Asset-level vulnerability and mitigation

Abstract

To obtain an insight into the hazard vulnerability of a parcel / asset.SOLUTION: There are provided methods, systems, and apparatus for receiving a request for a damage propensity score for a parcel, receiving imaging data for the parcel, wherein the imaging data comprises street-view imaging data of the parcel, extracting, by a machine-learned model including multiple classifiers, characteristics of vulnerability features for the parcel from the imaging data, determining, by the machine-learned model and from the characteristics of the vulnerability features, a damage propensity score for the parcel, and providing a representation of the damage propensity score for display.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] Cross - Reference to Related Applications This application claims priority to U.S. Patent Application No. 17 / 158,585, titled "ASSET - LEVEL VULNERABILITY AND MITIGATION," filed on January 26, 2021, the disclosure of which is incorporated herein by reference.

Background Art

[0002] As land development encroaches on the interface between wilderness and urban areas and environmental changes bring about long - term droughts, wildfires are becoming an increasingly serious problem. Insurers and risk assessment managers look at the various assets present in a parcel and use regression techniques and known vulnerabilities to generate wildfire risk assessments. Generating a risk assessment for a parcel may require asset inspections and other one - time static appraisals, and as a result, changes to the parcel can create a need for costly updated re - evaluations to continue to recognize the current risk.

Summary of the Invention

[0003] This specification describes systems, methods, devices, and other techniques related to using machine learning to obtain insights about the hazard vulnerability of a parcel / asset from imaging data that captures the parcel / asset.

[0004] Generally, one innovative aspect of the subject matter described in this specification can be embodied in a method of receiving a request for a damage propensity score for a parcel and receiving imaging data for the parcel, where the imaging data includes street - view imaging data of the parcel. A machine - learning model that includes a plurality of classifiers extracts the characteristics of a plurality of vulnerability features of the parcel from the imaging data and determines a damage propensity score for the parcel from the characteristics of the plurality of vulnerability features. A representation of the damage propensity score is provided for display.

[0005] These and other implementations can each optionally include one or more of the following features. In some embodiments, the method further includes generating a set of characteristic partitions from the characteristics of a plurality of vulnerability features.

[0006] In some embodiments, the method further includes generating a three-dimensional model of the partition from the characteristics and imaging data of the partition of a plurality of vulnerability features.

[0007] In some embodiments, the imaging data of the partition includes the imaging data captured within a time threshold from the requested time.

[0008] In some embodiments, receiving a request for a damage propensity score includes receiving hazard event data of a hazard event and determining a damage propensity score of the partition for the hazard event from the characteristics of a plurality of vulnerability features and the hazard event data of the hazard event. The method can further include receiving updated hazard event data for the hazard event and determining an updated damage propensity score of the partition for the hazard event from the characteristics of a plurality of vulnerability features, the hazard event data, and the updated hazard event data.

[0009] In some embodiments, the method further includes determining, by a machine learning model, one or more mitigation steps for the partition, determining, by the machine learning model, an updated damage propensity score based on the one or more mitigation steps, and providing a representation of the one or more mitigation steps and the updated damage propensity score. The one or more mitigation steps can include adjustments to the characteristics of a plurality of vulnerability features extracted from the imaging data.

[0010] In some embodiments, determining one or more mitigation steps further includes iteratively determining an updated damage propensity score based on adjusted characteristics of a plurality of vulnerability features. In some embodiments, determining an updated damage propensity score further includes determining that the updated damage propensity score meets a threshold damage propensity score.

[0011] In some embodiments, determining one or more mitigation steps includes determining one or more mitigation steps for a particular type of hazard event, and the one or more mitigation steps for a first type of hazard event are different from the one or more mitigation steps for a second type of hazard event.

[0012] In some embodiments, the method includes generating training data for a machine learning model, including receiving, for a hazard event, a plurality of sections located within the vicinity of the hazard event, each of the plurality of sections having received at least a threshold exposure to the hazard event; receiving, for each of the plurality of sections, imaging data of the section including street view imaging data; and extracting, from the imaging data, characteristics of a plurality of vulnerability features of a first subset of the plurality of sections that did not burn during the hazard event and a second subset of the plurality of sections that burned. The method further includes providing the training data to the machine learning model.

[0013] In some embodiments, extracting characteristics of a plurality of vulnerability features includes providing the imaging data to a plurality of classifiers. Extracting characteristics of a plurality of vulnerability features can include identifying a plurality of objects within the imaging data by the plurality of classifiers.

[0014] In some embodiments, the method further includes, for each of a plurality of compartments, receiving additional structural characteristics, and from the additional structural characteristics, extracting a second set of a plurality of vulnerability characteristics for a first subset of the plurality of compartments that did not burn during a hazard event and a second subset of the plurality of compartments that burned, and providing the second set of the plurality of vulnerability characteristics to a machine learning model.

[0015] In some embodiments, the additional structural characteristics include a post-hazard event inspection of the plurality of compartments.

[0016] The present disclosure also provides a non-transitory computer-readable storage medium coupled to one or more processors and storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations in accordance with implementations of the methods provided herein.

[0017] It will be understood that the methods and systems according to the present disclosure can include any combination of the aspects and features described herein. That is, the methods and systems according to the present disclosure are not limited to the combinations of aspects and features specifically described herein, but also include any combination of the provided aspects and features.

[0018] Certain embodiments of the subject matter described herein can be implemented to realize one or more of the following advantages. The advantages of the present technology can develop a new understanding of hazard vulnerability for a substantially larger number of vulnerability characteristics than conventional methods that use a trained machine learning model that takes into account the configuration of the characteristics of the vulnerability characteristics according to a specific set of hazard conditions and exposure levels, which can be more complex than the sum of the risk factors and can reflect non-obvious characteristics that contribute to the degree of damage or damage / no-damage results. The assessment of hazard vulnerability for a particular hazard and exposure level can be determined for a compartment using imaging data and may not require additional asset inspections. The hazard vulnerability assessment can be utilized in determining asset valuation, sales, taxes, and the like.

[0019] By using the street view images of the blocks, access can be obtained to the unique features of the blocks that are not available in other ways using other imaging data, such as features that reflect the current state of the houses on the blocks (e.g., ivy growing on the side of the house, the position of the cars parked on the private road). For example, in real-time hazard conditions where the mitigation response can be updated as the hazard conditions change to identify vulnerable blocks based on the respective vulnerability of each block under the current conditions of the hazard event, a vulnerability trend score can be determined. Optimized mitigation steps, such as risk reduction plans and / or cost-benefit estimations, can be determined in an iterative process by a trained machine learning model based on the extracted characteristics of the vulnerability features of the blocks and in response to the hazard event.

[0020] The uses of the present technology generally include insurance risk assessment, real-time risk assessment and response, and general natural disaster hazard assessment and mitigation. More specifically, the present technology can be utilized by local governments, states, or central governments to more accurately conduct risk assessments and design and formulate risk mitigation plans.

[0021] Details of one or more embodiments of the subject matter described in this specification are set forth in the accompanying drawings and the following description. Other features, aspects, and advantages of the subject matter will become apparent from the description, drawings, and claims.

Brief Description of the Drawings

[0022]

Figure 1

Figure 2A

Figure 2B

Figure 2C

Figure 3

Figure 4

Figure 5

[0023] Similar reference numerals and names in the various drawings indicate similar elements. **DETAILED DESCRIPTION**

[0024] Overview The technology of this patent application aims to utilize machine learning to obtain insights into the hazard vulnerability of compartments / assets from imaging data that captures the compartments / assets.

[0025] More specifically, the technology of this application utilizes a trained machine learning model to identify vulnerability features and identify the characteristics of vulnerability features within imaging data of compartments, such as street view imaging data, LIDAR data, high-resolution satellite image data, aerial image data, infrared image data, user-provided images, etc. The characteristics of the vulnerability features can be used to generate a vulnerability tendency score and / or identify mitigation strategies for reducing the hazard vulnerability of a specific compartment according to the degree of specific hazards and exposures.

[0026] Generating training data for training the machine learning model can include selecting a set of compartments located within the vicinity of a hazard event, for example, within the radius of a burn scar. The hazard event can be, for example, wildfire, flood, tornado, etc., and each of the set of compartments has experienced a certain degree of exposure to the hazard event. For each compartment in the set of compartments, imaging data that captures the compartment before the hazard event, such as a photo of a house / asset before experiencing a wildfire, is collected.

[0027] The vulnerability characteristics of a section can be defined using existing risk assessment data, such as defensible space, building structure, or other characteristics known to be related to increases / decreases in hazard vulnerability. The vulnerability characteristics can be further extracted from imaging data showing sections with damage / no-damage results and / or damage severity results for a specific hazard event, and one or more neural networks can be utilized to process the imaging data and extract additional vulnerability characteristics determined to distinguish the damage / no-damage results and / or damage severity results of the section.

[0028] Characteristics of the vulnerability characteristics, such as the material of the roof structure, the distance between the tree and the house, the manufacturing information of the building materials, the frontage, the type of fence, irrigation, etc., can be extracted from the imaging data of the section using multiple classifiers and object recognition technology. Training data can be generated for multiple sets of sections and each hazard event, and can include the extracted characteristics of the vulnerability characteristics, the location of the section relative to the hazard event, the degree of exposure / damage during the hazard event, and the damage / no-damage results. In addition, public records of the section, information about the hazard event, etc. can be utilized when generating training data for training the machine learning model.

[0029] The trained machine learning model can receive the requirements for the vulnerability assessment of a specific section and the requirements for a hazard event including the degree of exposure. The imaging data can be collected for the section using, for example, the known address, geographical location, etc. of the section, and can include only the imaging data captured within a time threshold, for example, collected within the past six months. The imaging data can reflect the current state of the section, such as the current state of the vegetation surrounding the section, the location of the vehicle, the structures built within the section (e.g., shed), etc. The machine learning model can receive the imaging data, public records (e.g., year of construction, setback, special permit, etc.), and other relevant geospatial information (e.g., density of neighboring houses, distance to the fire station / emergency services, distance to the main road, etc.) as inputs and extract the characteristics of the vulnerability characteristics of the section. The determined vulnerability tendency score can be provided as an output.

[0030] In some embodiments, the model can determine mitigation steps for reducing the risk score for a section based on the vulnerability characteristics of the section. Determining the mitigation steps by a machine learning model can include identifying the characteristics of the vulnerability features that can be adjusted (e.g., mowing overgrowth, changing roofing materials, changing siding materials), and iterating the risk score determination based on the adjusted characteristics of the vulnerability features. The permutations of the mitigation steps can be evaluated for various hazard event scenarios to provide an optimized subset of the mitigation steps for a particular section.

[0031] In some embodiments, the real-time hazard vulnerability can be determined for a particular section based on real-time hazard events or potential future hazard risks, such as ongoing wildfires, drought occurrences, severe weather patterns, etc. As the hazard event progresses, e.g., as the degree of exposure changes, the hazard vulnerability of the section is updated, and in response, real-time alerts can be generated, e.g., notifying the homeowner or emergency responders of the real-time hazard vulnerability and / or countermeasures.

[0032] Exemplary Operating Environment FIG. 1 is a block diagram of an exemplary operating environment 100 of a hazard vulnerability system 102. The hazard vulnerability system 102 can be hosted on one or more local servers, cloud-based services, or a combination thereof.

[0033] The hazard vulnerability system 102 can communicate with a network, and the network can be configured to enable the exchange of electronic communications between devices connected to the network. The network may include, for example, the Internet, Wide Area Networks (WANs), Local Area Networks (LANs), analog or digital wired and wireless telephone networks (e.g., public switched telephone network (PSTN), Integrated Services Digital Network (ISDN), cellular network, and Digital Subscriber Line (DSL)), wireless, television, cable, satellite, or any other delivery or tunneling mechanism for carrying data, and may include one or more of them. The network may include multiple networks or sub-networks, each of which may include, for example, wired or wireless data paths. The network may include a circuit-switched network, a packet-switched data network, or any other network capable of carrying electronic communications (e.g., data or voice communications). For example, the network may include a packet-switched network based on Internet protocol (IP), asynchronous transfer mode (ATM), PSTN, IP, X.25, or frame relay, or a network based on other equivalent technologies, and may support voice using, for example, VoIP or other equivalent protocols used for voice communications. The network may include one or more networks including wireless data channels and wireless voice channels. The network may be a wireless network, a broadband network, or a combination of networks including a wireless network and a broadband network.

[0034] The hazard vulnerability system 102 includes a training data generator 104 and a damage tendency model 106. Optionally, the hazard vulnerability system 102 includes a mitigation engine 108. The training data generator 104, the damage tendency model 106, and the mitigation engine 108 are described herein, but the operations described can be performed by more or fewer subcomponents.

[0035] The training data generator 104 includes a vulnerability feature extractor 110, a compartment hazard event module 112, and a vector generator 114. The training data generator 104 receives imaging data 116 from a repository such as satellite and / or aerial images 118, street view images 117, and provides the training data 120 as an output. The output training data 120 can be used to train the damage tendency model 106.

[0036] The damage tendency model 106 includes a plurality of classifiers 107, such as one or more neural networks or machine learning models, such as random forests. The classifier can be configured to classify the damage tendency as a binary result, e.g., damaged or non-damaged, or, for example, using a regression task, can be configured to classify the degree of damage tendency. In some embodiments, the classifier can be used to estimate damage to specific subcomponents of a compartment, such as the roof of a building, the siding of a building, etc., to further refine the damage tendency model 106.

[0037] The damage propensity model 106 can receive training data 120 including a significant number of training vectors generated using a large sample of different hazard events documented in the imaging data 116 and the historical hazard event data 124. The damage propensity model 106 can be trained to infer the damage propensity of a particular section, based in part on the characteristics of the vulnerability features of the section. The plurality of classifiers 109 can include, for example, the same set or a different set of classifiers as described with reference to classifier 107, and can process the received imaging data 116 to identify and label the characteristics of the vulnerability features extracted from the imaging data 116.

[0038] The satellite / aerial image 118 includes any image that captures a geographic area and provides information about that geographic area. The information about the geographic area can include, for example, information about one or more sections located in the geographic area, such as structures, vegetation, terrain, etc. The satellite / aerial image can be, for example, a Landsat image, or other forms of aerial images. The satellite / aerial image 118 can be, for example, an RGB image or a hyperspectral image. The satellite / aerial image 118 can be captured using satellite technology, such as Landsat, or drone technology. In some implementations, the satellite / aerial image can be captured using other high-altitude technologies, such as drones, weather observation balloons, airplanes, etc. In some embodiments, in addition to satellite images as described herein, synthetic aperture radar (SAR) images can be utilized.

[0039] In some implementations, the satellite image or other aerial image can be captured using radar-based imaging, such as LIDAR images, RADAR images, or other types of imaging using the electromagnetic spectrum, or combinations thereof. The satellite / aerial image 118 can include images of geographic areas containing various natural features, including different terrains, vegetation, water bodies, and other features. The satellite / aerial image 118 can include images of human development, such as housing construction, roads, dams, retaining walls, etc.

[0040] The street view image 117 includes any image that captures the aspect of one or more plots from a frontage perspective, captured, for example, from a road or sidewalk facing the plot. In some embodiments, the street view image 117 can be captured by one or more cameras attached to a vehicle and configured to capture the street view image 117 of the plot as the vehicle passes through the plot. Depth information regarding the plot can be captured using optical and LIDAR street view images 117. In some embodiments, the street view image 117 can have a high spatial resolution. For example, the street view image can have a spatial resolution of less than 1 centimeter.

[0041] In some embodiments, the street view image 117 can be captured by a user, e.g., a plot homeowner, from the frontage view of the asset. The street view image can be captured using a built-in camera on a smart device, e.g., a smartphone or tablet, and / or using a handheld camera. In some embodiments, the street view image can be captured by a land surveyor, an insurance appraiser, a plot appraiser, or other person documenting the aspect of the plot.

[0042] In some embodiments, the street view image 117 can include additional views of the plot, e.g., views captured from the side of the plot, the back of the plot. For example, the user can capture a street view image 117 of the backyard of their asset, or a side view of their asset.

[0043] In some embodiments, the imaging data 116 can include images of sections before and after a hazard event, e.g., before and after a wildfire. The imaging data 116 associated with a particular hazard event can include burn scars, e.g., areas damaged by the hazard event. Further description of the imaging data 116 is presented with reference to FIGS. 2A - 2C.

[0044] The training data generator 104 receives the section data 122 from a repository of historical hazard event data 124. The section data 122 can include, for example, insurance appraisals, land surveys, appraisals, construction / construction records, code inspections / violations, and other public records.

[0045] In some embodiments, the section data 122 includes public records from post - hazard damage reports, such as from damage inspections (DINS) databases maintained by CalFIRE, from post - hazard insurance inspections, etc. The post - hazard damage report can include a direct inspection of structures exposed to a hazard event, e.g., a wildfire, and can include information regarding the vulnerability characteristics of the structures, e.g., roof type, eaves type, building materials, etc., as well as the damage level. The section data 122 can include damage / no - damage data and / or damage severity data for sections within the radius of the hazard event, e.g., within the radius of the burn radius.

[0046] The training data generator 104 receives the imaging data 116 as input. The imaging data 116 captures one or more sections at a particular location, e.g., at one or more houses, and at a particular point in time, e.g., before or after a hazard event. For example, the street view image 117 can capture a house at a particular street address and a first date / time, e.g., before a hazard event.

[0047] The vulnerability feature extractor 110 can include a plurality of classifiers 107, and the plurality of classifiers can identify vulnerability features, such as objects, within the imaging data 116. For each section shown in the imaging data 116, the vulnerability feature extractor 110 can extract the vulnerability features and provide the vulnerability features F1, F2,... FN of the section as an output to the vector generator module 114. Continuing with the above example, the vulnerability features F1, F2,... FN are extracted for a house at a specific location address and a first date / time and are, for example, the roof structure, vegetation, frontage distance, land gradient, etc.

[0048] The vulnerability features can include, but are not limited to, building materials, defensible space, section gradient, proximity to a road, etc. The vulnerability features can further include objects, such as trees, vehicles, etc. In some embodiments, ground truth labeling can be utilized to identify the vulnerability features, for example, by a human expert or in an automatic / semi-automatic manner. The vulnerability features utilized by an insurance appraiser / risk assessment manager can be identified in the imaging data 116.

[0049] The vulnerability characteristics can further include the vulnerability characteristics of the sections captured within the imaging data 116 that were not conventionally identified as risk hazards by insurance appraisers / risk assessments. In other words, characteristics of sections that may not have been conventionally labeled as hazard risks can be extracted as potential vulnerability characteristics to generate training data. For example, the structure of a private road, the distance between a parked car and a house, the type of grass seed used in a lawn, etc. The damage tendency model 106 can define vulnerability characteristics extracted from the imaging data 116 that are significant, for example, by conventional means and may not be predicted in other ways, as significant, whereby the vulnerability characteristics can be processed by a machine learning model to determine which of the potential vulnerability characteristics are significant in the damage tendency of the section, for example, have a statistical effect on the damage / no damage result and / or the damage severity result. In this way, novel and non-obvious characteristics can be identified as having significance in the damage tendency.

[0050] Each vulnerability characteristic of the section describes the aspect of the section shown within the imaging data 116, for example, within the street view image 117 and / or the satellite / aerial image 118. For each vulnerability characteristic extracted from the imaging data 116, one or more characteristics C1, C2,... CN of the vulnerability characteristic, for example, F1{C1, C2,... CN} are extracted. Further details of the feature extraction will be described with reference to FIGS. 2A - 2C. The characteristics of the vulnerability characteristic can include quantifiable and / or qualifiable aspects of the vulnerability characteristic. For example, the vulnerability characteristic of a roof structure can be characterized by the building material, the roof panel spacing, the age of construction, and the maintenance of the roof. In another example, the vulnerability characteristic of the slope of a section can be characterized by a slope measurement value of 0.5°.

[0051] The zoning hazard event module 112 receives, as inputs, historical hazard event data 124 including records of past hazard events, and zoning data 122 for each zone shown in the imaging data 116 processed by the vulnerability feature extractor 110. The zoning hazard event module 112 provides the zoning data 122 for the zone to the vector generator 114.

[0052] The historical hazard event data 124 can include the times of the hazard events, for example, the start time and end time of the hazard event. For example, the start time when a wildfire starts, and the end time when the wildfire is completely contained or completely extinguished. The historical hazard event data 124 can include the affected area affected by the hazard event, for example, the geographical location of the area including the burn scar, for example, GPS coordinates.

[0053] The zoning data 122 for a particular zone can include public records of the zone before and after the hazard event, for example, before and after a wildfire. For example, the zoning data 122 can include post-hazard insurance / appraisal records, for example, damage assessments from after the hazard event. In other words, the zoning data 122 can include the damage / non-damage results for the zone for a particular hazard event, for example, damage versus non-damage. The zoning hazard event module 112 can utilize the damage / non-damage results and / or damage degree results for a plurality of zones located within the combustion radius 208 of the hazard event as the ground truth of the training data 120.

[0054] In some embodiments, the zoning data 122 for a house can include the structural characteristics of the zone collected before the hazard event, for example, building records, building materials, roof type, and the like.

[0055] The training data generator 104 can generate training data from the image of the section using the imaging data 116 generated before and after the hazard event, and from the section data 122 of the section from the historical hazard event data 124 corresponding to before and after an event, such as a wildfire. The vulnerability feature extractor 110 can extract the vulnerability features and related characteristics of the sections in the imaging data 116 that each appear within the radius of the hazard event, for example, within the distance of the burn scar.

[0056] The vector generator 114 receives the vulnerability features and the characteristics of the vulnerability features extracted from the feature extraction module 110, and optionally receives the section data 122 from the section hazard event module 112 for a specific section as input. In some embodiments, the section data 122 can be used as the ground truth of the training data 120. For example, the section data 122 including the damage / non - damage result and / or the damage degree result for the hazard event can be used to label the section as either "burned" or "unburned".

[0057] The vector generator 114 can generate the training data 120 from the extracted vulnerability features, the characteristics of the vulnerability features, and the section data 122 of each section, for example, each training vector V. Further details of the generation of the training data will be described with reference to FIG. 4.

[0058] The damage tendency model 106 can receive the training data 120 as input for training a machine learning model, such as the damage tendency model 106, using the training data 120. In some implementations, the damage tendency model 106 can be trained using a significant number of training vectors generated using section data representing large samples at different locations and various historical hazard events. In one example, the training data 120 provided to the damage tendency model 106 can include thousands of sections that have experienced many different hazard events.

[0059] The hazard vulnerability system 102 receives a request 126 from a user of the user device 128. The user device 128 can include, for example, a mobile phone, a tablet, a computer, or another device including an operating system 129 and an application environment 130, through which the user can interact with the hazard vulnerability system 102. In one embodiment, the user device 128 is a mobile phone including an application environment 130 configured to display a view 132 including at least a portion of a section. In one embodiment, as shown in FIG. 1, the application environment 130 displays a view 132 including a street view of a house and surrounding assets such as, for example, trees, bushes, and shrubs.

[0060] The request 126 can include the location of the section specified by the user of the user device 128. The location of the section can include a geographical location, such as, for example, GPS coordinates, a location address, etc., and can be input by the user into the application environment 130.

[0061] The request 126 can further include a damage propensity score, such as, for example, a request for relative vulnerability to a hazard event, and the request 126 can specify a particular hazard event, such as, for example, a particular real-time hazard event, or a general type of hazard event, such as, for example, wildfire, flood, earthquake, etc. In one embodiment, the user can submit a request 126 specifying the location address of the section and request the damage propensity score of that section for a real-time hazard event, such as, for example, a wildfire that is occurring. In another embodiment, the user can submit a request 126 specifying the location of the section including GPS coordinates and request the damage propensity score specific to flooding of the section.

[0062] In some embodiments, the damage propensity score may be a relative measure of the risk that a particular section will be damaged by a hazard event. The damage propensity score may be a general measure of the risk for a particular section to be damaged by a certain type of hazard event, such as wildfire, or a specific measure of the risk for a particular section to be damaged by a specific hazard event, such as a real-time flood event.

[0063] In some embodiments, the damage propensity score may include the percentage of loss of a given section under a particular hazard scenario, such as the percentage of damage. For example, the damage propensity score may be a 10% loss for a particular section under a particular wildfire scenario.

[0064] In some embodiments, an end user, such as an asset owner, an insurance adjuster, a government official, etc., can provide a complete probabilistic hazard model, that is, the hazard characteristics and the associated probability distributions, whereby the hazard vulnerability system can provide the expected average annual loss (AAL) of a particular section.

[0065] In some embodiments, claim 126 can specify a location that includes a plurality of sections, such as a neighborhood, a street that includes a plurality of houses, a complex that includes a plurality of buildings, etc. The user may be interested in determining individual damage propensity scores for each structure in the location that includes a plurality of sections, or may be interested in determining a global damage propensity score for a plurality of sections.

[0066] In some embodiments, the hazard vulnerability system 102 receives, as input, a requirement 126 that includes a requirement for mitigation steps to reduce the hazard vulnerability of a compartment. The mitigation engine 108 receives the requirement for mitigation steps and can identify a set of mitigation steps 136 that a user, such as a homeowner, can take to reduce the hazard vulnerability of the compartment based on the imaging data 116 and the damage propensity score 140. The mitigation steps are quantifiable and / or quantifiable measures that can be taken by the user, for example, the homeowner, to reduce the damage propensity score 140. The mitigation steps can include, for example, removal / reduction of vegetation adjacent to the structure, building materials used in the structure, etc. For example, the mitigation step can be to trim the leaves within a radius of 2 feet surrounding the house. In another example, the mitigation step can be to change the siding material of the house. In yet another example, the mitigation step can be to dig irrigation ditches to collect the outflow water from the flooded stream area.

[0067] In some embodiments, the mitigation steps can be provided within an application environment 130 on the user device 128, and the mitigation steps 136 are visually identified, for example, an indicator is overlaid on the view 132 of the compartment. For example, a tree with branches overhanging the roof can be visually identified within the application environment using, for example, a box surrounding the tree and / or the branches.

[0068] In some embodiments, the mitigation engine 108 can receive real-time event data 142, e.g., real-time data about a hazard event occurring, and update the mitigation steps 136 in real time to provide the user with a real-time response to the hazard event. The real-time event data 142 can include, for example, the spread of the hazard, weather patterns, mitigation events, emergency responses, etc. For example, the real-time event data 142 about a wildfire can include the real-time perimeter of the fire, the percentage of control by firefighters, evacuation data, wind advisories, etc. In another example, the real-time event data 142 about a flood can include the real-time river / creek levels, flood levels, rain / weather forecasts, evacuation data, etc.

[0069] Feature extraction As described above with reference to FIG. 1, the vulnerability feature extractor 110 can receive imaging data 116, including satellite / aerial images 118 and street view images 117, as extracted vulnerability features having the input and related characteristics. FIG. 2A is a schematic diagram of an exemplary pair of satellite images including a plurality of sections before and after a hazard event. Satellite images 200a and 200b are captured at capture times T1 and T2 respectively, where T1 is a time occurring before the hazard event and T2 is a time occurring after the hazard event. Additionally, T1 and T2 can be selected based on the section data 122 of the sections included in the satellite images 200a, 200b, and the section data includes records of the sections at a time T1' before the hazard event and a time T2' after the hazard event.

[0070] Satellite images 200a and 200b show the same geographic area 202 including a set of sections 204. Satellite image 200a is captured at a time T1 occurring within a first threshold of the time before the start of the hazard event, and satellite image 200b is captured at a time T2 occurring within a second threshold of the time after the end of the hazard event.

[0071] Satellite image 200b captured after a hazard event includes burn marks 206 resulting from a hazard event, such as a fire. The burn marks 206 can indicate the area of the geographical region 202 damaged / affected by the hazard event. The burn radius 208 defines an outer perimeter that encompasses, surrounds, and buffers the burn marks 206 and includes additional areas extending outward from the burn marks, such as an additional 100 feet, an additional 1000 feet, and an additional 5000 feet. The burn radius 208 can include section A damaged / affected by the hazard event and section B not damaged / affected by the hazard event. Section A can be located within the burn marks 206 and be the section damaged / affected by the hazard event. Section B can be located outside the burn marks 206 but within the burn radius 208, or section B can be located within the burn marks 206 but not damaged / affected by the hazard event.

[0072] Satellite image 200b captured after a hazard event can include multiple burn marks 206 and burn radii, and the burn marks and / or burn radii can overlap with each other. The section can be located in the overlapping area of the burn marks and / or the overlapping area of the burn radii.

[0073] As described with reference to FIG. 1, the vulnerability feature extractor 110 receives satellite images 200a, 200b and extracts vulnerability features F1, F2,... FN and their respective characteristics from the images 200a, 200b. The vulnerability features extracted from the satellite images 200a, 200b can include, for example, the position of the section with respect to the roof structure, natural formations (e.g., forest / tree coverage, waterways, etc.), and artificial features (e.g., roads, irrigation ditches, farmland, etc.). Each characteristic can include, for example, the building materials for the roof structure, such as ceramic, metal, wood, etc. In another example, the characteristic of the position of the section with respect to the artificial feature can include the relative distance between the section and the artificial feature, such as the distance from a house to a street.

[0074] In some embodiments, for a plurality of sections appearing in a satellite image, vulnerability features can be extracted from the satellite image 118. Additional vulnerability features can be extracted for each of the sections of the plurality of sections using a higher resolution image, for example, using the street view image 117.

[0075] FIG. 2B is a schematic diagram of an exemplary pair of street view images of section A captured before and after a hazard event. The street view images 220a and 220b show, for example, the house 222 and the surrounding vegetation 224 located on section A, captured from street level and facing the section. The street view image 220b captured after the hazard event includes, for example, hazard event damage 226 to the house, an adjacent storage shed, and nearby trees.

[0076] Vulnerability features are extracted from the street view images 220a, 220b. As described with reference to FIG. 1, the vulnerability feature extractor 110 receives imaging data 116 including the street view image 117, for example, images 220a, 220b, and extracts vulnerability features F1, F2,... FN. Referring back to FIG. 2B, the vulnerability features extracted from the street view images 220a, 220b include a group of trees (F1), a group of thickets (F2), an outdoor shed (F3), and the roof structure of the house 222 (F4). More or fewer vulnerability features can be extracted from the street view images 220a, 220b, and the examples provided are not limiting.

[0077] For each of the vulnerability features extracted from the street view images 220a, 220b, the vulnerability feature extractor identifies the characteristics of each vulnerability feature. For example, a group of trees F1 can have associated quantifiable characteristics such as the distance of the trees to the house 222, the number of trees, the height of the trees, the proximity of the tree groups, etc., and quantifiable characteristics such as the soundness of the trees, being covered with ivy, etc. In another embodiment, the roof structure F4 can have associated characteristics such as building materials, eaves structure, roof slope, age of the roof, maintenance of the roof, fullness of the rain gutters, etc.

[0078] In some embodiments, the vulnerability feature extractor 110 can identify a section A that reflects damage from a hazard event, such as the vulnerability features of the street view image 220b. In other words, the vulnerability feature extractor 110 can focus on vulnerability features whose characteristics reflect damage as a result of a hazard event, such as a roof structure showing burn / smoke damage, burned trees, etc.

[0079] FIG. 2C is a schematic diagram of an exemplary pair of street view images of section B captured before and after a hazard event. The street view images 240a and 240b show the street view captured from street level and facing the section of the house 242 and the surrounding vegetation 224 located on section B shown outside the burn mark 206 and within the burn radius 208 of the hazard event. Different from section A described with reference to FIG. 2B, section B in FIG. 2C is shown as not having received damage from the hazard event.

[0080] Similarly, as described with reference to FIG. 2B, the vulnerability features extracted from the street view images 240a, 240b include a group of trees (F5), a group of thickets (F6), the frontage space (F7) between the house and the street, and the roof structure (F8) of the house 246. More or fewer vulnerability features can be extracted from the street view images 240a, 240b, and the examples provided are not limiting.

[0081] For each of the vulnerability features extracted from the Street View images 240a, 240b, the vulnerability feature extractor identifies the characteristics of each vulnerability feature. For example, a group of trees F5 can have associated quantifiable characteristics such as the distance of the trees to the house 242, the number of trees, the height of the trees, the proximity of the group of trees, etc., and quantifiable characteristics such as the soundness of the trees, being covered with ivy, etc. In another example, the frontage space F7 between the house and the street can have characteristics including, for example, distance, gradient, type of land cover (e.g., cement vs. grass), etc.

[0082] The vulnerability feature extractor 110 provides the imaging data 216, e.g., the vulnerability features and characteristics extracted from 200a, 200b, 220a, 220b, 240a, 240b, to the vector generator 114 to generate the training data 120.

[0083] Exemplary Process FIG. 3 is a flowchart of an exemplary process of the hazard vulnerability system 102. The system 102 receives a request for a damage propensity score for a section (302). The request 126 can be provided to the hazard vulnerability system 102 by a user via the graphical user interface of the application environment 130. The request 126 can include a request for a damage propensity score 140 for a specific section, and can further include a request for one or more mitigation steps 136 to reduce the damage propensity score 140. The request 126 can further specify a specific hazard event, e.g., an ongoing wildfire, or a general hazard event type, e.g., flood, and request a damage propensity score accordingly.

[0084] The system receives (304) imaging data of a section, including street view imaging data of the section. The hazard vulnerability system 102 can receive imaging data 116 including a street view image 117 from a repository of imaging data. Each image can include a specific designated section of interest to the user. In some embodiments, the user can capture additional street view images of the section with a camera, such as the built-in camera of the user device 128, and upload them to the hazard vulnerability system 102.

[0085] In some embodiments, the system receives imaging data of a section captured within a threshold time from the time of claim 126, for example, within 6 months, within 2 weeks, within 1 hour, etc.

[0086] A machine learning model including a classifier extracts (306) the characteristics of the vulnerability features of the section from the imaging data. The damage tendency model 106 can receive the imaging data 116 and extract the characteristics of the vulnerability features using a plurality of classifiers 107. The vulnerability feature extractor 110 receives an image of the section from the imaging data 116, extracts the vulnerability features F1, F2,... FN of the section, and for each vulnerability feature FN, the vulnerability feature extractor 110 extracts one or more characteristics C1, C2,... CN of the vulnerability feature FN. In one example, the vulnerability feature extractor 110 receives a street view image of a section that captures a house and surrounding assets, and uses a plurality of classifiers to identify vulnerability features including, for example, the house, vegetation, frontage area, fence, etc. The vulnerability feature extractor 110 identifies the characteristics of each of the extracted vulnerability features, for example, the building materials used for the house, such as siding type, roof type, eaves structure, etc.

[0087] The machine learning model determines a zoning trend score from the characteristics of the vulnerability features (308). The damage trend model 106 is trained on training data 120 that includes a plurality of hazard events, for example, hundreds of hazard events, and a plurality of zones for each hazard event, for example, thousands of zones, as will be described in more detail below with reference to FIG. 4, whereby the model 106 can infer between the characteristics of the vulnerability features of the zone and the damage trend of the zone. The model 106 generates a damage trend score 140 as an output.

[0088] The system provides a representation of the trend score for display (310). The damage trend score 140 can be provided within the graphical user interface of the application environment 130. The trend score 140 can be visually represented, for example, as a numerical value, and / or presented with contextual cues, for example, a relative scale of the hazard, color-coding (high, medium, or low risk). The trend score 140 can be presented with contextual information to help the user better understand the importance of the trend score 140.

[0089] In some embodiments, the characteristics of the vulnerability features of the zone and the imaging data can be utilized by the hazard vulnerability system 102 to generate a three-dimensional model of the zone. The three-dimensional model of the zone can be displayed within the graphical user interface to assist the user in understanding the damage trend score and / or one or more mitigation steps 136 for optimizing the damage trend score.

[0090] In some embodiments, the hazard vulnerability system 102 can determine one or more mitigation steps 136 to provide to a user as a way to reduce risk, e.g., optimize the damage propensity score 140. The mitigation engine 108 can receive the damage propensity score and characteristics of the vulnerability characteristics of the compartment and can determine one or more mitigation steps 136. The mitigation steps 136 can be identified by the damage propensity model 106 based on an inference of which characteristics of the vulnerability characteristics reduce the damage propensity. The mitigation steps 136 can include, for example, trimming tree branches away from the house. In another example, the mitigation steps 136 can include changing the roof structure material, changing the location of an outdoor shed, removing brush from the frontage area of the house, etc. The mitigation engine 108 can provide the proposed updated characteristics of the vulnerability characteristics to the damage propensity model 106. The damage propensity model 106 can receive the proposed updated characteristics of the vulnerability characteristics and can determine an updated damage propensity score 140.

[0091] In some embodiments, the mitigation engine 108 can determine the mitigation steps 136 by calculating the gradient of the damage propensity score 140 for each vulnerability characteristic vector. The vulnerability characteristic vectors having the threshold magnitude gradient and / or a subset of the gradients having the largest magnitude among the set of gradients can be utilized as a basis for the mitigation steps 136. In other words, the vulnerability characteristics having the greatest impact (larger magnitude gradient) on the damage / non-damage and / or damage severity results can be the focus of the mitigation steps as they can affect the damage propensity score 140 more than the vulnerability characteristics having a smaller impact (smaller magnitude gradient) on the results. For example, if the gradient of the damage propensity score 140 for the vulnerability characteristic vector regarding the roof structure material is at least the threshold magnitude or is ranked among the top in the gradients of the vulnerability characteristic vectors of the compartment, the roof structure material can be selected as a mitigation step.

[0092] In some embodiments, a neural network can be utilized to infer vulnerability features within the imaging data that can have the greatest impact on the damage tendency score 140 determined by the damage tendency model 106.

[0093] In some embodiments, the graphical user interface of the application environment 130 can present a visual representation of the mitigation step 136 on the user device 128. For example, the visual indicator 138 can identify the mitigation step and can indicate, for example, to remove vegetation from the plot. The visual indicator 138 can be a bounding box, a graphical arrow, or other indicator that surrounds the identified mitigation step 136. The visual indicator 138 can include text-based information regarding the mitigation step 136 that explains, for example, how and why the mitigation step 136 reduces the damage tendency score 140 of the plot.

[0094] In some embodiments, the system 102 can execute an optimization process by iterating on the proposed characteristics for the vulnerability features of the plot and calculating an updated tendency score 140 until an optimized, e.g., lowest, damage tendency score is found. The optimization process can continue until a threshold damage tendency score is met.

[0095] In some embodiments, the hazard vulnerability system 102 can further incorporate a cost analysis of the mitigation step 136. In other words, the hazard vulnerability system 102 can determine a mitigation step 136 that balances the optimization of the damage tendency score 140 while maintaining the cost of the mitigation step 136 below a threshold cost.

[0096] In some embodiments, the hazard vulnerability system 102 can determine one or more mitigation steps based on a specific type of hazard event, such as a flood versus a wildfire, and the mitigation steps for a first type of hazard event are different from the mitigation steps for a second type of hazard event. For example, the mitigation steps for a flood hazard can include digging outflow channels and updating the gutter system, while the mitigation steps for a wildfire can include mowing the vegetation surrounding the house.

[0097] In some embodiments, the hazard vulnerability system 102 receives a request 126 for a real-time damage propensity score 140 corresponding to a real-time hazard event. The hazard vulnerability system 102 receives real-time event data 142 about the hazard event and can determine a propensity score for a section corresponding to the hazard event from the characteristics of the vulnerability features and the real-time event data. The hazard vulnerability system 102 can re-evaluate the propensity score 140 in real time based on updated hazard event data about the hazard event, such as changes in weather conditions, containment, etc., and provide the user with an updated propensity score 140 according to the updated hazard event data.

[0098] In some embodiments, the mitigation engine 108 receives real-time event data 142. The real-time event data is used to provide the user with real-time mitigation steps 136 via the user device 128 and can reduce the section damage score 140 corresponding to the ongoing hazard event. For example, real-time event data including the spread and containment of wildfires, weather patterns, and emergency responder alerts can be used to help homeowners take immediate measures to address the spread of wildfires and reduce the tendency of their assets to be damaged by wildfires.

[0099] FIG. 4 is a flowchart of another exemplary process of a hazard vulnerability system. The hazard vulnerability system 102 generates training data for a machine learning model (402). Training a machine learning model, such as a damage propensity model 106, includes generating training data 120 that includes a large sample set of imaging data 116 and historical hazard event data 124 including compartment data 122, such as thousands of images for hundreds of hazard events. The generated training data represents various imaging conditions, such as weather conditions, lighting conditions, seasons, etc., for various hazard events, such as hazards of different scales, spread of hazards, location, type of hazard, and generalizing the damage propensity model 106 trained on the training data 120 to develop heuristics for a wide range of imaging conditions and hazard events.

[0100] The system receives, for a hazard event, a plurality of compartments located within the vicinity of the hazard event, each of the plurality of compartments having received at least a threshold exposure to the hazard event (404). The system can receive historical hazard event data 124 for the hazard event, and the historical hazard event data 124 can include compartment data 122 for each of the plurality of compartments located within the vicinity of the hazard event, such as within the combustion radius 208.

[0101] Proximity to a hazard event can be defined as being located within a combustion radius 208 surrounding the burn scar 206. As shown in FIG. 2A, the combustion radius 208 can define an extended area surrounding the burn scar 206. In some embodiments, proximity to a hazard event can be a threshold distance from the outer perimeter of the burn scar 206, such as within 1 mile, within 100 feet, within 5 miles, etc.

[0102] Threshold exposure is the minimum exposure amount to a hazard event by a compartment, and can be defined, for example, by the proximity of the compartment to the hazard event, the time the compartment is actively exposed to the hazard event, such as the time an asset is actively exposed to wildfire. In some embodiments, the threshold exposure can be defined using emergency responder metrics, such as high-risk or evacuation zones considered by emergency responders. In one example, a compartment can meet the threshold exposure by being considered to be within an evacuation zone for wildfire. In another example, a compartment can meet the threshold exposure by bringing flood water (or wildfire, or tornado, or earthquake, etc.) into contact with at least a portion of the compartment. In another example, a compartment can meet the threshold exposure by being located within or in the vicinity of a threshold of a burn scar.

[0103] In some embodiments, each of a plurality of compartments located within a burn scar can be considered to have received a threshold exposure amount.

[0104] In some embodiments, the fire radiative power (FRP) of a fire can be calculated from remotely sensed, mapped measurements of fire intensity. For example, using satellite data of active fires, any structure within a particular area and having a given threshold FRP value can be counted as having experienced the same threshold exposure amount.

[0105] For each of a plurality of sections located within the vicinity of a hazard event, the system receives (406) imaging data of the section including street view imaging data. The imaging data 116 can include, for example, satellite / aerial images 118 and street view images 117 collected by the hazard vulnerability system 102 from repositories of collected images located within various databases and sources. Each image of the imaging data 116 includes the capture time at which the image was captured and includes the geographical area including that section of the plurality of sections. The satellite / aerial image 118 includes the geographical area captured at a specific resolution and includes position information, such as GPS coordinates, that defines the geographical area captured within the frame of the image. The street view image 117 includes a street-level view of one or more of the plurality of sections captured at a specific resolution, such as one section, two sections, etc., and includes position information, such as the location address, that defines the position of the section captured in the street view image 117.

[0106] The system extracts (408) the characteristics of a plurality of vulnerability features of a first subset of sections that did not burn and a second subset of sections that burned during the hazard event from the imaging data. As described with reference to FIGS. 1, 2A-2C, the vulnerability feature extractor 110 can receive the imaging data 116 and use a plurality of classifiers to extract vulnerability features. In some embodiments, extracting the characteristics of the vulnerability features by the plurality of classifiers includes identifying objects within the imaging data 116 by the plurality of classifiers.

[0107] In some embodiments, the system can receive section data 122 of a first subset of sections that did not burn and a second subset of sections that burned during the hazard event. The section data 122 can include additional structural characteristics for each of the sections, such as post-hazard event inspections, building / construction records, appraisals, insurance appraisals, etc. The system can extract the vulnerability features and the characteristics of the vulnerability features of the first subset of non-burning sections and the second subset of burning sections from the additional structural data.

[0108] The system can generate training vectors from the extracted vulnerability features and the characteristics of the vulnerability features. In some embodiments, the vector generator module 114 generates training vectors from the extracted vulnerability features and the characteristics of the vulnerability features for a machine learning model.

[0109] System 102 can generate training data 120 for a specific hazard event using the extracted vulnerability features and the characteristics of the vulnerability features for each of a plurality of sections. Damage / non-damage records and / or damage degree records for a specific section among the plurality of sections from the historical hazard event data 124 can be utilized as ground truth for the burn / no-burn results of each section.

[0110] The system provides the training data to a machine learning model (410). The training data 120 is provided to a machine learning model, such as the damage tendency model 106, to train the damage tendency model 106 to infer the damage tendency of a specific section for a general hazard event type or a specific hazard event.

[0111] FIG. 5 is a block diagram of an exemplary computer system 500 that can be used to perform the operations described above. System 500 includes a processor 510, a memory 520, a storage device 530, and an input / output device 540. Each of the components 510, 520, 530, and 540 can be interconnected, for example, using a system bus 550. The processor 510 can process instructions for execution within the system 500. In one implementation, the processor 510 is a single-threaded processor. In another implementation, the processor 510 is a multi-threaded processor. The processor 510 can process instructions stored in the memory 520 or the storage device 530.

[0112] Memory 520 stores information within system 500. In one implementation, memory 520 is a computer-readable medium. In one implementation, memory 520 is a volatile memory unit. In another implementation, memory 520 is a non-volatile memory unit.

[0113] Storage device 530 can provide mass storage to system 500. In one implementation, storage device 530 is a computer-readable medium. In various different implementations, storage device 530 can include, for example, a hard disk device, an optical disk device, a storage device shared via a network by multiple computing devices (e.g., a cloud storage device), or some other large-capacity storage device.

[0114] Input / output device 540 provides input / output operations to system 500. In one implementation, input / output device 540 can include one or more network interface devices, such as an Ethernet card, serial communication devices, such as an RS-232 port, and / or wireless interface devices, such as a 502.11 card. In another implementation, the input / output device can include a driver device configured to receive input data and transmit output data to other input / output devices, such as a keyboard, a printer, and display device 560. However, other implementations, such as mobile computing devices, mobile communication devices, set-top box television client devices, etc., can also be used.

[0115] Although an exemplary processing system is depicted in FIG. 5, implementations of the subject matter and the functional operations described herein can be implemented in other types of digital electronic circuits, or in computer software, firmware, or hardware, or combinations of one or more of them, including the structures disclosed herein and their structural equivalents.

[0116] In this specification, the term "configured" is used in relation to systems and computer program components. A system of one or more computers being configured to perform a particular operation or action means that the system has installed in it software, firmware, hardware, or a combination of them that during operation cause the system to perform the operation or action. One or more computer programs being configured to perform a particular operation or action means that the one or more programs include instructions that, when executed by a data processing apparatus, cause the apparatus to perform the operation or action.

[0117] Embodiments of the subject matter and the functional operations described in this specification can be implemented in digital electronic circuitry, tangibly embodied computer software or firmware, computer hardware, or one or more combinations of them, including the structures disclosed in this specification and their structural equivalents. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., as one or more modules of computer program instructions encoded on a tangible non-transitory storage medium for execution by, or to control the operation of, a data processing apparatus. The computer storage medium can be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of them. Alternatively or additionally, the program instructions can be encoded on an artificially generated propagated signal, e.g., a machine-generated electrical, optical, or electromagnetic signal that is generated to encode information for transmission to a suitable receiver device for execution by a data processing apparatus.

[0118] The term "data processing apparatus" refers to data processing hardware and includes, by way of example, any kind of apparatus, device, and machine for processing data, including programmable processors, computers, or multiple processors or computers. The apparatus may also be, or further include, dedicated logic circuitry, such as, for example, an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit). Optionally, in addition to the hardware, the apparatus may include code for creating an execution environment for a computer program, such as, for example, processor firmware, a protocol stack, a database management system, an operating system, or code constituting one or more combinations thereof.

[0119] A computer program, which may also be referred to as or described as a program, software, software application, app, module, software module, script, or code, may be described in any form of programming language, including a compiler-type or interpreter-type language, or a declarative or procedural language, and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. The program may or may not correspond to a file in a file system. The program may be stored as part of a file that holds other programs or data, such as, for example, within a markup language document, within a single file dedicated to the program in question, or within one or more scripts stored within a plurality of cooperating files, such as, for example, one or more modules, subprograms, or files that hold portions of the code. The computer program may be executed on one computer, or located at one site, or distributed across multiple sites and executed on multiple computers interconnected by a data communication network.

[0120] As used herein, the term "engine" is widely used to refer to a software-based system, subsystem, or process that is programmed to perform one or more specific functions. Generally, an engine is implemented as one or more software modules or components and installed on one or more computers located at one or more locations. In some cases, one or more computers are dedicated to a particular engine. In other cases, multiple engines may be installed and operate on the same computer(s).

[0121] The processes and logical flows described herein can be performed by one or more programmable computers executing one or more computer programs so as to perform functions by operating on input data to generate output. The processes and logical flows can also be performed, for example, by dedicated logic circuitry, such as an FPGA or ASIC, or by a combination of application-specific logic circuitry and one or more programmed computers.

[0122] A computer suitable for the execution of a computer program can be based on a general-purpose or special-purpose microprocessor, or both, or any other type of central processing unit. Generally, the central processing unit receives instructions and data from a read-only memory, or a random access memory, or both. The essential elements of a computer are a central processing unit for executing or running instructions, and one or more memory devices for storing instructions and data. The central processing unit and the memory can be complemented by, or incorporated in, special-purpose logic circuitry. Generally, a computer also includes, or is operatively coupled to receive data from, or transfer data to, or both, one or more mass storage devices for storing data, such as magnetic disks, magneto-optical disks, or optical disks. However, a computer need not have such devices. Further, a computer can be incorporated in another device, such as, by way of example only, a cellular phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device, such as a universal serial bus (USB) flash drive.

[0123] Computer-readable media suitable for storing computer program instructions and data include, by way of example, all forms of non-volatile memory, media, and memory devices, including semiconductor memory devices, such as EPROM, EEPROM, and flash memory devices, magnetic disks, such as internal hard disks or removable disks, magneto-optical disks, and CD-ROM and DVD-ROM disks.

[0124] To provide interaction with a user, embodiments of the subject matter described herein can be implemented on a computer, which can have a display device for displaying information to the user, such as a CRT (cathode ray tube) or LCD (liquid crystal display) monitor, and a keyboard and a pointing device, such as a mouse or trackball, by which the user can provide input to the computer. Other types of devices can be used as well to provide interaction with the user. For example, the feedback provided to the user can be any form of sensory feedback, such as visual feedback, auditory feedback, or tactile feedback, and the input received from the user can be received in any form, including acoustic, voice, or tactile input. Additionally, the computer can interact with the user by sending documents to, and receiving documents from, the devices used by the user, such as by sending a web page to a web browser on the user's device in response to a request received from the web browser. Also, the computer can interact with the user by sending a text message or other form of message to a personal device, such as a smartphone running a messaging application, and receiving a response message from the user as a reply.

[0125] A data processing apparatus for implementing a machine learning model can include, for example, a dedicated hardware accelerator unit for processing general and computationally intensive parts of machine learning training or production operation, i.e., inference, workloads.

[0126] The machine learning model can be implemented and deployed using a machine learning framework such as the TensorFlow framework, the Microsoft Cognitive Toolkit framework, the Apache Singa framework, or the Apache MXNet framework.

[0127] Embodiments of the subject matter described herein can be implemented in a computing system that includes, for example, backend components as a data server, or includes middleware components, such as an application server, or includes frontend components, such as a graphical user interface, a web browser, or a client computer having an app with which a user can interact with an implementation of the subject matter described herein, or includes any combination of one or more such backend, middleware, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication, such as a communication network. Examples of communication networks include local area networks (“LANs”) and wide area networks (“WANs”), such as the Internet.

[0128] A computing system can include clients and servers. Clients and servers are generally remote from each other and typically interact over a communication network. The relationship of client and server arises by virtue of computer programs running on respective computers and having a client-server relationship to each other. In some embodiments, a server can send data, such as an HTML page, to a user device for the purpose of, for example, displaying data to a user interacting with a device operating as a client and receiving user input from the user. Data generated at the user device, such as the results of user interactions, can be received at the server from the device.

[0129] This specification includes the details of many individual implementations, which are not to be construed as limiting the scope of any features or the scope of what can be claimed, but rather as descriptions of features specific to particular embodiments. Certain features described herein in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, the various features described in the context of a single embodiment may be implemented separately in multiple embodiments or in any suitable partial combination. Further, features may be described above as acting in a particular combination and even initially claimed as such, but one or more features from the claimed combination may, in some cases, be deleted from the combination, and the claimed combination may be directed to a sub-combination or variation of a sub-combination.

[0130] Similarly, operations are depicted in the drawings in a particular order, but this should not be understood as requiring that such operations be performed in the particular order shown or in a sequential order to achieve the desired result, or that all of the illustrated operations be performed. In certain situations, multitasking and parallel processing may be advantageous. Further, the separation of various system components in the embodiments described above should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems may generally be integrated together in a single software product or packaged into multiple software products.

[0131] Accordingly, particular embodiments of the subject matter have been described. Other embodiments are within the scope of the following claims. In some examples, the acts recited in the claims can be executed in a different order and still achieve desirable results. Further, the processes shown in the accompanying figures do not necessarily require the particular order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing may be advantageous.

Claims

1. receiving a request for a damage propensity score for a parcel; receiving imaging data for the section, the imaging data including street view imaging data for the section; extracting a plurality of vulnerability signature characteristics of the compartment from the imaging data using a machine learning model including a plurality of classifiers; determining, by the machine learning model, the damage propensity score for the compartment from the characteristics of the plurality of vulnerability features; providing a representation of the injury propensity score for display; and A method comprising:

2. The method of claim 1 , further comprising generating a set of property partitions from the properties of the plurality of vulnerability features.

3. The method of claim 1 , further comprising generating a three-dimensional model of the compartment from the characteristics of the plurality of vulnerability features and imaging data of the compartment.

4. The method of claim 1 , wherein the imaging data for the partition comprises imaging data captured within a threshold time from a time of the request.

5. receiving the request for the injury propensity scores, receiving hazard event data about a hazard event; determining the damage propensity score of the subdivision for the hazard event from the characteristics of the plurality of vulnerability features and the hazard event data for the hazard event; The method of claim 1 , comprising:

6. receiving updated hazard event data for the hazard event; determining an updated damage propensity score for the partition for the hazard event from the characteristics of the plurality of vulnerability features, the hazard event data, and the updated hazard event data; The method of claim 5 further comprising:

7. determining one or more mitigation steps for the compartment via the machine learning model; determining, by the machine learning model, an updated injury propensity score based on the one or more mitigation steps; and providing a representation of the one or more mitigation steps and the updated injury propensity score; The method of claim 1 further comprising:

8. The method of claim 7 , wherein the one or more mitigation steps include adjustments to the characteristics of the plurality of vulnerability features extracted from the imaging data.

9. Determining one or more mitigation steps iterating an updated injury propensity score determination based on the adjusted characteristics of the plurality of vulnerability features; The method of claim 7 further comprising:

10. determining that the updated injury propensity score meets a threshold injury propensity score; The method of claim 9 further comprising:

11. determining the one or more mitigation steps determining the one or more mitigation steps for a particular type of hazard event, the one or more mitigation steps for a first type of hazard event being different from the one or more mitigation steps for a second type of hazard event; The method according to claim 7.

12. generating training data for the machine learning model, receiving, for a hazard event, a plurality of compartments located within a vicinity of the hazard event, each of the plurality of compartments having received at least a threshold exposure to the hazard event; receiving, for each partition of the plurality of partitions, imaging data for the partition including street view imaging data; extracting from the imaging data a plurality of vulnerability signature characteristics of a first subset of compartments of the plurality of compartments that did not burn during the hazard event and a second subset of compartments of the plurality of compartments that did burn; generating said training data, providing the training data to a machine learning model; The method of claim 1 further comprising:

13. The method of claim 12 , wherein extracting the characteristics of the plurality of vulnerability features comprises providing the imaging data to the plurality of classifiers.

14. The method of claim 13 , wherein extracting the plurality of vulnerability feature characteristics further comprises identifying a plurality of objects in the imaging data with the plurality of classifiers.

15. receiving additional structural characteristics for each section of the plurality of sections; extracting a second plurality of vulnerability signatures from the additional structural characteristics for the first subset of compartments of the plurality of compartments that did not burn during the hazard event and for the second subset of compartments of the plurality of compartments that did burn; providing the second plurality of vulnerability features to the machine learning model; The method of claim 12 further comprising:

16. The method of claim 15 , wherein the additional structural characteristic comprises a post hazard event inspection of the plurality of compartments.

17. A non-transitory computer storage medium encoded with a computer program, the computer program, when executed by a data processing apparatus, causing the data processing apparatus to: receiving a request for a damage propensity score for a parcel; receiving imaging data for the section, the imaging data including street view imaging data for the section; extracting a plurality of vulnerability signature characteristics of the compartment from the imaging data using a machine learning model including a plurality of classifiers; determining, by the machine learning model, the damage propensity score for the compartment from the characteristics of the plurality of vulnerability features; providing a representation of the injury propensity score for display; and A non-transitory computer storage medium containing instructions to cause a computer to perform operations including

18. generating training data for the machine learning model, receiving, for a hazard event, a plurality of compartments located within a vicinity of the hazard event, each of the plurality of compartments having received at least a threshold exposure to the hazard event; receiving, for each partition of the plurality of partitions, imaging data for the partition including street view imaging data; extracting from the imaging data a plurality of vulnerability signature characteristics of a first subset of compartments of the plurality of compartments that did not burn during the hazard event and a second subset of compartments of the plurality of compartments that did burn; generating said training data, providing the training data to a machine learning model; 20. The non-transitory computer storage medium of claim 17, further comprising:

19. receiving additional structural characteristics for each section of the plurality of sections; extracting a second plurality of vulnerability signatures for the first subset of compartments of the plurality of compartments that did not burn during the hazard event and for the second subset of compartments of the plurality of compartments that did burn from the additional structural characteristics; providing the training data to a machine learning model; 20. The non-transitory computer storage medium of claim 18, further comprising:

20. A user device; Interacting with the user device; receiving a request for an injury propensity score for a partition from the user device; receiving imaging data for the section, the imaging data including street view imaging data for the section; extracting a plurality of vulnerability signature characteristics of the compartment from the imaging data using a machine learning model including a plurality of classifiers; determining, by the machine learning model, the damage propensity score for the compartment from the characteristics of the plurality of vulnerability features; providing, to the user device, a representation of the injury propensity score for display; one or more computers operable to perform operations including: A system comprising:

Images