency spectrum of the sound waves generated by light and heavy vehicles, e.g., Figures 2 and 3, respectively, is valuable in how it allows one to discriminate between noise sources (engine A, rolling B, aerodynamics, etc.) based on frequency. It is valuable to be able to filter the transformation of the signal into a specific frequency band, e.g., 500-4000 Hz, and thereby process only the acoustic contribution due to rolling tires.
[0096] This is useful, for example, for counting the number of moving vehicles or for determining periods of heavy traffic congestion.
[0097] For the purposes of understanding the present invention, the term "comprising" and its derivatives, as used herein, are intended to be open-ended terms that specify the presence of the mentioned properties, elements, components, groups, integers, and / or steps, but do not exclude the presence of other properties, elements, components, groups, integers, and / or steps that are not mentioned. The above also applies to words with similar meanings, such as the terms "comprised," "have," and their derivatives. Furthermore, the terms "part," "section," "portion," "member," or "element," when used in the singular, can have a dual meaning of a single part or multiple parts. For purposes of describing the above-mentioned implementation embodiments, the following directional terms "forward," "rearward," "upward," "downward," "vertical," "horizontal," "below," and "transverse," as well as any other similar directional terms, refer to the described embodiment in an operational position. Finally, terms of degree, such as "substantially," "about," and "generally," as used herein, refer to modified terms that deviate by a reasonable amount so as not to significantly change the end result.
[0098] Although only selected embodiments have been chosen to illustrate the present invention, it will be apparent to those skilled in the art from this description that various modifications and variations may be made without departing from the scope of the present invention as defined in the appended claims. For example, the size, shape, location or orientation of various components may be changed as necessary and / or desired. Components shown directly connected or in contact with each other may have intermediate structures disposed between them. The function of one element may be performed by two and vice versa. The structure and function of one embodiment may be adopted in another embodiment. All advantages of a particular embodiment need not necessarily be present at the same time. Any feature that is novel compared to the prior art should also be considered as a separate description of the applicant's other inventions, including the structural and / or functional concepts embodied by such features, either by itself or in combination with other features. Thus, the foregoing description of embodiments according to the present invention is provided for illustrative purposes only and not for the purpose of limiting the present invention, as defined by the appended claims and their equivalents.
Claims
1. 1. A method for predictive adaptive lighting control, comprising: Providing a controllable lighting system for a given area (10, 12); Providing an acoustic recognition artificial intelligence (20); performing an acoustic training step of an image-based artificial intelligence that is taught to identify and classify traffic data from acoustic detection operations, characterized in that the acoustic training learns by simultaneously correlating between acoustic detection operations and image detection operations of the same scene, thereby creating a sample soundtrack (18); After said step of acoustic training with images, a) a1) generating a historical traffic data record for the area based on at least sound detection activity compared to said sample soundtrack (18); a2) determining at least illumination setting limits for an illuminated area based on said historical data records; performing at least one adaptation step, comprising: b) b1) predicting local traffic events based on at least the sound detection activity compared to said sample soundtrack (18); b2) controlling the lighting system of the area to adapt to the predicted traffic event within set limits established by the adaptation step; performing at least one prediction step, comprising: A predictive adaptive lighting control method comprising:
2. 2. The method of claim 1, wherein the step of determining set limits in the adapting step includes assigning a lighting technology rating category to the area, and the step of controlling the lighting system in the predicting step includes controlling the system based on the assigned rating category.
3. 3. The method according to claim 1 or 2, characterized in that the acoustic training step by image is performed by correlating local acoustic and image detection operations in one or more predefined sample zones (19), the entirety of which is smaller than the area (10, 12), while the acoustic detection operations can be localized in the area (10, 12) in the step of creating the past data record.
4. 4. A method according to claim 3, characterized in that during the acoustic training step with images there are different sample zones (19).
5. A method according to any one of claims 1 to 4, characterised in that the step of acoustic training with images comprises the steps of processing the image detection operations, identifying and classifying them based on the objects represented, and transferring this classification to the acoustic detection operations to produce a sample soundtrack (18) corresponding to said image detection operations.
6. The acoustic training step comprises: Simultaneously detecting a film (96) and a spectrogram (98) of the videotaped scene; performing said visual identification and relative segmentation of at least a portion of the video with visual identification software; creating a segmented portion of the spectrogram, called a sample soundtrack (18), by associating a portion of the video with a corresponding portion of the spectrogram and transferring this segment; The method according to any one of claims 1 to 5, characterized in that
7. 7. The method according to claim 1, further comprising the steps of: predicting a traffic event based on one or more sound detection operations compared to the past records and / or the sample soundtrack (18); and controlling the lighting of a zone of the area involved in the event and adjacent zones based on the prediction.
8. The acoustic training step with images is characterized in that it includes a step of simultaneously processing the acoustic and image sensing operations of the same scene by correlating acoustic and image data to obtain a sample soundtrack (18), comprising the following information: Sample acoustic sensing operating characteristics upon passing of an object; sample acoustic sensing performance characteristics in passing two or more objects; sample acoustic sensing performance characteristics for a given number of objects; Sample acoustic sensing performance characteristics for a given type of object; Sample acoustic detection operating characteristics at target stop; A sample acoustic sensing operating characteristic in a given direction that a given target is tracking; The method according to any one of claims 1 to 7, characterized in that at least one of
9. Creating the historical data record is based solely on the acoustic sensing activity and includes at least the following learning steps: - cleaning the acoustic signal of ambient noise not belonging to vehicles or pedestrians; determining a type of moving object; determining a velocity of a moving object; determining a direction of a moving object; determining a change in orientation of a moving object; Counting moving objects during a given time interval; The method according to any one of claims 1 to 8, characterized in that it comprises:
10. 10. The method of claim 9, wherein the act of counting moving objects includes more than one of the acts of clearing a traffic light, determining a type of object, determining a speed of an object, determining a direction of an object, and determining a change in direction of an object.
11. In the adaptation step, mapping the predetermined area to be illuminated by dividing it into corresponding zones using lighting technique classes assigned based on at least one traffic characteristic learned by the artificial intelligence; - an operation of correcting the initial lighting behavior of each zone based on the assigned lighting technical class in the prediction step, said correction comprising adjusting at least one of the following: attenuation, power, operation period, operation schedule, alternation of operation periods, turning on or increasing brightness for anticipation of a moving event, priority between zones, switch-off priority for controlled outages, demand in the presence of traffic and adjustment of light flow in real time; 11. The method according to claim 1, further comprising carrying out the steps:
12. 12. The method according to any one of claims 1 to 11, characterized in that the historical data and / or traffic forecasts are determined by identifying at least vehicle movements from at least characteristic operating sounds of the vehicle, said characteristic operating sounds including sounds from engine noise, tire rolling noise, aerodynamic noise, artificially generated noise by hybrid and electric vehicles, or a combination of sounds.
13. A method according to claim 11, characterised in that the characteristic motion sounds are identified by a frequency filter, for example to isolate frequencies between 500 and 4000 Hz.
14. 1. A lighting system comprising: an area to be illuminated (10, 12); a plurality of lighting devices (15) distributed in said area and controllable to vary the lighting; An acoustic recognition artificial intelligence (20), a) a plurality of intelligent devices (22) arranged in said illuminated area, each associated with one or more lighting devices which they control, each comprising, for said associated lighting device, a local neural network (21a) of acoustic based traffic data, and acoustic sensing means (25) interacting with said local neural network, in order to predict local traffic events and, within predefined limits, to process subsequent local lighting commands; b) at least one general-purpose neural network (21b) remote from said intelligent devices (22) and in communication with each of them for receiving and processing traffic data and for determining relevant adaptive changes of the lighting setting limits of said lighting devices (15), said artificial intelligence comprising means for transmitting said setting limits to said intelligent devices; An acoustic recognition artificial intelligence (20) comprising: Equipped with The lighting system comprises: The lighting system further comprises an image-based acoustic training means (32) for the artificial intelligence, the image detection means (30) and the acoustic detection means (25) being at the same site, the neural training network (21c) cooperating therewith to simultaneously correlate the image detection operations and the acoustic detection operations to produce a sample soundtrack (18).
15. 15. The system according to claim 14, characterized in that said image sensing means (30) are removable from said system.
16. 16. System according to claim 14 or 15, characterized in that the image sensing means (30) are arranged in a localized manner in one or more sample zones (19) which are smaller overall than the area (10, 12).
17. The adaptive change of lighting setting limits may include: modifying at least one lighting technology class pre-assigned to the area or zone; At least one of The local lighting command may include one or more sound sensing operations; a) the historical record of acoustic detection behavior obtained after a training step; b) at least one sample soundtrack (18) obtained during said training step; temporarily modifying the lighting of a zone of the area based on a prediction of a transient traffic event predicted by the processing means by a comparison between at least one of A system according to any one of claims 14 to 16, characterized in that it comprises:
Citation Information
Patent Citations
Systems and methods for road acoustics and road video-feed based traffic estimation and prediction
US20120188102A1
Smart optical sensor for adaptive, predictive, and on-demand control of public lighting
US20160050397A1
Method of controlling a light intensity of a light source in a light network
US20190008019A1
Lighting system and method of controlling a lighting system
WO2018185132A1
Noise-flow monitoring and sound localization via intelligent lighting
WO2018210588A1
