Lighting system integrated hail detection

Abstract

A hail detection and analysis method includes performing, by a processor of an outdoor luminaire (102, 300), a hail detection at least based on sounds captured by one or more microphones of the outdoor luminaire. The processor is configured to execute a first artificial intelligence (AI) model to classify the sounds to detect hail. The method further includes estimating, by the processor, a severity level of the hail at least based on a terminal velocity of hail stones of the hail and an intensity level of the hail. The terminal velocity of the hail stones and the intensity level of the hail are determined based on at least radar signals emitted by one or more radar devices of the outdoor luminaire. The processor is configured to execute a second AI model to estimate the severity level of the hail.

Description

[0001] Lighting system integrated hail detection

[0002] FIELD OF THE INVENTION

[0003] The present disclosure relates generally to hail detection, and more particularly to artificial intelligence (Al) based hail detection and hail severity estimation integrated in outdoor light system.

[0004] BACKGROUND OF THE INVENTION

[0005] Hail can cause damage to building structures, vehicles, and even crop, and may also cause injury to people. Damages caused by hail to structures, such as roofs and glass windows of houses, cars, and crops can be costly. Hail can also cause traffic chaos, for example, as drivers that are caught up in a hailstorm seek shelter. In some cases, detecting hail may enable the notification of drivers and others to avoid an area affected by hail or to take shelter or other protective measures. Determining the severity of hail in addition to detecting the hail events may enable the assessment of expected damage to property. For example, if a person claims extensive hail damage to a property in an area where light hail was detected, an insurance company may determine that the claim is unfounded. As another example, if a person claims property damage in an area where hail was not detected but that is close to a location of a hail event, the insurance company may determine the claim to be unfounded. Thus, a solution that enables detecting hail and determining the severity of hail may be desirable.

[0006] SUMMARY OF THE INVENTION

[0007] The present disclosure relates generally to hail detection, and more particularly to Al based hail detection and hail severity estimation integrated in outdoor light system. In an example embodiment, a hail detection and analysis method includes performing, by a processor of an outdoor luminaire, a hail detection at least based on sounds captured by one or more microphones of the outdoor luminaire. The processor is configured to execute a first Al model to classify the sounds to detect hail. The method further includes estimating, by the processor, a severity level of the hail at least based on a terminal velocity of hail stones of the hail and an intensity level of the hail. The terminal velocity of the hail stones and the intensity level of the hail are determined based on at least radar signals emitted by one or more radar devices of the outdoor luminaire. The processor is configured to execute a second Al model to estimate the severity level of the hail.

[0008] In some another example embodiment, an outdoor luminaire system for hail detection and hail severity estimation includes an outdoor luminaire comprising a processor, one or more radar devices, and one or more microphones. The processor is configured to execute a first Al model to detect hail at least by classifying sounds captured by the one or more microphones. The processor is further configured to execute a second Al model to estimate a severity level of the hail at least based on a terminal velocity of hail stones of the hail and an intensity level of the hail. The terminal velocity of the hail stones and the intensity level of the hail are determined based on at least radar signals emitted by the one or more radar devices.

[0009] These and other aspects, objects, features, and embodiments will be apparent from the following description and the appended claims.

[0010] BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Reference will now be made to the accompanying drawings, which are not necessarily drawn to scale, and wherein:

[0012] FIG. 1 illustrates an outdoor luminaire system for detecting hail and estimating hail severity according to an example embodiment;

[0013] FIG. 2 illustrates a method of detecting hail and estimating hail severity performed by an outdoor luminaire of the outdoor luminaire system of FIG. 1 according to an example embodiment;

[0014] FIG. 3 illustrates a block diagram of an outdoor luminaire of the outdoor luminaire system of FIG. 1 according to an example embodiment; and

[0015] FIG. 4 illustrates a method of detecting hail and estimating hail severity by the luminaire system of FIG. 1 according to an example embodiment.

[0016] The drawings illustrate only example embodiments and are therefore not to be considered limiting in scope. The elements and features shown in the drawings are not necessarily to scale, emphasis instead being placed upon clearly illustrating the principles of the example embodiments. Additionally, certain dimensions or placements may be exaggerated to help visually convey such principles. In the drawings, the same reference numerals used in different figures may designate like or corresponding but not necessarily identical elements. DETAILED DESCRIPTION OF THE EXAMPLE EMBODIMENTS

[0017] In the following paragraphs, example embodiments will be described in further detail with reference to the figures. In the description, well known components, methods, and / or processing techniques are omitted or briefly described. Furthermore, reference to various feature(s) of the embodiments is not to suggest that all embodiments must include the referenced feature(s).

[0018] FIG. 1 illustrates an outdoor luminaire system 100 for detecting hail and estimating hail severity according to an example embodiment. In some example embodiments, the system 100 includes outdoor luminaires 102, 104, 106, 108, 110, and 112. For example, the luminaires 102-112 may be streetlight luminaires. The luminaires 102-112 may include sensors and other devices that may be used in the detection of hail and to estimate the severity level of the hail in the area that includes the luminaires 102-112. For example, the luminaires 102. 104, 106 may be on one side of a street 122, and the luminaires 108, 110, 112 may be on the opposite side of the street 122 generally across from the luminaires 102-106.

[0019] In some example embodiments, some luminaires of the luminaires 102-112 may be the same type of luminaires (e.g., the same SKU number) with respect to each other but may be different from other luminaires of the luminaires 102-112. For example, the luminaires 102-106 may be the same type of luminaires (e.g., the same SKU number or have similar housing) with respect to each other but may be different types (e.g., different SKU numbers) from the luminaires 108-112. To illustrate, the luminaires 102-106 may have the same type or similar housing such that the audio effect of hail stones hitting the housing of each of the luminaires 102-106 is similar and such that sensor devices, such as microphones and radar devices, are attached in generally the same orientation across the luminaires 102- 106. The luminaires 108-112 may be the same type of luminaires with respect to each other but may be different from the luminaires 102-106. To illustrate, the luminaires 108-112 may be located between the street 122 and the buildings 130, 132 that may be, for example, residential houses. For such or other reasons, the luminaires 108-112 may be different types of luminaires from the luminaires 102-106.

[0020] In some example embodiments, if the luminaires 102-106 are a different type from the luminaires 108-112, training data for artificial intelligence (Al) models executed by the luminaires 102-112 may be obtained on a per luminaire-type basis. For example, at least some Al models executed by the luminaires 102-106 may be trained using training data applicable to the type of the luminaires 102-106, and at least some Al models executed by the luminaires 108-112 may be trained using training data applicable to the type of the luminaires 108-112. In some cases, ground truth training data may be obtained using each luminaire of the luminaires 102-112 to additionally train the particular luminaire with training data specific to the luminaire.

[0021] In some example embodiments, the luminaires 102-112 are located in close enough proximity to each other to experience the same or substantially the same level of hail intensity (e.g., number of hail stones) as well as hail stone size distribution and terminal speeds of hail stones. The luminaires 102-112 may also communicate with each other, for example, to pass information among the luminaires 102-112. For example, the luminaires 102-112 may form a wireless mesh network (e.g., a Bluetooth mesh network) that can be used, for example, for local out-of-band message passing between the luminaires 102-112. Alternatively or in addition, the luminaires 102-112 may communicate using one or more other communication networks and / or protocols. The luminaires 102-112 may also communicate with other devices and systems such as a cloud server, etc.

[0022] In some example embodiments, the individual luminaires 102-112 of the system 100 may detect the occurrence of hail. For example, the luminaires 102-112 may each detect hail based on sounds produced by the hail. To illustrate, the luminaires 102-112 may each detect hail based on sounds produced by hail stones impacting the housing of the respective one of the luminaires 102-112. The sounds may be captured by respective one or more microphones of the luminaires 102-112. Each luminaire of the luminaires 102-112 may execute one or more Al models to classify sounds captured by the microphones of the particular luminaire and thereby determine whether the sounds resulted from hail.

[0023] In some cases, because sounds captured by microphones of the luminaires 102-112 may include or may be entirely other sounds, such as traffic sounds and wind gust sounds, the sounds captured by the microphones may be filtered to remove or at least reduce noise before classification is performed by an Al model. Alternatively or in addition, sounds captured at the same time as vibrations being detected may also be ignored entirely. For example, the luminaires 102-112 may include a respective vibration sensor.

[0024] In some example embodiments, upon the detection of hail by one or more of the luminaires 102-112, a notification of the hail detection may be sent, for example, to insurance companies or traffic authorities that can notify drivers and others to stay away from the area of hail detection or to take shelter. For example, one of the luminaires 102-112 or another component (e.g., a server) of the system 100 may transmit the notification. To illustrate, the system 100 may transmit a notification based on the detection of hail by one luminaire of the luminaires 102-112 or based on detections by some or all of the luminaires 102-112. As a non-limiting example, if a majority of the luminaires 102-112 detect hail, one or more of the luminaires 102-112 or another device of the system 100 may transmit a notification indicating hail detection.

[0025] In some example embodiments, traffic behavior patterns that are observed during hail events may be used to detect hail. The luminaires 102-112 may be used to detect traffic behavior patterns associated with driving during a hailstorm. Speed reduction, movements of cars toward the curb, stoppage of cars, and / or other driving behaviors observed during hailstorms may be used to detect hail. For example, during a hailstorm, most drivers tend to stop and search for cover (e.g., under a tree or a roof) to protect their cars. To illustrate, the movement of cars, such as cars 124, 16, 128, within radar detection regions of the luminaires 102-108 may be used to determine traffic behavior patterns of the cars. Training data of traffic behavior patterns can be collected to train an Al model to understand the temporal and spatial traffic patterns in case of hail events, and the trained Al model can thereby detect the occurrence of hail based on the traffic behavior patterns detected, for example, using radar. The detection of hail based on traffic behavior patterns may be performed instead of or in addition to hail detection based on sounds produced by hail.

[0026] In some example embodiments, one or more of the luminaires 102-112 may be excluded from use to detect hail. For example, if the luminaire 102 is at least partially covered by a branch of a tree 142, the tree branch may block the hail stones from impacting the luminaire 102 or otherwise interfere with hail stones directed toward the luminaire 102. As such, the sounds captured by microphones of the luminaire 102 may not be reliable in indicating the occurrence of hail. The tree 142 may also interfere with the reliable use of radar, for example, to detect cars and people in the vicinity of the luminaire 102 and for other purposes such as the measurement of terminal velocities of hail stones. Whether a tree or another structure interferes with hail detection by a particular luminaire of the system 100 may be determined, for example, by comparing captured sounds and radar signals against expected sounds and radar signals, respectively, under known conditions. Alternatively or in addition, images from cameras of one or more of the luminaires 102-112 may be used to determine whether a tree or another structure interferes with hail detection and other tasks that can be performed by a particular luminaire of the system 100.

[0027] In some example embodiments, barometric pressure and wind speed may be used to confirm the occurrence of hail detected based on sounds produced by the hail (e.g., sounds produced by hail stones impacting housings of the luminaires 102-112) and / or based on traffic behavior patterns determined based on radar signals transmitted by the luminaires 102-112. For example, because barometric pressure is relatively lower (e.g., below a threshold pressure) and wind speed is relatively higher (e.g., above a threshold wind speed) during hailstorms, barometric pressure measured by a barometer and / or windspeed estimated based on vibrations sensed by a vibration sensor of a luminaire may be used to confirm or invalidate hail detection performed based on sounds and / or radar. To illustrate, one or more of the luminaires 102-112 may include a barometer and / or a vibration sensor. Alternatively, pressure and windspeed information may be obtained from another source such as a nearby public weather station. Hail related information from a nearby public weather may also be used to confirm or invalidate hail detection by the luminaires 102-112 and to fine-tune Al models that are used by the luminaires 102-112 to detect hail by classifying sounds and / or traffic behavior patterns.

[0028] In some example embodiments, light level sensed by one or more photosensors of the system 100 may be used to confirm or invalidate the detection of hail. For example, the luminaires 102-112 may include a respective photosensor. Because light level is relatively low during hail events, if a particular luminaire of the system 100 indicates the detection of hail while light level determined using a photosensor of the particular luminaire is above a threshold level, the hail detection by the luminaire may be ignored or may be further verified using other parameters such atmospheric pressure, wind speed, etc.

[0029] In some example embodiments, the luminaires 102-112 may individually estimate the severity level of hail. The severity of hail may be one of multiple categories of severity. For example, possible severity levels may be “low severity,” “medium severity,” and “high severity.” To illustrate, the luminaire 104 may estimate the severity of hail detected by the luminaire 104 as low severity. As another example, the severity level of hail estimated by the luminaire 104 may be expressed as probability values (e.g., a percentage values) associated with categories of hail severity. In general, the severity level of hail may be expressed using labels and / or values corresponding to different degrees of severity.

[0030] In some example embodiments, the luminaires 102-112 may each estimate the severity level of hail in response to the detection of hail by the particular luminaire. For example, each one of the luminaires 102-112 may estimate the severity level of hail based on sounds produced by hail stones of the hail impacting the particular luminaire. For example, if the tree 142 affects the number of hail stones or otherwise interferes with hail stones that impact the housing of the luminaire 102, the sounds produced by hail stones impacting the housing may not accurately reflect the severity level of the hail. In such cases, the sounds captured by the microphones of the luminaire 102 may not be used to estimate the severity level of the hail. The luminaires 102-112 may each execute one or more Al models to estimate the respective severity level of hail based on sounds resulting from hail stones impacting the luminaires 102-112.

[0031] Alternatively or in addition to sounds, an individual luminaire of the system 100 may estimate the severity level of the hail based on traffic behavior patterns as determined using radar signals transmitted by a radar device of the particular luminaire. For example, if the tree 142 interferes with the radar signals transmitted by the radar device of the luminaire 102 to detect the presence, location, speed, etc. of cars and people on the ground, traffic behavior patterns determined based on the radar signals may not accurately reflect the severity level of the hail. In such cases, traffic behavior patterns may not be used by the luminaire 102 to estimate the severity level of the hail. In some alternative embodiments, to determine traffic behavior pattern that is subsequently used by the luminaire 102 to estimate hail severity level, the luminaire 102 may use traffic information from one or more of the luminaires 104-112 along with traffic information determined using radar signals transmitted by the luminaire 102 itself. The luminaires 102-112 may each execute one or more Al models to estimate the respective severity level of hail based on traffic behavior pattern.

[0032] Alternatively or in addition to using sounds and traffic behavior patterns, an individual luminaire of the system 100 may estimate the severity level of hail based on the terminal velocity (e.g., average terminal velocity) of hail stones of the hail and based on the intensity level of hail that may be determined using radar signals transmitted by a radar device of the particular luminaire. For example, a radar device of a particular luminaire may be positioned such that at least some of the hail stones that impact the housing of the particular luminaire are detected using radar signals from the radar device. To be clear, in contrast to the severity level of hail, intensity level of hail refers to number of hail stones, for example, in reference to a time period (e.g., 1 second or 1 minute) and an area (e.g., square feet). As an illustrative example, the intensity level of the hail may be N (e.g., N hail stones per second per square feet). Alternatively or in addition to using radar signals, a particular luminaire (e.g., the luminaire 106) may determine the intensity level of hail based on sounds produced by hail stones impacting the particular luminaire (e.g., hails stones impacting the housing of the luminaire). In general, the luminaires 102-112 may each execute one or more Al models to estimate a respective severity level of hail. After estimating the severity level of hail, one or more luminaires of the system 100 may transmit information indicating the severity level of hail, for example, to insurance companies. In some example embodiments, the luminaires 102-112 may each pass a respective severity level estimated by the particular luminaire to other luminaires of the system 100 such that the luminaires 102-112 may each use severity levels of hail estimated by multiple luminaires to determine an overall severity level of hail. For example, each luminaire of the system 100 may pass information indicating the severity level (e.g., low severity, medium severity, or high severity) estimated by the particular luminaire to one or more of the other luminaires of the system 100, where a luminaire of the system 100 that receives the severity level information from other luminaires may determine an overall severity level of hail based on, for example, a majority rule. For example, if the severity level of hail estimated by most of the luminaires 102-112 is low severity, one or more of the luminaires 102-112 that receive the severity levels may determine that the overall severity level is low severity. In some alternative embodiments, individual luminaires of the system 100 may use severity levels from a subset of the luminaires 102-112 instead of from all luminaires of the system 100 to determine an overall severity level. Alternatively, maxproduct based message passing may be used to estimate an overall severity level.

[0033] In some example embodiments, the luminaires 102-112 may each pass severity levels to each other using probability values corresponding to different categories of severity. To illustrate with respect to the luminaires 102 and 104, the luminaire 102 may pass to the luminaire 104 probability values corresponding to different severity levels of hail as estimated by the luminaire 102. For example, the probability values may be outputs of an Al model executed by the luminaire 102 to estimate the severity level of hail based on sounds, traffic behavior patterns, and / or hail intensity level and terminal velocity of hail stones as described above. As an illustrative example, the probability values from the luminaire 102 may be such that “medium severity” has a high probability value compared to the other severity categories.

[0034] In some example embodiments, the luminaire 104 may use the probability values received from the luminaire 102 along with probability values corresponding to different severity levels estimated by the luminaire 104 to derive resultant probability values. For example, for each possible severity level (e.g., low severity, medium severity, and high severity), the luminaire 104 may derive a respective resultant probability value by determining, for example, the average value of the probability values from the luminaires 102 and 104 with respect to the particular severity level. The luminaire 104 may then determine an overall severity level of hail from the resultant probability values by selecting, for example, the particular severity level that has the highest probability. For example, medium severity or another severity level (e.g., high severity) that has the highest probability value from among the resultant probability values may be selected as the overall severity level of hail.

[0035] In some example embodiments, the luminaire 104 may receive severity levels (e.g., as probability values corresponding to different severity levels) from multiple luminaires such as the luminaires 102, 106-112 and determine an overall severity level based on probability values from the luminaires 102-112. In general, each luminaire of the system 100 may determine an overall severity level in the same manner as described with respect to the luminaire 104.

[0036] In some example embodiments, after determining an overall severity level of the hail, one or more luminaires from among the luminaires 102-112 may transmit information indicating the severity level of the hail. For example, the overall severity level of the hail may be sent to one or more insurance companies. The one or more luminaires may also transit other information, such as locations (e.g., GPS locations) of the luminaires 102- 112 to enable identifying areas that are affected by the hail. Alternatively or in addition, another device (e.g., a server) of the system 100 may determine and / or transmit information indicating the overall severity level of the hail. As a non-limiting example, the information conveying the overall severity level of hail may indicate one of multiple categories of severity level, such as low severity, medium severity, high severity, and / or another category of severity level.

[0037] In some example embodiments, other information that can be used to assess expected damages due to hail in the area near the luminaires 102-112 may be determined and sent to, for example, insurance companies. For example, the number of cars that were at or that went by the area near the luminaires 102-112 during the hail event may be determined using radar signals transmitted by one or more of the luminaires 102-112. To illustrate, based on the locations of the luminaires 102-112 that experienced a hail event and the severity level of the hail as provided by each of the luminaires 102-112 and / or the overall severity level of hail, expected damage to cars, to structures such as the buildings 130, 132, and other property or persons may be assessed by insurance companies. Other information such as speeds of cars, etc. during the hail event can also be provided to insurance companies. To illustrate, information such as locations and speeds of cars in the area can be provided, for example, to insurance companies from other sources, and the information can be analyzed along with the hail severity and other information from the system 100 to estimate damages. In some alternative embodiments, the system 100 may perform expected damage assessments and send the damage related information to insurance companies.

[0038] In some example embodiments, the luminaires 102-112 may each distinguish between ice pellets (sleet), graupel, and hail stones (e.g., between 5 mm and 15 cm) based on the sounds captured by microphones and / or based on radar signals. For example, an Al model that is trained based on audio and radar data using labeled ground truth precipitation types or elements may be used to distinguish between ice pellets, graupel, and hail stones. In some example embodiments, thunder can be used to distinguish hail stone size categories. To illustrate, relatively louder thunder and more frequent thunders are associated with larger size hail stones. As such, the loudness and frequency of thunder may be used to estimate hail stone sizes. In some example embodiments, an accelerometer of one or more of the luminaires 102-112 may be used to determine kinetic energy of hail stone impacts on the respective luminaires and the estimated kinetic energy may be used to estimate the size (e.g., average diameter) of hail stones. Because damage caused by hail stones are associated with hail stone sizes, the size information may be used as additional information to assess expected damage from hail. For example, the hail stone size information along with other information may be transmitted to insurance companies or many be used by the system 100 to estimate damage to property, etc.

[0039] In some example embodiments, other outdoor luminaires, such as outdoor luminaires 114, 116, 118, 120, may be part of the system 100. For example, the luminaires 114-120 may detect hail and estimate the severity level of hail along with the luminaires 102- 112 and in the manner described with respect to the luminaires 102-112. Alternatively, the luminaires 114-120 may detect hail and estimate severity level of hail as a separate cluster of luminaires. For example, the luminaires 114-120 may be located too far from the luminaires 102-112 and thus may not be expected to experience the same severity level of hail. To illustrate, the severity level of hail in the area near the luminaires 102-112 may be, for example, low, while the severity level of hail in the area near the luminaires 114-120 is medium or high. As another example, the luminaires 114-120 may be too far from the luminaires 102-112 and may not experience the hail event.

[0040] In some example embodiments, the luminaires 114-120 may detect hail and estimate severity level of hail in the manner described above with respect to the luminaires 102-112. Sounds produced by hail, such as sounds produced by hail stones impacting the luminaires 114-120, may be used to detect hail and to estimate the severity level of the hail. Traffic behavior patterns determined based on cars such as cars 134, 136 can be used in the manner described with respect to the luminaires 102-112 to detect hail and to estimate the severity level of the hail. A notification of the detection of hail and information indicating the severity level of the hail may be sent, for example, to a governmental agency (e.g., traffic control or roadway safety agency), to insurance companies, etc. by one or more of the luminaires 114-120. For example, the luminaires 114-120 may be located near a farmland 140 and structures such as a building 138, and hail damage to crops and / or other properties in the area can be estimated or assessed based on the estimated severity level of the hail and the locations of the luminaires 114-120 with respect to the farmland 140 and other properties.

[0041] In some example embodiments, if one of the luminaires 102-112 that was indicating, for example, a high severity level of hail starts indicating a medium or low severity level of hail while the majority of the remaining luminaires continue to indicate a high severity level, the possibility that a temporary partial blockage of the particular luminaire may be checked. For example, radar or camera-based checking of the particular luminaire for interference by a tree or another object / structure may be performed. If a determination is made that the particular luminaire is blocked by a tree or another object / structure, the severity level estimate provided by the particular luminaire may be excluded from further use. For example, the particular luminaire itself may perform the analysis and checking related to change in the severity level.

[0042] In some example embodiments, each luminaire of the system 100 may execute an Al model to estimate the level of blockage or interference of the luminaire by tree branches or another object or structure that can prevent or otherwise interfere with hail stones impacting the luminaire. For example, for each type of the luminaires of the system 100, a respective Al model may be trained with labeled ground truth audio data corresponding to different levels of blockage or interference. During inference, the luminaires of the system 100 may classify the sounds into different categories of blockage / interference. The blockage / interference estimates can be used to decide whether a particular luminaire can be used to reliably detect hail and / or estimate hail severity. The blockage / interference information can also be provided to the entity responsible for clearing the blockage / interference.

[0043] By using one or more of sounds produced by hail (e.g., such as sounds from hail stones hitting the housing of a luminaire), traffic behavior patterns associated with hail events, terminal velocity of hail stones, and hail intensity, each luminaire of the system 100 may detect hail and estimate the severity level of the hail by executing Al models and send notifications of hail detection and information indicating the severity of the detected hail. By detecting occurrences of hail using the outdoor luminaires 102-120 and based on the locations of the luminaires 102-120, areas that experience a hail event can be generally identified. By providing notifications of hail detection, the system 100 can enable drivers to avoid the particular area where hail is detected and to take other measures to reduce hail damage. If the luminaires of the system 100 do not detect hail, insurance claims for hail damage in the areas where the luminaires are located can be considered as potentially fraudulent. By estimating the severity of hail, the system 100 can enable insurance companies to estimate damages and check insurance claims for consistency with the expected damages. The hail detection and severity level information can enable insurance companies to estimate claim damages and send claim appraisers to affected areas.

[0044] In some alternative embodiments, the system 100 may include more or fewer luminaires than shown without departing from the scope of this disclosure. In some alternative embodiments, the luminaires of the system 100 may be located in different configuration and in a different area than shown without departing from the scope of this disclosure.

[0045] FIG. 2 illustrates a method 200 of detecting hail and estimating hail severity performed by an outdoor luminaire of the outdoor luminaire system 100 of FIG. 1 according to an example embodiment. Referring to FIGS. 1 and 2, to illustrate with respect to the luminaire 102 as a representative luminaire of the luminaires 104-120 of the system 100, the luminaire 102 may include a processor that performs operations described herein. At step 202, the luminaire 102 may perform audio sensing and filtering operations. For example, one or more microphones of the luminaire 102 may capture sounds from different sources. During hail events, the sounds captured by the microphones may include sounds generated by hail stones impacting the luminaire 102. The sounds captured by the microphones may also include sounds generated by hail stones impacting the ground and other structures in the vicinity of the luminaire 102. The sounds may also include unwanted sounds / noise for the purpose of hail detection. For example, the sounds may include wind just sounds, traffic sounds, etc. The luminaire 102 may filter out noise such as wind gust sound, traffic sound, etc. from the sounds captured by microphones of the luminaire 102. For example, the luminaire 102 may perform frequency -based audio filtering and / or Al-based processing to filter out the noise as can be readily understood by those of ordinary skill in the art with the benefit of this disclosure.

[0046] In some example embodiments, sounds captured by the microphones while vibrations of the luminaire 102 are detected may be ignored. To illustrate, vibrations of the luminaire 102 may indicate strong winds and / or relatively large vehicles passing by the luminaire 102 that are likely to distort the sounds captured by the microphones of the luminaire 102. The sounds captured by the microphones may also be ignored if relatively heavy traffic is detected, for example, by the radar system / device of the luminaire 102.

[0047] In some example embodiments, the luminaire 102 may perform hail detection at step 204 based on the sounds captured by the microphones of the luminaire 102. In particular, the luminaire 102 may perform hail detection based on filtered sounds derived from the captured sounds. The luminaire 102 may detect hail by processing time-series audio data corresponding to the filtered sounds. For example, the luminaire 102 may execute an Al model to classify the sounds and determine whether the sounds indicate or otherwise correspond to the occurrence of hail. The Al model may have been trained using labeled ground truth data obtained for the particular type (e.g., an SKU number) of the luminaire 102. The type of the luminaire 102 may be known from installation information, from satellite images, etc.

[0048] In some example embodiments, the luminaire 102 may detect hail by executing a self-learning Al model to classify the audio data derived from the sounds captured by the microphones of the luminaire 102. To illustrate, the self-learning Al model may be first trained with unlabeled audio data to classify audio data into clusters. For example, the Al model may be an audio spectrogram transformer (AST) based model. The unlabeled audio data may be derived from or may otherwise correspond to sounds captured by microphones of luminaires that are the same type as the luminaire 102. The Al model may then be trained with a relatively small amount of labeled ground truth data specific to the type of the luminaire 102. The trained Al model can be executed by the luminaire 102 to classify the audio data and thus the sounds captured by the microphones of the luminaire 102. The audio data is classified by the luminaire 102 as indicating or otherwise corresponding to a hail event if the sounds captured by the microphones of the luminaire 102 are generated by hail stones impacting, for example, the housing of the luminaire 102 and / or other parts of the luminaire 102. That is, hail is detected by the luminaire 102 if the luminaire 102 determines, by classifying the sounds, that the sounds captured by the microphones of the luminaire 102 indicate or otherwise correspond to a hail event. The luminaire 102 may transmit a notification or other message, for example, to traffic control authority, to insurance companies, etc. indicating the detection of hail.

[0049] In some example embodiments, at step 206, the luminaire 102 may execute a self-learning Al model that is trained to classify sounds captured by the microphones of the luminaire 102 into categories of severity levels of the hail. To illustrate, sounds caused by hail stones impacting a luminaire may be correspond to different severity levels of hail. For example, the frequency components of the sounds, the distribution of the frequency components, and amplitudes of the frequency components may be related to different sounds based on the severity level of the hail. The self-learning Al model executed by the luminaire 102 may be or may include an AST-based model that is trained with unlabeled audio data. The unlabeled audio data may be derived from or may otherwise correspond to sounds captured by microphones of luminaires that are the same type as the luminaire 102. The Al model may be subsequently trained with labeled ground truth data specific to the type of the luminaire 102. For example, the ground truth data may be labeled with categories of severity levels such as, for example, low severity, medium severity, and high severity. The number of categories of the severity level and the particular categories may be defined by a user based on the user’s preference. As can be readily understood by a person of ordinary skill in the art, the Al model may output, during inference, probability values associated with the different categories of severity level, where, for example, the particular category of severity level that has the highest probability value is selected as the severity level estimation by the luminaire 102.

[0050] In some example embodiments, at step 208, the luminaire 102 may estimate the intensity level of hail based on the sounds captured by the microphones of the luminaire 102. For example, after unlabeled audio data is used to train an Al model, labeled ground truth audio data that is labeled with numbers of hails stones may be used to train the Al model. The unlabeled and the labeled audio data may be derived for the specific type of the luminaire 102, for example, from luminaires that are the same type as the luminaire 102. After training, the Al model may be used during inference at step 208 to estimate the intensity level of hail presented as a number of hail stones during a time period (e.g., 1 second or 1 minute) in an area (e.g., square foot or square meter).

[0051] In some example embodiments, at step 208, the luminaire 102 may estimate the intensity level of hail based on radar signals instead of or in addition to using sounds captured by the microphones of the luminaire 102. For example, at step 210, the luminaire 102 may perform radar sensing using radar signals transmitted by a radar device of the luminaire 102. To illustrate, the radar signals transmitted by the radar device may be directed to detect hail stones that are, for example, about to impact the housing of the luminaire 102. The near-field radar sensing operation performed at step 210 provides hail stone sensing information that is processed at step 208 to estimate the intensity of hail. In some example embodiments, the near-field radar sensing operation performed at step 210 may also provide terminal velocities of hail stones, for example, that impact the luminaire 102. For example, the radar device of the luminaire 102 may be oriented to transmit radar signals to sense speeds of hail stones that are, for example, about to impact the housing of the luminaire 102. At step 214, the luminaire 102 may determine a terminal velocity value representative of the terminal velocities of hail stones sensed at step 210. For example, at step 214, the luminaire 102 may determine an average terminal velocity from the terminal velocity values sensed at step 210 and provide the average velocity value to step 216 as the terminal velocity of the hail stones. The terminal velocity may be regularly updated during a hail event based on updated radar sensing at step 210.

[0052] In some example embodiments, at step 216, the luminaire 102 may perform hail severity estimation based on the hail intensity produced at step 208 and the terminal velocity from step 214. For example, at step 216, the luminaire 102 may execute an Al model that is trained with labeled ground truth data with respect to hail intensity and terminal velocity inputs. For example, the ground truth data may be labeled with categories of severity levels such as, for example, low severity, medium severity, and high severity depending on the severity level. The number of categories of the severity level and the particular categories may be defined by a user based on the user’s preference. As can be readily understood by a person of ordinary skill in the art, the Al model may output, during inference at step 216, probability values associated with the different categories of severity level, where, for example, the particular category of severity level that has the highest probability value is selected as the estimated severity level of hail by the luminaire 102.

[0053] In some example embodiments, at step 210, the luminaire 102 may perform radar sensing using radar signals transmitted by a radar device of the luminaire 102. For example, the radar signals transmitted at step 212 may be from a different or same radar device used to transmit radar signals at step 210. The radar signals transmitted by the radar device at step 212 may be directed to detect cars, pedestrians, and other moving objects, for example, in the street 122 and / or sidewalks of the street 122. The far-field radar sensing operation performed at step 212 can provide speed and location information that is processed at step 218 to estimate traffic behavior patterns. For example, changes in speed of cars in the street 122 that are within a radar detection range, the rate of speed changes (e.g., how quickly cars slow down), movements of cars toward the curbside of the street 122, etc. may be determined at step 218 based on information from the radar sensing operation at step 212. In some example embodiments, at step 220, the luminaire 102 may execute an Al model to estimate severity level of hail based on the traffic behavior patterns detected at step 218. As described above, some traffic behavior patterns are associated with the occurrence of hail and the severity level of the hail. The Al model may have been first trained using unlabeled traffic pattern data and subsequently trained using a relatively small amount of labeled ground truth data, where, for example, the ground truth data may be labeled with categories of severity levels such as, for example, low severity, medium severity, and high severity depending on the severity level. As can be readily understood by a person of ordinary skill in the art, the Al model may output, during inference at step 220, probability values associated with the different categories of severity level, where, for example, the particular category of severity level that has the highest probability value is selected as the estimated severity level of hail by the luminaire 102.

[0054] In some example embodiments, at step 222, the luminaire 102 may perform fusing of the estimated hail severity levels from steps 206, 216, 220 to derive a severity level of hail (i.e., a per-luminaire severity level of the hail) by the luminaire 102. For example, the luminaire 102 may execution a Bayesian Fusion algorithm to fuse the severity levels from steps 206, 216, 220. In some example embodiments, the luminaire 102 may take the averages of the probability values of the different severity categories from the steps 206, 216, 220 and select the severity category that has the highest probability value as the overall severity level of hail estimated by the luminaire 102.

[0055] In some example embodiments, at step 224, the luminaire 102 may receive estimates of hail severity level (indicated in FIG. 2 as Hail Severity Level #2, . . ., Hail Severity Level #N) from one or more other luminaires of the system 100 such as the luminaires 104-112 that operate in the same manner as the luminaire 102 to estimate a respective severity level of hail. The luminaire 102 may derive, at step 224, an overall hail severity level from the hail severity levels estimated by the luminaire 102 and one or more the luminaires 104-112. For example, at step 224, the luminaire may use a majority rule to determine the overall hail severity level. Alternatively, the hail severity levels from the luminaire 102 as well as the one or more other luminaires may include probability levels associated with different categories (e.g., low severity, medium severity, and high severity), and the luminaire 102 may, for example, generate an average of the probability values for each category and select the category with the highest average value as the severity level of the hail detected at step 204. The luminaire 102 may subsequently send the overall severity level of hail determined at step 224, for example, to a cloud server, insurance companies, etc. In some example embodiments, instead of or in addition to sending the overall severity level determined at step 224, the luminaire 102 may transmit the hail severity level (labeled Hail Severity Level #1 in FIG. 2), for example, to a cloud server, insurance companies, etc. The luminaire 102 may also transmit the hail severity level (labeled Hail Severity Level #1 in FIG. 2) to one or more other luminaires of the system 100 that may determine a respective overall hail severity level in the manner described with respect to the luminaire 102.

[0056] In some example embodiments, if the sounds captured by the microphones of the luminaire 102 are unreliable, the hail severity level estimation at step 206 may not be performed or if performed, the severity level estimated at step 206 may not be used at step 222. For example, if the housing of the luminaire 102 is partially or fully blocked by a branch of the tree 142, the sounds produced by hail stones impacting the housing of the luminaire 102 may not accurately indicate the severity level of the hail.

[0057] In some example embodiments, if the radar sensing operation at step 210 is unreliable, the hail severity level estimation at step 216 may not be performed or if performed, the severity level estimated at step 216 may not be used at step 222. For example, if radar signals transmitted by the radar device of the luminaire 102 are partially or fully blocked by a branch of the tree 142, the radar sensing operation at step 210 may not accurately sense hail stones. As such, the luminaire 102 may not accurately sense the severity level of the hail based on the terminal velocity determined at step 214 based on the radar sensing at step 210.

[0058] In some example embodiments, if the radar sensing operation at step 212 is unreliable, the hail severity level estimation at step 220 may not be performed or if performed, the severity level estimated at step 220 may not be used at step 222. For example, if radar signals transmitted by the radar device of the luminaire 102 are partially or fully blocked by a branch of the tree 142, the radar sensing operation at step 212 may not accurately detect people and cars in the street 122, speed of cars, etc. As such, the luminaire 102 may not accurately sense the severity level of the hail based on the traffic behavior patterns determined at step 218 based on the radar sensing at step 212.

[0059] In some example embodiments, the luminaire 102 may transmit other information that can be used by insurance companies to assess expected damages caused by hail. For example, the luminaire 102 may send hail stone size information. The luminaire 102 may also send the location of the luminaire 102 along with the notification of the detection of hail, the severity level as determined at step 222, and / or the overall severity level as determined at step 224. Alternatively or in addition, the luminaire 102 may process some of the information including severity level of the hail to estimate damages caused by the hail detected by the luminaire 102.

[0060] In general, Al model training data used in the method 200 is specific to the particular type of luminaire. For example, the training data used to train Al models executed by the luminaire 102 may be different from the training data used to train Al models executed by the luminaire 108 if the type of the luminaire 108 (e.g., SKU number) is different from the type of the luminaire 102. In some example embodiments, estimating hail severity level is performed at one or more of steps 206, 216, 220 only if hail is detected at step 204. For example, while the radar devices of the luminaires of the system 100 may be active for nonhail related sensing, the near-field radar sensing at step 210 and / or far-field radar sensing at step 212 with respect to hail related operations of the method 200 may not be performed or activated until after the detection of hail is indicated at step 204. To illustrate, terminal velocity measurement at step 214 and traffic pattern detection at step 218 and related subsequent operations may not be performed until after the hail is detected at step 204. In some example embodiments, the hail severity level provided by the fusing operation at step 222 is transmitted, for example, to other luminaires and / or insurance companies only if hail is detected at step 204. Alternatively, hail severity level may be estimated at steps 206, 216, 220 and the hail severity level determined at step 222 is transmitted to other luminaires and / or to insurance companies regardless of whether hail is detected at step 204. In some example embodiments, unlabeled and labeled training data used to train some or all of the Al models executed by the luminaire 102 may be obtained for the particular type of the luminaire 102. In general, the luminaires 104-120 may operate in the manner described herein with respect to the luminaire 102.

[0061] In some alternative embodiments, the method 200 may include other operations than shown in FIG. 2 without departing from the scope of this disclosure. In some alternative embodiments, one or more operations of the method 200 may be omitted without departing from the scope of this disclosure. For example, estimating hail intensity using information from step 202 may be omitted. In some alternative embodiments, Autoregressive Predictive Coding (APC) based model instead of AST-based model may be used to detect hail at step 204 and to estimate the severity level of the hail at step 206 using time series audio data without departing from the scope of this disclosure.

[0062] FIG. 3 illustrates a block diagram of an outdoor luminaire 300 of the outdoor luminaire system 100 of FIG. 1 according to an example embodiment. In general, the luminaire 300 may correspond to each one of the luminaires 102-120 of the system 100. Referring to FIGS. 1-3, in some example embodiments, the luminaire 300 may each include a processor 302 (e.g., a microprocessor), a communication interface unit 316, and a memory device 318 (e.g., a nonvolatile memory device such as a flash memory device). The processor 302 may execute software code stored in the memory device 318 to perform operations described herein with respect to the luminaires 102-120. For example, Al models 322 described in this description may be stored in the memory device 318. The communication interface unit 316 may transmit and receive signals complaint with one or more wireless communication standards. For example, the communication interface unit 316 may transmit and receive Bluetooth Low Energy (BLE) signals, Wi-Fi signals, and / or other wireless signals. The processor 302 may use the communication interface unit 316 to transmit information, for example, to other luminaires, the cloud, insurance companies, and / or other destinations. The processor 302 may also receive information through the communication interface unit 316, for example, from other luminaires.

[0063] In some example embodiments, the luminaire 300 may include a light module 304 that provides a light provided by the luminaire 300. For example, the luminaire 300 may be a streetlight that is attached to a pole. The processor 302 may control the operation of the light module 304 that may, for example, include one or more light sources such as light emitting diode light sources. The luminaire 300 may further include one or more microphones 306 (e.g., an array of microphones) that can capture sounds. For example, the microphones 306 may sense sounds produced by hail stones impacting the housing and other parts of the luminaire 300. The microphones 306 may also capture sounds such as wind gust sounds, automobile traffic sounds. The microphones 306 may also capture sounds produced by hails stones impacting the ground or roofs of nearby houses. The processor 302 may process audio data derived from the sounds captured by the microphones 306. For example, the processor 302 may filter out noise such as wind gust sounds and traffic sounds from the sounds captured by the microphones 306. The processor 302 may execute Al models as described with respect to FIGS. 1 and 2 to perform operations shown in FIG. 2 to detect hail and estimate hail intensity and hail severity level based on the sounds captured by the microphones 306 and filtered to remove noise. If no wind or low speed wind is detected and if moving cars are present in the vicinity of the luminaire 300 (e.g., the luminaire 102 installed as shown in FIG. 2), filtering the sounds captured by the microphones 306 may not be done. In some example embodiments, the luminaire 300 may include one or more radar devices 308 that can transmit and receive radar signals that can be used to detect people, cars and determine the locations and speeds of cars, etc. For example, the processor 302 can use information from the radar devices for hail detection as described with respect to step 204 in FIG. 2, to measure terminal velocity of hail stones as described with respect to step 214 of FIG. 2, and to determine traffic behavior patterns as described with respect to step 218 of FIG. 2.

[0064] In some example embodiments, the luminaire 300 may include a photosensor 310 to detect light level. The processor 302 may use the light level information from the photosensor 310 to confirm the hail detection performed at step 204 of FIG. 2. To illustrate, because ambient light level is relatively low during hail events, the processor 302 can invalidate hail detection at step 204 if the light level indicated by the photosensor 310 exceeds a threshold light level.

[0065] In some example embodiments, the luminaire 300 may include a barometer 312 to measure atmospheric pressure. The processor 302 may use the atmospheric pressure information from the barometer 312 to confirm or invalidate the hail detection performed at step 204 of FIG. 2. To illustrate, because atmospheric pressure is relatively lower during hail events, the processor 302 can invalidate hail detection at step 204 if the atmospheric pressure indicated by the barometer 312 exceeds a threshold atmospheric pressure.

[0066] In some example embodiments, the luminaire 300 may include a vibration sensor 314 to measure vibrations that may be caused by wind, cars, etc. The processor 302 may use the vibration information from the vibration sensor 314 to filter out noise from the sounds captured by the microphones 306 to determine whether the filtered sounds correspond to a hail event. The processor 302 may also discard audio data derived from sounds captured while vibrations exceeding a threshold are sensed by the vibration sensor 314.

[0067] In some example embodiments, the luminaire 300 may include an accelerometer 320 to measure the acceleration of hail stones. The processor 302 may use the acceleration information from the accelerometer 320 to determine sizes (e.g., diameter) of hail stones by determining the kinetic energy of the hail stones as can be readily understood by those of ordinary skill in the art with the benefit of this disclosure.

[0068] In some example embodiments, the luminaire 300 may include a driver that may receive AC power, for example, from a municipality power line, and provide DC power to the components of the luminaire 300 as can be readily understood by those of ordinary skill in the art with the benefit of this disclosure. In some alternative embodiments, the luminaire 300 may include more or fewer components than shown in FIG. 3 without departing from the scope of this disclosure. In some alternative embodiments, some of the components of the luminaire 300 may be integrated into a single component without departing from the scope of this disclosure.

[0069] FIG. 4 illustrates a method 400 of detecting hail and estimating hail severity by the luminaire system 100 of FIG. 1 according to an example embodiment. As described above, the luminaire 300 of FIG. 3 corresponds to each luminaire of the luminaires 102-120 of the system 100, and the components of the luminaire 300 correspond to respective components of the luminaires 102-120. Referring to FIGS. 1-4, in some example embodiments, at step 402, the method 400 includes performing hail detection at least based on sounds captured by one or more microphones of an outdoor luminaire. For example, the processor of a first luminaire (e.g., the luminaire 102) may detect hail based on sounds produced by hail and captured by the microphones of the first luminaire. To illustrate, the processor of the luminaire 102 may detect hail from the sounds produced by hail stones of the hail impacting, for example, the housing of the luminaire 102. The processor may execute an Al model to classify the sounds, for example, using time series audio data derived from the sounds.

[0070] In some example embodiments, at step 404, the method 400 includes estimating, by the first luminaire (e.g., the luminaire 102) a first severity level of the hail based on a terminal velocity of hail stones of the hail and an intensity level of the hail, based on the sounds, and / or based on traffic behavior patterns. For example, the processor of the luminaire 102 may execute a second Al model to estimate the severity level of hail based on a terminal velocity of hail stones of the hail and based on an intensity level of the hail. The severity level of hail refers to categories of the hail severity such as low severity, medium severity, high severity, and / or other categories as may be defined to describe different degrees of severity. The intensity level of the hail refers to number of hail stones, for example, in reference to a time period (e.g., 1 second or 1 minute) and an area (e.g., square feet). The terminal velocity of the hail stones and the intensity level of the hail are determined based on at least radar signals emitted by one or more radar devices (e.g., radar devices corresponding to the radar devices 308 of the luminaire 300) of the outdoor luminaire (e.g., the luminaire 300).

[0071] At step 406, the method 400 includes classifying precipitation elements as ice pellets, graupel, or the hail stones based on the sounds and the radar signals. For example, the processor of the luminaire 102 may execute an Al model to classify audio data derived from the sounds and radar sensing data into categories of ice pellets, graupel, or the hail stones. For example, because the ice pellets may not produce a sound that is easy to detect with a microphone, the radar signals may be more effective to detect ice pellets. In contrast, sounds produced by hail stones can be more readily useful to detect hail stones. The Al model executed by the processor of the luminaire 102 may be trained using training data collected for the specific type of the luminaire 102.

[0072] In some example embodiments, at step 408, the method 400 includes estimating by a second luminaire (e.g., the luminaire 104) a second severity level of the hail. For example, the processor of the luminaire 104 may execute one or more Al models to estimate the second severity level of the hail. For example, the severity level of the hail detected by the luminaire 102 (i.e., the first luminaire) may be Hail Severity Level #1 in FIG. 2, and Hail Severity Level #2 in FIG. 2 may correspond to the severity level of the hail estimated by the luminaire 104. The first luminaire (e.g., the luminaire 102) and the second luminaire (e.g., the luminaire 104) may be within a vicinity of each other such that the first luminaire and the second luminaire are expected to be exposed to intensity levels of the hail that are expected to correspond to the same hail severity level such as low severity, medium severity, or high severity. In general, the luminaires 102-112 may be expected to experience intensity levels of the hail that correspond to a particular hail severity level, and the luminaires 114-120 may be expected to experience intensity levels of the hail that correspond to another hail severity level. For example, such expectations may be based on historical hail intensity level data from previous hail events and / or based on weather conditions and patterns. Historical hail intensity level data may account for obstructions of luminaires by trees and other objects or structures.

[0073] In some example embodiments, at step 410, the method 400 includes deriving an overall severity level of the hail based on the first hail level and the second hail level. For example, the overall severity level of the hail may be derived from Hail Severity Level #1 and Hail Severity Level #2 shown in FIG. 2 in the manner described with step 224 of the method 200 of FIG. 2.

[0074] In some example embodiments, at step 412, the method 400 includes sending / transmitting a notification of the hail detection. For example, the processor of the luminaire 102 may send a notification of hail detection via the communication interface unit of the luminaire 102 to a traffic control authority, insurance companies, etc. To illustrate, a traffic control authority and insurance companies may, for example, send messages to drivers to avoid the area of hail detection. Each luminaire of the system 100 that detects hail may send a respective notification of hail detection. The luminaires may also send location information that may help identify the areas experiencing a hail event. Alternatively, one of the luminaires 102-120 of the system or another component (e.g., a server) of the system 100 may send the notification.

[0075] In some example embodiments, at step 414, the method 400 includes sending / transmitting information indicating the first severity level, the second severity level, and / or the overall severity level of the hail. For example, the luminaire 102 may transmit information indicating the severity level of the hail as estimated by the luminaire 102 independent of severity levels estimated by other luminaires of the system 100. As another example, the luminaire 104 may transmit information indicating the severity level of the hail as estimated by the luminaire 104 independent of severity levels estimated by other luminaires of the system 100. Alternatively, the luminaire 102 or another component (e.g., a server) of the system 100 may transmit the overall severity level of the hail derived at step 410.

[0076] In some example embodiments, the method 400 includes filtering out noise from the sounds. For example, the processor of each one of the luminaires 102-120 may filter out wind gust sound and traffic sound from the sounds captured by the microphones of the particular luminaire.

[0077] In some example embodiments, the method 400 may include determining audio sensing and / or radar sensing by the luminaires 102-120 of the system 100 is obstructed by an object such as a tree branch or a building. For example, if the luminaire 102 is obstructed by the tree 142 as shown in FIG. 1 such that audio sensing and radar sensing by the luminaire 102 cannot be performed reliably, hail detection and / or estimating severity level of hail may not be performed by the luminaire 102 or such information from the luminaire 102 may be unused (e.g., not sent to insurance companies, not combined with information from other luminaires, etc.).

[0078] In some example embodiments, an Al model may be trained with audio data corresponding to different levels of blockage of a luminaire, for example, by a tree branch or another object or structure. During a hail even, each one of the luminaires 102-120 may execute the Al model to identify the degree of blockage of the particular luminaire. To be clear, different versions of the Al models may be trained for different types (e.g., different SKU numbers) of luminaires. Based on the blockage information, hail detection or severity level information from some luminaires from among the luminaires 102-120 may be deemed unreliable and thus may not be used. In some example embodiments, traffic behavior patterns determined using radar signals may be used to estimate the dominant size (e.g., small, medium, or large) of hail stones of hail. For example, because the behavior of drivers during a hail event depends on the dominant size of hail stones, an Al model may be trained with labeled traffic behavior pattern data. Durin g inference, the processor of each luminaire of the system 100 may estimate the dominant size of hail stones. The information may be sent, for example, to insurance companies that can use the information to estimate or refine hail damage assessment.

[0079] In some alternative embodiments, the method 400 may include other steps without departing from the scope of this disclosure. In some alternative embodiments, some of the steps of the method 400 may be performed in a different order than shown without departing from the scope of this disclosure. In some alternative embodiments, some of the steps of the method 400 may be omitted without departing from the scope of this disclosure.

[0080] Although particular embodiments have been described herein in detail, the descriptions are by way of example. The features of the example embodiments described herein are representative and, in alternative embodiments, certain features, elements, and / or steps may be added or omitted. Additionally, modifications to aspects of the example embodiments described herein may be made by those skilled in the art without departing from the scope of the following claims, the scope of which are to be accorded the broadest interpretation so as to encompass modifications and equivalent structures.

Claims

CLAIMS1. A hail detection and analysis method, comprising: performing (204), by a processor (302) of an outdoor luminaire (102, 300), a hail detection at least based on sounds captured by one or more microphones (306) of the outdoor luminaire, wherein the processor is configured to execute a first artificial intelligence (Al) model (322) to classify the sounds to detect hail; and estimating (216), by the processor, a severity level of the hail at least based on a terminal velocity of hail stones of the hail and an intensity level of the hail, wherein the terminal velocity of the hail stones and the intensity level of the hail are determined based on at least radar signals emitted by one or more radar devices (308) of the outdoor luminaire, and wherein the processor is configured to execute a second Al model (322) to estimate the severity level of the hail.

2. The method of Claim 1, wherein the sounds are produced at least partially by the hail stones impacting the outdoor luminaire, wherein estimating the severity level of the hail is performed further based on the sounds, and wherein estimating the severity level of the hail is performed after detecting the hail based on the sounds, and wherein estimating the severity level of the hail is performed after the hail is detected based on the sounds.

3. The method of Claim 2, wherein estimating the severity level of the hail is performed further based on traffic patterns of automobiles that are within a radar detection range of the outdoor luminaire.

4. The method of Claim 3, wherein the severity level of the hail is derived from a first severity level, a second severity level, and a third severity level, wherein the first severity level is estimated (216) based on the terminal velocity of the hail stones and the intensity level of the hail, wherein the second severity level is estimated (206) based on the sounds, and wherein the third severity level of the hail is estimated (220) based on the traffic pattern of the automobiles.

5. The method of Claim 2, further comprising filtering out wind sounds and traffic sounds from the sounds.

6. The method of Claim 1, further comprising: transmitting (412) a notification indicating a detection of the hail; and transmitting ( 14) information indicating the severity level of the hail.

7. The method of Claim 6, wherein the information includes a number of automobiles detected using the radar signals during the hail.

8. The method of Claim 1, further comprising classifying precipitation elements as ice pellets, graupel, or the hail stones based on the sounds and the radar signals.

9. The method of Claim 1, further comprising estimating (408), by a processor (302) of a second outdoor luminaire (104, 300), a second severity level of the hail (Hail Severity Level #2), wherein the outdoor luminaire (102) and the second outdoor luminaire (104) are within a vicinity of each other such that the outdoor luminaire and the second outdoor luminaire are expected to experience intensity levels of the hail that correspond same severity level of the hail.

10. The method of Claim 9, further comprising: deriving (224, 410), by the processor (302) of the outdoor luminaire (102, 300), an overall severity level of the hail from at least the severity level of the hail (Hail Severity Level #1) and the second severity level of the hail (Hail Severity Level #2); and send (414), by the outdoor luminaire (102, 300), information indicating the overall severity level of the hail.

11. An outdoor luminaire system (100) for hail detection and hail severity estimation, the system comprising: an outdoor luminaire (102, 300) comprising a processor (302), one or more radar devices (308), and one or more microphones (306), wherein the processor is configured to: execute a first Al model (322) to detect hail at least by classifying sounds captured by the one or more microphones (306); andexecute (226) a second Al model (322) to estimate a severity level of the hail at least based on a terminal velocity of hail stones of the hail and an intensity level of the hail, wherein the terminal velocity of the hail stones and the intensity level of the hail are determined based on at least radar signals emitted by the one or more radar devices (308).

12. The outdoor luminaire system of Claim 11, wherein the processor (302) is configured to: estimate (206) a second severity level of the hail further based on the sounds; estimate (220) a third severity level of the hail further based on traffic patterns of automobiles that are within a radar detection range; and derive (222) a per-luminaire severity level of the hail from the severity level, the second severity level, and the third severity level.

13. The outdoor luminaire system of Claim 12, wherein the outdoor luminaire is configured to transmit a notification of the hail detection.

14. The outdoor luminaire system of Claim 11, further comprising a second outdoor luminaire (104, 300) that comprises a second processor (302), second one or more radar devices (308), and second one or more microphones (306), wherein the second processor is configured to detect the hail and to estimate a second severity level of the hail.

15. The outdoor luminaire system of Claim 14, the processor is configured to derive an overall severity level of the hail from at least the severity level of the hail and the second severity level of the hail and send information indicating the overall severity level of the hail.

Images