Lighting fixtures with precipitation estimation capabilities

Abstract

A lighting fixture has a precipitation sensor for detecting precipitation at its location based on detected sound. The precipitation sensor receives information from other lighting fixtures in the vicinity of the lighting fixture, and the lighting fixture and the other nearby lighting fixtures form a set of lighting fixtures. The sound detected at the lighting fixture and the received information are combined to derive a precipitation estimate. In particular, a subset of the set of lighting fixtures is selected, and the information associated with that subset is used to derive the precipitation estimate.

Description

Technical Field

[0001] The present invention relates to lighting fixtures, such as road lighting fixtures, and more particularly to the detection of rain or other precipitation using lighting fixtures.

Background Art

[0002] It is known that the intensity of heavy rain reduces visibility and causes many traffic accidents. It is desirable to be able to detect rain with high granularity (i.e., hyperlocal detection) so that countermeasures can be taken according to the local rain situation. For this purpose, it is particularly interesting that existing infrastructure, such as street lighting infrastructure, can be used.

[0003] Various precipitation sensing techniques are known, including tipping-bucket rain gauges, optical (infrared) rain gauges (commonly used on the windshields of vehicles), and capacitive sensors.

[0004] It is also known to use a microphone that measures the rainfall intensity based on the sound produced when rain hits the sensor. For example, WO 2016 / 156563 discloses the detection of precipitation in the outdoor environment of a lighting fixture. The lighting fixture has a precipitation detection module configured to convert an accelerometer output signal and / or a microphone output signal into a frequency domain waveform by executing a fast Fourier transform algorithm. Once the frequency domain representation of the accelerometer output signal and / or the microphone output signal is obtained, the frequency domain waveform is compared with reference precipitation information to detect the occurrence of precipitation in the outdoor environment of the lighting fixture. The stored reference precipitation information includes prior information regarding one or more frequency ranges indicative of precipitation and associated amplitude levels.

[0005] For streetlights, the most obvious architecture is to use an outdoor sensor bundle with a microphone mounted on the bottom of the light fixture, using audio data to detect the sound of raindrops on top of the fixture. However, tree branches above the streetlights strongly affect the amount of rain that falls on the top of the light fixture housing. Leaves cause a delay from the start of the rain until raindrops fall on the light fixture. Similarly, after the rain has stopped, leaves continue to drip water onto the light fixture. Compared to unobstructed rain, the presence of leaves also changes the number of raindrops that fall on the light fixture housing and the size of each individual raindrop. [Overview of the project] [Problems that the invention aims to solve]

[0006] This problem could potentially be addressed by calibrating the audio-based rain detection algorithms for individual light fixtures to account for rain-shielding by different light fixture housings, tree branches, and other objects. However, calibrating the audio-response curve for rainfall is cumbersome, and the calibration parameters are highly sensitive to changes in the streetlight environment (e.g., as the tree canopy above the streetlight changes seasonally, or when overhanging branches of trees are regularly pruned to protect power lines, for example).

[0007] There is a need for improved rain (or other precipitation) sensing approaches. [Means for solving the problem]

[0008] The present invention is defined by the claims.

[0009] According to one aspect of the present invention, a lighting fixture is provided, Light source and A precipitation sensor for detecting precipitation at the location of the lighting fixture, Includes, The precipitation sensor is, A microphone for detecting sound in the lighting fixture, An input unit for receiving information derived from sound detected in other lighting fixtures in the vicinity of the lighting fixture, wherein the lighting fixture and the other nearby lighting fixtures form a set of lighting fixtures, An output unit for sending information derived from the sound detected in the lighting fixture to other lighting fixtures in the vicinity of the lighting fixture, A processor adapted to process the sound and information detected by the lighting fixture and derive a precipitation estimate from it, The processor is provided with luminaires, which include a processor adapted to use information related to a subset of luminaires to derive precipitation estimates.

[0010] This lighting fixture measures precipitation levels, such as rainfall. Precipitation levels can be used to control lighting as a function of weather conditions; for example, if the lighting fixture is part of a streetlamp, it can provide different lighting for different road conditions.

[0011] The system of the present invention estimates precipitation levels using a subset of lighting fixtures, rather than simply combining data from individual fixtures or from sets of fixtures. Instead, an intelligent selection of lighting fixtures is made to contribute to an overall precipitation level estimate. The lighting fixtures in the set from which the subset is selected are located in the same general vicinity, i.e., are expected to experience the same precipitation levels. The set is, for example, the lighting fixtures of streetlights along a particular road. The subset is, for example, the lighting fixtures that are least likely to be obstructed by tree branches or leaves, have minimal exposure to audio noise (other than rain), and therefore give the best estimate of precipitation levels.

[0012] The subset is determined, for example, by the installer during installation.

[0013] However, the system (i.e., the processor) itself may learn which luminaires should be in the subset. Furthermore, the subset may be dynamically changed over time. For example, the precipitation level estimated by a luminaire covered with leaves may differ from the precipitation level estimated by other luminaires. Thus, the system itself can infer which luminaires are best selected to form the subset.

[0014] Sensor hardware is typically already installed in lighting fixtures. For example, lighting fixtures equipped with radar sensors, tilt sensors, vibration sensors, microphones, and temperature sensors are known. These sensors are collectively known as a sensor bundle.

[0015] Unsupervised or weakly supervised machine learning may be used to analyze the audio signals of a set of lighting fixtures located in close proximity to each other, the lighting fixtures in the set may pass rainfall-audio-related messages to each other.

[0016] Precipitation levels are determined, for example, based on the audio effect of water droplets on the housing of a lighting fixture. For instance, the processor for each lighting fixture applies a self-learning algorithm to estimate the precipitation level, and this learning takes into account the specific design of the lighting fixture.

[0017] A precipitation sensor is a single-design sensor bundle that can be used with many different types of lighting fixtures, for example, those with different surface areas exposed to rain, or those with different audio responses when raindrops hit the upper surface. Thus, as a result of the self-learning process, different processing algorithms will be used for different lighting fixture designs.

[0018] If the subset is fixed, for example, at installation, each light fixture within the subset will provide a precipitation estimate. If the processor can determine and dynamically adapt the subset, all nearby light fixtures (whether within the subset or not) may provide their own local precipitation estimates.

[0019] A subset is selected, for example, from lighting fixtures that are not blocked by leaves, bridges, tunnels, etc. A master lighting fixture is not required, but one lighting fixture may initiate message passing and be the last lighting fixture to receive messages passed back from its neighboring fixtures. Preferably, each lighting fixture has the capability to estimate the overall precipitation level by processing input from its neighboring fixtures.

[0020] For example, precipitation estimation is, Occurrence of precipitation Intensity of precipitation, Precipitation droplet size, This includes one or more estimations of the following.

[0021] Thus, the probability of precipitation, as well as the type and intensity of precipitation, may be estimated.

[0022] For example, the processor is adapted to select a time-varying subset. For instance, the subset may change between summer and winter depending on the level of foliage. Since branches also grow over time and are sometimes pruned, a subset of the most suitable lighting fixtures may be selected. If the branches above a streetlamp are pruned, that streetlamp may become more suitable for detecting rain.

[0023] The time-varying subset may have a different number of members at different times. For example, in the leafless winter, fewer lighting fixtures may be used in the sensing group to save connection costs (e.g., cellular charges for data transmission from each lighting fixture to the cloud, etc.).

[0024] For example, the precipitation sensor may further include a vibration sensor.

[0025] This may be used to detect road usage and thus to estimate road noise, and noise cancellation may be used.

[0026] For example, the precipitation sensor may further include a wind speed sensor. This may be used to enable cancellation of wind-related noise.

[0027] For example, the precipitation sensor may further include a temperature sensor. This may be used to detect weather conditions and thus to enable precipitation to be detected by additional sensing, for example, after a rapid temperature drop.

[0028] For example, the precipitation sensor may further include a humidity sensor. Humidity changes occur with weather changes, and time-series humidity information may enable precipitation to be detected.

[0029] For example, the precipitation sensor may further include a pressure sensor. Air pressure changes occur with weather changes, and time-series pressure information may enable precipitation and precipitation type to be detected.

[0030] For example, the precipitation sensor may further include a daylight sensor. The daylight sensor can be used to recognize whether a streetlight is overgrown by a tree (e.g., when the streetlight is in the sun in winter and in the shade at other times).

[0031] For example, a precipitation sensor may further include a radar sensor. Radar sensing can be used to detect the presence and distance of trees or other obstacles.

[0032] For example, a precipitation sensor may also include an RF sensor. RF sensing can be used to detect atmospheric conditions, and it can also be used to detect tree branches.

[0033] Lighting fixtures may further include wireless systems for sending and receiving information from other lighting fixtures. Bluetooth® Low Energy (BLE) mesh networks are already available in known urban lighting systems. However, other wireless communication systems may be used.

[0034] For example, the information received and transmitted includes a precipitation level probability distribution, which is the probability of precipitation based on the sensed conditions.

[0035] For example, a subset of lighting fixtures is defined for each lighting fixture as follows: Obstruction above by vegetation, Obstruction above by man-made structures Ambient non-precipitation noise level It is selected based on one or more of the following.

[0036] In one example, the precipitation sensor includes a radar sensor, and the processor is adapted to select a subset of lighting fixtures based on the proximity of trees or branches determined by the radar sensor.

[0037] For example, the processor is adapted to derive the shared / fused precipitation level for an area of ​​a set of lighting fixtures. For example, multiple probabilities are combined using any appropriate statistical approach, such as Bayesian fusion, to combine them into a better estimate.

[0038] Thus, multiple lighting fixtures within an area are used to estimate the shared precipitation level for that area. The message-passing approach reduces the effect of a single lighting fixture being (partially) blocked by tree branches, or the effect of another lighting fixture being located near a (building) structure that partially shields raindrops. In this case, the lighting fixtures in the set may determine the maximum likelihood rain level for the area.

[0039] For example, the processor may be adapted to cluster a set of lighting fixtures into different types of fixtures. In this way, precipitation levels may be estimated using different algorithms, each tailored to different lighting fixture types and / or mounting orientations.

[0040] For example, precipitation estimation is the estimated rainfall and / or raindrop size.

[0041] The present invention also, Paul and, The lighting fixture as defined above, wherein the lighting fixture is attached to the pole, It provides streetlights, including those included.

[0042] These and other aspects of the present invention will become apparent and clarified by reference to the embodiments described below. [Brief explanation of the drawing]

[0043] For a better understanding of the present invention and to more clearly illustrate how the present invention may be carried out, the accompanying drawings are referred to only as examples. [Figure 1] This shows a lighting fixture with a precipitation sensing function. [Figure 2] This demonstrates how lighting fixtures can sense their local precipitation levels. [Figure 3] This document describes a method for a lighting fixture to derive an overall precipitation level based on monitoring by multiple nearby lighting fixtures. [Modes for carrying out the invention]

[0044] The present invention will be described with reference to the figures.

[0045] The detailed descriptions and specific examples illustrate exemplary embodiments of the apparatus, systems, and methods, but should be understood to be for illustrative purposes only and not intended to limit the scope of the invention. These and other features, aspects, and advantages of the apparatus, systems, and methods of the invention will be better understood from the following description, the appended claims, and the appended drawings. Please understand that the drawings are for illustrative purposes only and are not drawn to scale. Please also understand that the same reference numerals are used to indicate the same or similar parts throughout the drawings.

[0046] The present invention provides a lighting fixture having a precipitation sensor for detecting precipitation at the location of the lighting fixture based on detected sound. The precipitation sensor receives information from other lighting fixtures in the vicinity of the lighting fixture, and the lighting fixture and the other nearby lighting fixtures form a set of lighting fixtures. The sound detected at the lighting fixture and the received information are combined to derive a precipitation estimate. In particular, a subset of the set of lighting fixtures is selected, and the information associated with that subset is used to derive a precipitation estimate.

[0047] Thus, instead of focusing on (looking at) individual light fixtures, the audio signals of groups of light fixtures located in direct proximity to each other are analyzed, and these groups of light fixtures pass (pass) rainfall audio-related messages to one another. The radius of the neighborhood is such that a large number of light fixtures experience similar rainfall amounts. (Based on selection criteria,) a subset of light fixtures is automatically selected that has the lowest probability of being obstructed by tree branches / leaves and therefore gives the best estimate of rainfall in that neighborhood. Since branches grow over time and are sometimes pruned, it is preferable that the algorithm can dynamically select the currently optimal subset of light fixtures with an appropriate outdoor sensor bundle. For example, different subsets of light fixtures may be used in winter, summer, and spring.

[0048] Figure 1 shows a lighting fixture 10 that includes a light source 12 such as an LED configuration and a precipitation sensor for detecting precipitation at the location of the lighting fixture. The precipitation sensor may have one or more sensing modalities, as described later.

[0049] At a minimum, the precipitation sensor includes a microphone 14 for detecting sound in the lighting fixture. The microphone is an existing sensor in an existing lighting fixture's sensor bundle, for example, so that precipitation sensing can be added to the existing infrastructure. An existing street light with a sensor bundle has sufficient capabilities for the ultra-local detection of rainfall. The present invention utilizes selection criteria for purposefully choosing a subset of lighting fixtures to be used for inferring rainfall intensity at a given time.

[0050] Localized sound can be interpreted to determine the localized probability of precipitation sensed by the luminaire. However, the precipitation sensor also includes an input unit for receiving information derived from sound detected by other luminaires in the vicinity of the luminaire. This input unit is represented by a receiver 16. There is also an output unit for sending information derived from sound detected by the luminaire to the aforementioned other luminaires in the vicinity of the luminaire. This output unit is represented by a transmitter 18.

[0051] In this example, the light fixture is part of a streetlamp. This light fixture and other nearby light fixtures form a set of light fixtures. This set is a geographically related set of light fixtures, such as all the light fixtures along a particular road or section of road.

[0052] The processor 20 is adapted to process locally detected sounds and received information at the lighting fixture and derive a precipitation estimate from them. In particular, the processor uses information related to a subset of lighting fixtures to derive the final precipitation estimate.

[0053] This invention utilizes a four-step approach to estimate rainfall levels based on an audio sensor and optionally a vibration sensor, which are incorporated into a lighting fixture.

[0054] Figure 2 shows the method used to derive the local probability of precipitation for each individual lighting fixture.

[0055] In step 30, sensor signals that enable rain detection are collected.

[0056] In step 32, a data preprocessing algorithm is applied to the signals from each sensor bundle. A classification algorithm detects whether the detected sound is due to raindrops. This preprocessing algorithm is similar, for example, to Voice Activity Detection (VAD) in speech recognition. For example, filters are applied to reduce traffic noise caused by gusts of wind or passing cars or trucks. Each sensor bundle may also include a vibration sensor, which allows for correlation analysis between time-series audio events and time-series vibrations associated with gusts of wind and passing trucks. Raindrops do not cause significant vibrations in the lamp head or pole. Therefore, the use of a vibration sensor in the sensor bundle allows for filtering out non-rain-related audio noise. This data preprocessing is possible if training data has been collected for a specific lamp installation.

[0057] The lighting fixtures present in the set may be clustered by lighting fixture type. The exact lighting fixture types may be available as data forming part of the lighting system network, or they may be obtained from satellite imagery or Google Street View imagery.

[0058] Urban lighting systems can have many different types of lighting fixtures (with varying surface areas exposed to rain, and significantly different audio responses when raindrops hit the top surface of the fixtures).

[0059] If it is not possible to select lighting fixtures from exactly the same manufacturer (for example, similar housing material, similar size of the top surface of the lighting fixture, similar structure of the top surface of the lighting fixture, etc.), different lighting fixture manufacturers with similar audio response characteristics to rain may be selected.

[0060] Conversely, in more advanced embodiments, two (or more) lighting fixture manufacturers with very different audio responses to rain may be selected. This may allow the rain classification algorithm to focus on different rain-related audio features. For example, a first lighting fixture housing may produce an audio sound that can detect even very light rain (however, the first lighting fixture housing may produce excessive noise to detect large raindrops, especially if many raindrops fall on the fixture per second). However, the first lighting fixture housing may be suitable for counting a small number of individual heavy raindrops falling on the fixture per second.

[0061] The second lighting fixture housing may generate audio features suitable for classifying heavy rain.

[0062] In such cases, a machine learning algorithm may be used to capture rain-related audio features from both the first and second lighting fixture housings, thereby enabling the detection of rain occurrence, rain intensity, and rain characteristics.

[0063] Similarly, when selecting lighting fixtures, it is advantageous to ensure that each lighting fixture type is represented by multiple fixtures within the same ultra-local area. This allows rain inference to be performed by combining raw audio data from multiple fixtures of the same lighting fixture hardware design within a lighting fixture housing.

[0064] For example, k-means clustering may first be used to identify the number of luminaire classes present in this particular streetlight installation from rain audio data (and optionally from radar / RF sensing sensor data) collected by various sensor bundles. Then, k-means clustering may be used to segment streetlights of a single luminaire type into different levels of vegetation shielding (e.g., a first luminaire class is not covered by tree branches, a second luminaire class is partially covered by branches, a third luminaire class is completely covered by branches, etc.). Classes may include different current densities of leaf canopy.

[0065] In step 34, a self-learning algorithm is applied to estimate the rainfall level for each of the lighting fixture types present in this set. Similar to acoustic event detection, a masked audio spectrum transformer (AST) or autoregressive predictive coding (APC) may be applied so that the machine learning system can learn time-series audio data related to rainfall.

[0066] Furthermore, AST or APC may be used to classify audio recordings (of rain on luminaire housings) into different categories corresponding to rainfall levels. Only very limited ground truth data is needed to fine-tune and map these audio classes to actual rainfall levels. Ground truth data may come from local weather stations or utilize precipitation estimates from transit satellites. High-resolution satellite imagery can have a resolution of 30cm to 5m, and this granularity can be useful not only for obtaining information on rainfall levels for each luminaire but also for checking whether luminaires are blocked by tree branches, etc. However, satellite data is very expensive to acquire for cities and is therefore preferable to use only during the training phase of a low-cost rainfall inference system. A typical distance between luminaires equipped with sensor bundles is 30m to 50m.

[0067] Self-learning allows the sensor bundle, including a microphone and any other sensors that may be used, to be applicable to many different types of lighting fixtures (such as those with different surface areas exposed to rain, or those with different audio responses when raindrops hit the top surface of the lighting fixture).

[0068] In step 36, mean pooling (or average pooling) is optionally performed to fine-tune the data using small batches of labeled data. This divides the input region (e.g., pooling window or filter) into non-overlapping regions and calculates the mean value within each region. The resulting output feature map has reduced spatial dimension compared to the input.

[0069] In step 38, the algorithm generates a probability of precipitation, taking into account the specific type of lighting fixture to which the sensing system is applied.

[0070] For example, an algorithm might use a simple linear layer or linear transformation (called a linear head) applied to a representation learned from a pre-trained model. Linear heads are often added on top of pre-trained models to perform specific downstream tasks or extract useful features. Adding a linear head maps the pre-trained model's representation to a task-specific output space, enabling the model to perform well on the desired task. Essentially, a linear head adapts the pre-trained representation to a specific downstream task by learning task-specific weights.

[0071] This method is performed on each luminaire selected to report individually or locally estimated precipitation levels.

[0072] Figure 3 illustrates the method for combining these estimated precipitation levels from multiple lighting fixtures. Thus, this illustrates the process undertaken by multiple lighting fixtures and is used to explain the selection process of a subset of lighting fixtures.

[0073] Figure 3 shows a set of lighting fixtures L1 through LN. Lighting fixture L2 is covered by a tree canopy, and lighting fixture L3 is under a bridge. Thus, a suitable subset is L1, L4, and LN (and other lighting fixtures between numbers 4 and N). A full set may be, for example, all the lighting fixtures along one or more streets, or some of the lighting fixtures along a (long) street.

[0074] Lighting fixtures L2 and L3 are marked as unsuitable for contributing to combined precipitation estimates. Thus, they do not form part of the subset of lighting fixtures used to derive more accurate precipitation estimates. These lighting fixtures are selected based on predetermined selection criteria. Lighting fixtures are selected for the subset that is least likely to be obstructed by tree branches / leaves, has minimal exposure to audio noise, and therefore gives the best estimate of rainfall.

[0075] Lighting fixtures may be marked as unsuitable for use in a subset of the system during installation. However, the system may learn or determine on its own which lighting fixtures are unsuitable for inferring precipitation levels. For example, lighting fixtures far from trees, as determined by the radar sensor in the sensor bundle, or lighting fixtures that minimize exposure to audio noise (other than rain sounds), as determined by audio analysis using microphones, may be selected.

[0076] A subset of lighting fixtures may consist entirely of the same type, or it may contain multiple types of lighting fixtures within a subset.

[0077] In this example, lighting fixture L1 is used to derive the overall precipitation probability. Any lighting fixture may be used to derive the overall probability.

[0078] Lighting fixtures L1, L4, and LN form a subset used to derive an overall precipitation estimate. A local detection method is performed for each of these lighting fixtures.

[0079] In step 40, raindrop detection is performed based on the audio effect of water droplets on the selected lighting fixture housing, and in step 42, traffic-induced audio noise cancellation is performed.

[0080] In step 44, a self-learning algorithm is applied to each lighting fixture to estimate the rainfall level.

[0081] The estimated rainfall level is in the form of a probability distribution,460, for example, the respective probabilities for different types and / or intensities of precipitation.

[0082] During overall precipitation level inference, a message-passing algorithm is applied within a selected subset of lighting fixtures to obtain robust rain level estimates.

[0083] As shown in Figure 3, a forward message F is sent from lighting fixture L1 to another lighting fixture, and a backward reply message B is returned. Messaging includes local out-of-band message passing, which can be easily implemented using existing Bluetooth® Mesh radios already installed in the networked streetlights.

[0084] As an example, a "max-product" based message-passing inference system may be used to estimate rainfall in an area using multiple lighting fixtures within that area. The message-passing approach reduces the impact of a single lighting fixture being (partially) blocked by tree branches, or another lighting fixture being located near a (building) structure that partially obstructs raindrops. A message-passing group is generated, which the first lighting fixture passes its rainfall level prediction to neighboring lighting fixtures within the same sensing subset. This is repeated for all lighting fixtures in the subset. These lighting fixtures then collectively determine the maximum likelihood rainfall level for the area.

[0085] For example, the multiple probabilities received by the lighting fixture L1 are combined using any appropriate statistical approach, such as Bayesian fusion, to arrive at a better estimate.

[0086] One option is for all luminaires to report their precipitation estimates, but only the estimates from luminaires within a subset are used for overall precipitation level inference. For example, the system can additionally learn from these precipitation estimates (and other sensing information mentioned above) whether one or more luminaires are not exposed to the same rain conditions as their neighbors. Thus, the system can learn which luminaires in the set are suitable to be included in the subset used to derive the overall precipitation estimate.

[0087] In another option, shown in Figure 3, only the luminaires within a subset perform their precipitation estimation. Luminaires L2 and L3 do not need to run the estimation algorithm.

[0088] The adaptation to the subset members may be performed manually by the system administrator.

[0089] The overall precipitation probability of 50 is generated by lighting fixture L1. This is a real-time inference for the hyperlocal rain precipitation level. This information is then provided to a third-party road safety system.

[0090] As mentioned above, subsets may be defined dynamically. For example, if the system detects that one of the luminaires in a subset (i.e., a rain sensing group) is suddenly showing a much smaller or much larger rainfall level (e.g., a 3σ deviation) compared to its neighboring luminaires, the sensor bundle radar may be used to check whether the luminaire is obscured by tree branches or structures such as large trucks parked in front of it. In this way, it is possible to detect accurate rainfall levels even if some of the street luminaires are (temporarily) obscured from raindrops by tree branches or trucks.

[0091] Several proposed improvements and alternative options for the system are discussed below.

[0092] As mentioned above, a subset of lighting fixtures may include different types of lighting fixtures. For example, at an intersection, a first type of street-lighting luminaire may be located on a side street in a residential area, while a second type of road-lighting luminaire may be located on a main road. The algorithm can automatically cluster the two different lighting fixture types into a first classifier and a second classifier, and then perform rainfall inference for the first and second lighting fixture types using the first and second neural network weight sets.

[0093] Furthermore, the algorithm may identify which lighting fixtures within a city are located near public weather stations. This can provide accurate rainfall readings that should be used to fine-tune the rain inference system to specific urban environmental conditions and to fine-tune it to specific types of lighting fixtures used in street lighting installations.

[0094] To customize a rain detection system for a specific light fixture, few-shot learning may be applied, utilizing data from both public weather stations and the nearest unobstructed light fixture. This allows the weights of the rain inference neural network to be adjusted for the specific light fixture. In other words, the audio response curves for each sensor bundle's local environment (wind speed, traffic noise, current canopy, etc.) and each light fixture type can be fine-tuned.

[0095] In reality, the same type of streetlight fixture is mounted on a pole at a first height and angle on the first street, and at a second height and angle on the pole on the second street, resulting in different measurements from the audio sensor bundle even with the same rainfall.

[0096] The rain algorithm may also utilize wind speed data. Wind speed data may be derived, for example, from pole swing (using the tilt sensor in the sensor bundle). For example, time-series wind data can be used to selectively mask audio data from the rain detection audio dataset at the moment a strong gust of wind hits the sensor bundle.

[0097] The amplitude of time-domain waveforms output by a microphone is generally proportional to precipitation (e.g., increases during rainy periods). This allows for the detection of precipitation and its type / intensity. However, in some cases, precipitation inference based solely on microphone data may not yield inconclusive results between two or more precipitation types. For example, temperature data may be used to help classify detected precipitation as hail rather than heavy rain if the temperature is below 0°C. Similarly, sharp drops in temperature often occur during precipitation periods. Thus, temperature sensors can be used to detect the onset of rain or other precipitation types from temperature drops in time-series temperature data.

[0098] When temperature sensing is used, lighting fixtures may be selected to form part of a subset where the direct environment is not affected by nearby airflow from the building (e.g., HVAC exhausts, retail store sliding doors, subway exhaust vents, etc.).

[0099] Another sensor modality that may be used is RF sensing. In fact, it has been proposed to use RF sensing performed by streetlights to detect atmospheric conditions.

[0100] Another sensor modality that may be used is motion sensing. The aforementioned WO 2016 / 156563 utilizes accelerometers to detect precipitation from vibrations experienced by accelerometers incorporated into lighting fixtures, let alone accelerometers inside lighting fixtures. Thus, accelerometer sensors may additionally be used to detect the onset of rain or other types of precipitation from time-series accelerometer data.

[0101] When accelerometer data is used, lighting fixtures may be selected for a subset that is not affected by vibrations from traffic, machinery, subways, etc., where the direct environment may introduce noise into the accelerometer data.

[0102] Outdoor humidity sensors may also be used to detect rain or other precipitation types from changes in time-series humidity data. Instead of dedicated humidity sensors, it has been proposed to sense air humidity levels using RF sensing.

[0103] When humidity sensing is used, lighting fixtures may be selected to form part of a subset that is not affected by nearby exhaust vents, such as heating and ventilation system vents, retail store sliding doors, and subway vents, where the direct environment may alter the humidity of the air around nearby lighting fixtures.

[0104] Another sensor modality that may be used is pressure sensing for detecting rain or other precipitation types based on time-series pressure changes. Instead of a classic pressure sensor, changes in atmospheric pressure may be indirectly inferred by piezoelectric sensor means positioned in the housing of the lighting fixture. The opening of the housing of the outdoor lighting fixture is sealed to prevent water from entering the housing. This results in a pressure difference between the inside and outside of the housing of the lighting fixture when atmospheric pressure changes rapidly, as is the case with most of the onset of rain.

[0105] For example, the lighting fixture housing may include at least a flexible portion and a rigid portion, the flexible portion being more flexible than the rigid portion, and the piezoelectric sensor may be adapted to sense the deformation of the housing by sensing the deformation of the flexible portion of the housing.

[0106] Another sensor modality that may be used is a daylight sensor. A daylight sensor can be used to determine whether a streetlamp is obscured by a tree (for example, if the streetlamp is in the sun during winter and in the shade at other times of the year).

[0107] As mentioned earlier, precipitation estimates from a group of lighting fixtures allow us to infer the extent to which a particular streetlamp is obscured by tree branches. This data can be used by the system to map which streetlamps are located under (large) trees. This information may also be used to monitor tree overgrowth as part of a highway monitoring function.

[0108] If all nearby streetlights are at least partially obscured by trees, a subset of streetlights exhibiting similar levels of rain blocking by overhead tree canopy may be selected. To improve the robustness of the rainfall inference, detected rainfall is averaged across the subset of streetlights, and the rainfall estimate is then corrected to account for tree canopy based on ground truth data. Ground truth data may be obtained, for example, from a local weather station somewhere in the city.

[0109] The algorithm required for this invention is not computationally intensive. The algorithm may be unsupervised or weakly supervised.

[0110] In addition to detecting rain, the detected precipitation can take the form of one of several precipitation types, such as rain, snow, hail, or sleet.

[0111] Rainfall monitoring for many applications, such as road safety applications, does not require rainfall data of the same quality as that of a professional weather station, as long as the rain sensing data is accurate enough to infer a rough class of rainfall intensity, especially during heavy rain.

[0112] By examining the drawings, this disclosure, and the appended claims, modifications to the disclosed embodiments can be understood by those skilled in the art and can be implemented in carrying out the claimed invention. In the claims, the word “comprising” does not exclude other components or steps, and the indefinite article “a” or “an” does not exclude plural.

[0113] The functions performed by the processor may be performed by a single processor or by a number of separate processing units that can be considered collectively as a “processor.” Such processing units may be remote from one another and may communicate with one another by wired or wireless means (optional).

[0114] The mere fact that certain means are enumerated in different dependent claims does not indicate that combinations of these means cannot be used advantageously.

[0115] Computer programs may be stored and distributed on suitable media such as optical storage media or solid-state media, supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems.

[0116] Note that when the term "adapted to" is used in the claims or specification, it is intended to be equivalent to the term "configured to." Note that when the term "arrangement" is used in the claims or specification, it is intended to be equivalent to the term "system," and vice versa.

[0117] No reference numeral in a claim should be construed as limiting the scope.

Claims

1. A lighting fixture, said lighting fixture is Light source and A precipitation sensor for detecting precipitation at the location of the lighting fixture, Includes, The aforementioned precipitation sensor is A microphone for detecting sound in the lighting fixture, An input unit for receiving information derived from sound detected in other lighting fixtures in the vicinity of the lighting fixture, wherein the lighting fixture and the other nearby lighting fixtures form a set of lighting fixtures, An output unit for sending information derived from the sound detected in the lighting fixture to other lighting fixtures in the vicinity of the lighting fixture, A processor adapted to process the sound detected by the lighting fixture and the received information, and to derive a precipitation estimate from there, The processor includes, Select a subset of lighting fixtures based on at least one predetermined selection criterion, and To derive a precipitation estimate, the information relating to the subset of lighting fixtures is used. Lighting fixtures that are adapted to this purpose.

2. At least one predetermined selection criterion is: Obstruction from above by plants, Obstruction from above by artificial structures, Environmental non-precipitation noise level, A lighting fixture according to claim 1, comprising one or more of the following.

3. Precipitation estimates are, Occurrence of precipitation, Intensity of precipitation, Precipitation droplet size, A lighting fixture according to claim 1 or 2, comprising one or more presumptions among the following.

4. The luminaire according to claim 3, wherein the processor is adapted to select a time-varying subset.

5. The lighting fixture according to claim 4, wherein the time-varying subset has a different number of members at different times.

6. The aforementioned precipitation sensor further, Vibration sensor, Wind speed sensor, Temperature sensor, Humidity sensor, pressure sensor, Daytime running light sensor, RF sensor, A lighting fixture according to any one of claims 1 to 5, comprising one or more of the following.

7. The lighting fixture according to any one of claims 1 to 6, wherein the precipitation sensor further includes a radar sensor.

8. The lighting fixture according to claim 7, wherein at least one predetermined selection criterion includes the proximity of a tree or branch determined by the radar sensor.

9. The lighting fixture according to any one of claims 1 to 8, wherein the lighting fixture includes a wireless system for sending information to and receiving information from other lighting fixtures.

10. The lighting fixture according to any one of claims 1 to 9, wherein the information received and transmitted includes a precipitation level probability distribution.

11. The luminaire according to any one of claims 1 to 10, wherein the processor is adapted to derive a shared precipitation level for an area of ​​the set of luminaires.

12. The lighting fixture according to any one of claims 1 to 11, wherein at least one predetermined selection criterion includes the type and / or orientation of the other lighting fixtures.

13. The lighting fixture according to any one of claims 1 to 12, wherein the precipitation estimate is the estimated rainfall and / or raindrop size.

14. Paul and, A lighting fixture according to any one of claims 1 to 13, wherein the lighting fixture is attached to the pole, Streetlights, including

Images