System and method for geospatially-contextualised behavioral anomaly detection for remote worker safety

Abstract

A computerised system for environment-based accident, incident, or risk detection for human subjects in outdoor environments comprises at least one user device configured to be carried or worn by a human subject. The at least one user device comprises a location determination module; a movement detection module comprising at least one of an accelerometer, gyroscope, or other motion sensor; and a communication module configured to transmit data via non- cellular-dependent protocols. The system further comprises a central server comprising at least one processor and memory. The central server is configured to receive, from the at least one user device, subject-specific data comprising: current location data indicating a current location of the human subject in an outdoor environment; and movement data relating to the human subject, including at least one of: speed, direction, acceleration, or stationary duration. The central server is configured to receive or access environment-specific data indicating environmental conditions for the outdoor environment. Said environment-specific data comprises terrain characteristic data indicating terrain characteristics, said terrain characteristic data comprising one or more of: geospatial data including topographical and geographical features; terrain classification data. Said environment-specific data further comprises meteorological data indicating localised weather conditions for the outdoor environment including at least one of: temperature, precipitation, wind, or visibility. The central server is configured to generate a digital geospatial representation of the outdoor environment, comprising: a plurality of geospatial polygons defining one or more active monitoring zones and one or more non-active monitoring zones; wherein active monitoring zones designate outdoor areas where accident, incident or risk detection monitoring is enabled; and wherein non-active monitoring zones designate areas where accident, incident or risk detection monitoring is suspended or not enabled. The central server is configured to determine, for the current location of the human subject, whether the location falls within an active monitoring zone or a non-active monitoring zone by performing geospatial polygon intersection analysis. If the location falls within an active monitoring zone, the central server is configured to perform behavioral anomaly analysis comprising: retrieving historical movement data for the human subject from a historical movement database, said historical movement data comprising prior movement data of the human subject under the same or similar terrain characteristics and the same or similar weather conditions as those indicated by the environment-specific data; calculating expected movement data for the human subject at the current location based on: (i) the historical movement data, (ii) the terrain characteristics indicated by the environment- specific data, and (iii) the weather conditions indicated by the environment-specific data; and comparing the movement data of the human subject against the expected movement data to generate an anomaly score. If the anomaly score exceeds a predetermined threshold, the central server is configured to generate an output alert signal to at least one recipient device.

Description

[0001] System and Method for Geospatially-Contextualised Behavioral Anomaly Detection for Remote Worker Safety FIELD OF THE INVENTION

[0002] The present invention relates to a system for environment-based accident, incident or risk detection. The invention has particular application for monitoring human subjects located in remote, outdoor or isolated environments and / or environments with potential hazards such as difficult topography or hazardous machinery or other assets, such as agricultural workers, miners, or hikers. More particularly, the invention relates to an automated system that dynamically assesses accident, incident, or risk conditions by analysing behavioral patterns, environmental (weather) context, and geospatial data (terrain characteristics) without primary reliance on physiological biometric sensors. However, the invention may also have various other applications.

[0003] BACKGROUND

[0004] Workers in agriculture, construction, emergency response, military operations, and other outdoor industries face significant safety risks when working in remote locations where direct human oversight is impractical or impossible.

[0005] Existing accident detection technologies may tend to require specific hardware and / or software to operate, and / or may tend to require the subject to be in range of cell towers, meaning that such solutions may not be suitable if the subject does not have the required software / hardware and / or is in a remote area that is out of range (or where network infrastructure is absent or unreliable).

[0006] Furthermore, existing accident detection technologies may only be based on one parameter, or a limited range of parameters, that do not fully reflect the realities and complexities of the subject’s environment. For instance, such technologies may be based only on detected movement of the subject, or only on a single parameter type, or only on generic, preloaded thresholds that may be inflexible, inaccurate, and incapable of accurately mapping onto complex and dynamic outdoor environments. A further limitation of existing accident detection technologies is that they may require manual triggering of an alert or alarm by the subject, or other input by the subject.

[0007] These problems may result in failure to generate an alert or alarm when required. They may also result in the generation of false alerts or alarms based on problems with transmitting or accurately analysing data.

[0008] It is accordingly an object of the present invention to provide a system that ameliorates one or more of the above-noted problems with the prior art. At the very least, it is an object of the invention to provide the public with a useful choice.

[0009] STATEMENTS OF THE INVENTION

[0010] According to one aspect of the invention, there is provided a computerised system for environment-based accident, incident or risk detection, the system comprising a server having a processor and a memory, the system adapted to:

[0011] receive inputs comprising:

[0012] subject-specific data from at least one device associated with a human subject in the environment, comprising at least one subject-specific data item relating to movement and / or physiology of the human subject;

[0013] environment-specific data comprising at least one environment-specific data item relating to the environment;

[0014] based on the received inputs, compare the at least one subject-specific data item against a corresponding expected subject-specific data item for the environment,

[0015] the corresponding expected subject-specific data item for the environment being determined based on one or more features and / or characteristics of the environment, wherein the corresponding expected subject-specific data item for a first location in the environment is different from the corresponding expected subject-specific data item for a second location in the environment; and if the at least one subject-specific data item is outside of a predetermined range relative to the corresponding expected subject-specific data item for the environment, generate an output signal. Throughout the present specification, reference to a “subject” will be understood as being a reference to a human subject, unless the context otherwise requires.

[0016] In some examples, the at least one subject-specific data item comprises data relating to one or more of: identity ofthe subject; bodily movement of the subject; speed or velocity of the subject within the environment; location of the subject within the environment; physiological parameters of the subject such as heart rate, respiratory rate, blood pressure, body temperature. In some examples, the subject-specific data is obtained and transmitted by a single device associated with the subject; for instance, a single device may be adapted to obtain and transmit different types of subject-specific data items. In other examples, there are multiple devices associated with the subject for obtaining and transmitting different types of subject-specific data items.

[0017] In some examples, the at least one device associated with the subject comprises one or more of: a GPS device; a GNSS device; a 3-axis accelerometer; a device adapted to effect signal triangulation using LoRaWan or Wi-Fi

[0018] In some examples, the at least one device associated with the subject comprises a physiological sensor, such as a heartrate sensor, for example an optical (PPG) sensor.

[0019] In some examples, the at least one device associated with the subject comprises a mobile device of the subject, which may be programmed with a mobile app to transmit data to the system and / or to perform at least some of the steps for which the system is adapted. The mobile device of the subject may itself comprise one or more sensors (such as GPS sensors or physiological sensors), and / or it may be adapted to receive data from other sensors (being other devices associated with the subject), which it may then transmit to the system and / or based on which it may perform at least some of the steps for which the system is adapted.

[0020] In some examples, the at least one device associated with the subject is communicable with the system without reliance on cell towers.

[0021] In some examples, the at least one device associated with the subject is communicable with the system using Internet of Things (loT) technology. In some examples, the at least one device associated with the subject is communicable with the system using a satellite connection(s), such as low Earth orbit and / or geostationary satellite connect! on(s), via a suitable communication protocol(s).

[0022] In some examples, the at least one device associated with the subject is communicable with the system using a ground based network(s) such as LoraWan, Bluetooth, and / or Wi-Fi.

[0023] In some examples, the at least one device associated with the subject is adapted to transmit the at least one subject-specific data item to the system periodically, such as at predetermined intervals or upon occurrence of a trigger event.

[0024] In some examples, the system is adapted to receive, as an input, further subject-specific data, comprising at least one further subject-specific data item relating to one or more of: at least one planned activity of the subject in the environment; past subject-specific data of the subject; health data and / or mental health data of the subject.

[0025] In some examples, said further subject-specific data is inputted via the at least one device associated with the subject. In other examples, said further subject-specific data is obtained by the system from another device, or from data stored in the system or stored externally such as in an external database.

[0026] In some examples, said further subject-specific data relating to at least one planned activity of the subject in the environment includes data relating to one or more of: the location of the at least one planned activity in the environment; the time of the planned activity; machinery or assets involved in the planned activity.

[0027] In some examples, said past subject-specific data of the subject includes data as to the environment corresponding to said past subject-specific data.

[0028] In some examples, the at least one environment-specific data item comprises data relating to one or more of: topographical or geographical features of the environment; property boundaries; non-natural features of the environment; assets or machinery in the environment; light conditions in the environment (which may also be considered a subset of weather conditions); temperature conditions in the environment (which may also be considered a subset of weather conditions); current and / or predicted weather conditions in the environment such as precipitation and / or wind speed and / or direction. In some examples, at least some of the environment-specific data may be preprogrammed into the system prior to the subject being in the environment. For instance, data as to topographical or geographical features of the environment, property boundaries and / or non-natural features of the environment such as buildings, may be preprogrammed into the system, such as based on publicly available databases (such as public title data to determine property boundaries) and / or based on a landowner and / or occupant (and / or the subject) scanning or otherwise inputting the relevant environment-specific data into the system using a suitably adapted device such as a mobile device. A combination of multiple sources of environment-specific data may be used, for example a publicly available database for some geographical or topographical data and / or property boundaries combined with images, videos or scans from the landowner and / or occupant (and / or the subject) to pinpoint aspects like non-natural features and assets such as machinery.

[0029] In some examples, the system is adapted to use at least some of the environment-specific data to generate and store a digital representation of the environment. For instance, a digital representation of the topographical or geographical features of the environment, and / or non-natural features of the environment such as buildings, may be generated using polygon geospatial data techniques, and stored in the system. Subsequently, data such as the subject’s location may be overlaid by the system onto the digital representation of the environment. In some examples, at least some of the environment-specific data may be obtained by the system via dedicated devices. For instance, mobile assets such as vehicles or machinery may be associated with GPS devices or RFID tags adapted to periodically update the system as to their location in the environment. In another example, the dedicated devices may relate to determining weather conditions, and may include one or more of: a satellite weather radar system(s); a temperature sensor(s); a light sensor(s); a sensor(s) for determining precipitation, wind and / or lightning.

[0030] The system may be adapted for use with 35.5 - 36GHz Ka Band and / or High (>400 MHz) and Low bandwidth. The dedicated devices may be communicable with the system without reliance on cell towers, such as: using Internet of Things (loT) technology; using satellite connection(s), such as low Earth orbit and / or geostationary satellite connection(s), via a suitable communication protocol(s); and / or using a ground based network(s) such as LoraWan, Bluetooth, and / or Wi-Fi. In some examples, at least some of the environment-specific data may be inputted by the subject in use, using the at least one device associated with the subject or another device of the subject, such as a mobile device.

[0031] In some examples, at least some of the environment-specific data may be obtained or retrieved by the system upon occurrence of a trigger; for instance, data as to weather conditions may be obtained by the system (such as from satellite data) upon occurrence of a trigger. The trigger may be, for example, a preliminary indication determined by the system that the subject may be involved in an accident, incident or risk condition.

[0032] In some examples, said comparing of the at least one subject-specific data item against a corresponding expected subject-specific data item for the environment is performed by the system using at least one algorithm, and / or using artificial intelligence (Al).

[0033] In some examples, said corresponding expected subject-specific data item for the environment is preprogrammed and stored in the system or in a manner accessible by the system. In other examples, said corresponding expected subject-specific data item for the environment is dynamically generated by the system on an as-needed basis. In some examples, said corresponding expected subject-specific data item for the environment is derived from data pertaining to previous subjects in a similar environment.

[0034] In some examples, said comparing of the at least one subject-specific data item against a corresponding expected subject-specific data item for the environment comprises retrieving previously-transmitted or recorded instances of the subject’s at least one subject-specific data item over a predetermined period of time, and comparing said previously-transmitted or recorded instances against the corresponding expected subject-specific data item.

[0035] In some examples, obtaining or determining the corresponding expected subject-specific data item for the environment comprises identifying the subject’s location within the environment, and identifying the features and / or characteristics of the environment at that location (from the environment-specific data and / or from the digital representation of the environment), in order to determine the corresponding expected subject-specific data item for the environment. It will be understood that a corresponding expected subject-specific data item for a first location in the environment will be different than a corresponding expected subject-specific data item for a second location in the environment. The first and second locations in the environment may have different features and / or characteristics, as reflected in the respective environmentspecific data for those locations. For instance, if the first location has more difficult terrain than the second location, then the corresponding expected subject-specific data item, such as expected pace across the ground, at the first location may be slower than at the second location. In some examples, said comparing of the at least one subject-specific data item against a corresponding expected subject-specific data item for the environment further comprises using the further subject-specific data, such as to modify the corresponding expected subject-specific data item. For instance, the further subject-specific data may comprise past subject-specific data of the subject, and may indicate that the subject generally moves more slowly than an average subject, or that the subject has a generally higher blood pressure than an average subject. Accordingly, the corresponding expected subject-specific data item for the relevant parameter may be modified to take this into account. In another example, the further subjectspecific data may comprise a planned activity of the subject at that location and / or that time. Accordingly, the corresponding expected subject-specific data item may be modified to take this into account - for instance the expected speed of the subject may be increased if the planned activity at that location and / or that time involves use of a vehicle (as opposed to walking), or the expected speed may be decreased if the planned activity is a rest stop or lunch break involving a stationary period.

[0036] Alternatively, instead of using the further subject-specific data to modify the corresponding expected subject-specific data item, the further subject-specific data may be used to modify the size of the predetermined range between the at least one subject-specific data item and the corresponding expected subject-specific data item. For instance, if it is known that the subject generally moves more slowly than the average subject, the system may allow a greater range (or deviation) as between the subject’s speed and the expected speed in the environment. In some examples, at least some subject-specific data items may automatically be classed as falling outside of the predetermined range. For instance, absence of (or irregular) heart rate may automatically be classed as falling outside of the predetermined range, and thus indicate generation of an output signal.

[0037] As an example of interpreting received subject-specific data: If the heart rate was to drop 20% below the lower observed rolling average, at a similar time of day and day of week, a warning would be triggered. A 40% deviation below the lower or 40% higher observed rolling average, at a similar time of day and day of week would trigger an alarm with immediate sharing of the individual’s location to pre registered contacts in the system.

[0038] Outside of biometrics, the system may in some embodiments also look at typical behaviours and movement patterns for specific locations in the designated active areas (also referred to herein as “the environment” in which the subject is located, and which, depending on the input into the system, may vary in size and may include private properties, public properties such as national parks, or larger geographical areas such as states, territories or even countries). Just like biometrics the system would look at the deviation from the norm, but in this case would look at clustering algorithms to handle the behavioural observations, these sometimes cluster around time of year, or micro climate conditions as one example.

[0039] In some examples, the system is adapted to perform a preliminary step of comparing the at least one subject-specific data item against a corresponding expected subject-specific data item for the environment, and determining whether the at least one subject-specific data item is outside of a predetermined range relative to the corresponding expected subject-specific data item for the environment; and subsequently, if the at least one subject-specific data item is outside of the predetermined range, perform a further step of assessing or parsing at least some of the environment-specific data (optionally including by obtaining further environmentspecific data) to detect an environmental risk condition; and if an environmental risk condition is detected, generate the output signal.

[0040] In some examples, the environmental risk condition comprises one or more of: adverse current or predicted weather conditions; topographical or geographical features associated with risk; or assets or machinery associated with risk.

[0041] In some examples, the output signal generated by the system comprises an alert or alarm. In some examples, the alert or alarm is generated automatically by the system.

[0042] In some examples, the alert or alarm is sent to a recipient device, being a communication device, such as a mobile device or computer, of at least one third party, such as one or more of: emergency services; an employer or colleague of the subject; a family member, friend, or other designated third party. In some examples, the alert or alarm comprises one or more of: an SMS or text message; an email.

[0043] In some examples, the alert or alarm comprises information including one or more of: the identity of the subject; the location of the subject; an indication or type of accident, incident or risk; an environmental risk condition; at least some of the subject-specific data; at least some of the environment-specific data; a risk factor calculated by the system; a suggested route to reach the location of the subject.

[0044] In some examples, the location of the subject may be expressed as a latitude and a longitude, and / or may be displayed on a map of the environment.

[0045] In some examples, the output signal comprises a message to the subject sent to a communication device, such as a mobile device, of the subject (aka a recipient device of the subject).

[0046] In some examples, the message to the subject indicates that an alert or alarm will be sent in the absence of a response from the subject.

[0047] In some examples, the subject can deactivate the proposed alert or alarm by responding to the message within a predetermined period of time; or if no response is received within the predetermined period of time, the alert or alarm is sent to the at least one third party.

[0048] In some examples, the subject can pre-emptively suspend the generation of output signals by the system. For instance, if the subject intends to pause or take a break, which may otherwise cause generation of an output signal due to a change in the subject’s pace, the subject may input a command into the system (for instance via the at least one device, or via another communication device of the subject) to prevent generation of an output signal.

[0049] In some examples, the subject can terminate the alert or alarm after the alert or alarm has been sent.

[0050] In some examples, the at least one third party can terminate the alert or alarm.

[0051] In some examples, the system is adapted to utilise at least one algorithm, and / or artificial intelligence (Al). In some examples, the at least one algorithm, and / or the artificial intelligence (Al), is adapted to perform for one or more of the following aspects, protocols or steps:

[0052] Identification of, and / or determining one or more characteristics or parameters of, one or more of the at least one environment-specific data item, such as:

[0053] identifying and / or determining characteristics / parameters of assets or machinery; identifying and / or determining characteristics / parameters of nonnatural features of the environment; identifying and / or determining characteristics / parameters of topographical or geographical features of the environment;

[0054] modelling current weather conditions in the environment;

[0055] predicting future weather conditions in the environment;

[0056] Identification of, and / or determining one or more characteristics or parameters of, one or more of the at least one subject-specific data item, such as to assess mental and / or physical health of the subject;

[0057] Identification of, and / or determining one or more characteristics or parameters of, the further subject-specific data item;

[0058] Determining the corresponding expected subject-specific data item for the environment;

[0059] Using the further subject-specific data item to modify the corresponding expected subject-specific data item and / or to modify the size of the predetermined range between the at least one subject-specific data item and the corresponding expected subject-specific data item;

[0060] Comparing the at least one subject-specific data item against the corresponding expected subject-specific data item for the environment;

[0061] Determining whether the at least one subject-specific data item is outside of the predetermined range relative to the corresponding expected subject-specific data item; Obtaining, assessing and / or parsing at least some of the environment-specific data and / or further environment-specific data to detect an environmental risk condition;

[0062] Generate the output signal.

[0063] According to another aspect of the invention, there is provided a method for environment-based accident, incident or risk detection, the method being for execution by a computerised system comprising a server having a processor and a memory, the method comprising:

[0064] receiving inputs comprising:

[0065] subject-specific data from at least one device associated with a human subject in the environment, comprising at least one subject-specific data item relating to movement and / or physiology of the human subject;

[0066] environment-specific data comprising at least one environment-specific data item relating to the environment;

[0067] based on the received inputs, comparing the at least one subject-specific data item against a corresponding expected subject-specific data item for the environment, the corresponding expected subject-specific data item for the environment being determined based on one or more features and / or characteristics of the environment, wherein the corresponding expected subject-specific data item for a first location in the environment is different from the corresponding expected subject-specific data item for a second location in the environment; and

[0068] if the at least one subject-specific data item is outside of a predetermined range relative to the corresponding expected subject-specific data item for the environment, generating an output signal.

[0069] The invention may otherwise be substantially as described above.

[0070] According to another aspect of the invention, there is provided a computerised system for environment-based accident, incident or risk detection, the system comprising a server having a processor and a memory, the system comprising:

[0071] at least one device associated with a human subject in the environment, the at least one device being communicable with the system without reliance on cell towers, the at least one device being adapted to obtain and transmit subject-specific data comprising at least one subject-specific data item relating to movement and / or physiology of the human subject;

[0072] the system adapted to also receive environment-specific data comprising at least one environment-specific data item relating to the environment;

[0073] the system being adapted to:

[0074] based on the received inputs, compare the at least one subject-specific data item against a corresponding expected subject-specific data item for the environment,

[0075] the corresponding expected subject-specific data item for the environment being determined based on one or more features and / or characteristics of the environment, wherein the corresponding expected subject-specific data item for a first location in the environment is different from the corresponding expected subject-specific data item for a second location in the environment; and if the at least one subject-specific data item is outside of a predetermined range relative to the corresponding expected subject-specific data item for the environment, generate an output signal for sending to a recipient device.

[0076] The invention may otherwise be substantially as described above.

[0077] According to another aspect of the invention, there is provided a computerised system for environment-based accident, incident or risk detection, the system comprising a server having a processor and a memory, the system adapted to:

[0078] receive inputs comprising:

[0079] subject-specific data from at least one device associated with a human subject in the environment, comprising the location of the subject in the environment and at least one subject-specific data item relating to movement and / or physiology of the human subject;

[0080] environment-specific data comprising at least two different types of environment-specific data item relating to the environment; based on the at least two different types of environment-specific data item, generate a digital representation of the environment;

[0081] for the received location of the subject in the environment:

[0082] based on the digital representation of the environment and / or based on the received environment-specific data, identify one or more features and / or characteristics of the environment at the location of the subject; based on the identified one or more features and / or characteristics of the environment at the location of the subject, determine a corresponding expected subject-specific data item for the environment;

[0083] for the received at least one subject-specific data item:

[0084] compare the at least one subject-specific data item against the corresponding expected subject-specific data item for the environment; and

[0085] if the at least one subject-specific data item is outside of a predetermined range relative to the corresponding expected subject-specific data item for the environment, generate an output signal.

[0086] The invention may otherwise be substantially as described above.

[0087] According to another aspect of the invention, there is provided a computerised system for environment-based accident, incident, or risk detection for human subjects in outdoor environments, the system comprising:

[0088] (a) at least one user device configured to be carried or worn by a human subject, the at least one user device comprising:

[0089] a location determination module;

[0090] a movement detection module comprising at least one of an accelerometer, gyroscope, or other motion sensor; and

[0091] a communication module configured to transmit data via non-cellular- dependent protocols; and (b) a central server comprising at least one processor and memory, configured to: receive, from the at least one user device, subject-specific data comprising: current location data indicating a current location of the human subject in an outdoor environment; and

[0092] movement data relating to the human subject, including at least one of: speed, direction, acceleration, or stationary duration;

[0093] receive or access environment-specific data indicating environmental conditions for the outdoor environment, said environment-specific data comprising:

[0094] terrain characteristic data indicating terrain characteristics, said terrain characteristic data comprising one or more of:

[0095] geospatial data including topographical and geographical features;

[0096] terrain classification data;

[0097] meteorological data indicating localised weather conditions for the outdoor environment including at least one of: temperature, precipitation, wind, or visibility; and

[0098] generate a digital geospatial representation of the outdoor environment, comprising:

[0099] a plurality of geospatial polygons defining one or more active monitoring zones and one or more non-active monitoring zones; wherein active monitoring zones designate outdoor areas where accident, incident or risk detection monitoring is enabled; and wherein non-active monitoring zones designate areas where accident, incident or risk detection monitoring is suspended or not enabled; determine, for the current location of the human subject, whether the location falls within an active monitoring zone or a non-active monitoring zone by performing geospatial polygon intersection analysis;

[0100] if the location falls within an active monitoring zone, perform behavioral anomaly analysis comprising:

[0101] retrieving historical movement data for the human subject from a historical movement database, said historical movement data comprising prior movement data of the human subject under the same or similar terrain characteristics and the same or similar weather conditions as those indicated by the environment-specific data calculating expected movement data for the human subject at the current location based on: (i) the historical movement data, (ii) the terrain characteristics indicated by the environment-specific data, and (iii) the weather conditions indicated by the environment-specific data; comparing the movement data of the human subject against the expected movement data to generate an anomaly score;

[0102] if the anomaly score exceeds a predetermined threshold, generate an output alert signal to at least one recipient device.

[0103] According to another aspect of the invention, there is provided a computerised system for environment-based accident, incident, or risk detection for human subjects in outdoor environments, the system being configured to be communicable with at least one user device configured to be carried or worn by a human subject, the at least one user device comprising:

[0104] a location determination module;

[0105] a movement detection module comprising at least one of an accelerometer, gyroscope, or other motion sensor; and

[0106] a communication module configured to transmit data via non-cellular-dependent protocols; and the system comprising a central server comprising a at least one processor and memory, configured to:

[0107] receive, from the at least one user device, subject-specific data comprising:

[0108] current location data indicating a current location of the human subject in an outdoor environment; and

[0109] movement data relating to the human subject, including at least one of: speed, direction, acceleration, or stationary duration;

[0110] receive or access environment-specific data indicating environmental conditions for the outdoor environment, said environment-specific data comprising:

[0111] terrain characteristic data indicating terrain characteristics, said terrain characteristic data comprising one or more of:

[0112] geospatial data including topographical and geographical features; terrain classification data;

[0113] meteorological data indicating localised weather conditions for the outdoor environment including at least one of: temperature, precipitation, wind, or visibility; and

[0114] generate a digital geospatial representation of the outdoor environment, comprising:

[0115] a plurality of geospatial polygons defining one or more active monitoring zones and one or more non-active monitoring zones;

[0116] wherein active monitoring zones designate outdoor areas where accident, incident or risk detection monitoring is enabled; and

[0117] wherein non-active monitoring zones designate areas where accident, incident or risk detection monitoring is suspended or not enabled;

[0118] determine, for the current location of the human subject, whether the location falls within an active monitoring zone or a non-active monitoring zone by performing geospatial polygon intersection analysis; if the location falls within an active monitoring zone, perform behavioral anomaly analysis comprising:

[0119] retrieving historical movement data for the human subject from a historical movement database, said historical movement data comprising prior movement data of the human subject under the same or similar terrain characteristics and the same or similar weather conditions as those indicated by the environmentspecific data

[0120] calculating expected movement data for the human subject at the current location based on: (i) the historical movement data, (ii) the terrain characteristics indicated by the environment-specific data, and (iii) the weather conditions indicated by the environment-specific data;

[0121] comparing the movement data of the human subject against the expected movement data to generate an anomaly score;

[0122] if the anomaly score exceeds a predetermined threshold, generate an output alert signal to at least one recipient device.

[0123] The subject-specific data may comprise movement data and current location data, without requiring continuous physiological biometric data for the anomaly analysis, thereby reducing power consumption and hardware cost of the system compared to biometric-dependent systems.

[0124] The subject-specific data may further comprise data relating to physiological parameters of the human subject selected from at least one of heart rate, respiratory rate, blood pressure, or body temperature, wherein said physiological parameters are used as supplementary context data and not as primary data for the anomaly analysis.

[0125] The at least one user device may comprise one or more of: a GPS device; a GNSS device; a 3-axis accelerometer; a device adapted to effect signal triangulation using LoRaWAN or Wi-Fi; or a physiological sensor comprising an optical photoplethysmography (PPG) sensor.

[0126] The at least user device may comprise a mobile device programmed with a mobile application configured to transmit data to the central server. The at least one user device may be communicable with the system without reliance on cell towers, wherein the at least one user device may be communicable with the system using one or more of: Internet of Things (loT) technology; a satellite connection via low Earth orbit or geostationary satellite, via a suitable communication protocol; or a ground-based network comprising LoRaWAN, Bluetooth, or Wi-Fi.

[0127] The central server may be configured to receive, as an input, further subject-specific data relating to one or more of: at least one planned activity of the human subj ect in the environment; past subject-specific data of the human subject; or health data and / or mental health data of the human subject.

[0128] The further subject-specific data relating to at least one planned activity of the human subject in the environment may include data relating to one or more of: the location of the at least one planned activity in the environment; the time of the planned activity; or machinery or assets involved in the planned activity.

[0129] The past subject-specific data of the human subject may include data as to the environmental conditions corresponding to said past subject-specific data, enabling correlation of historical movement data with historical environmental conditions.

[0130] The environment-specific data may further comprise data relating to one or more of: property boundaries; non-natural features of the environment; assets or machinery in the environment; light conditions in the environment; temperature conditions in the environment; or current and / or predicted weather conditions in the environment selected from precipitation, wind speed, or wind direction.

[0131] At least some of the environment-specific data may be pre-programmed into the system prior to the subject being in the environment.

[0132] At least some of the environment-specific data may be obtained from multiple sources of environment-specific data, comprising: a publicly available database or source; and data from a party associated with the environment and / or data from the human subject; wherein at least some geographical and / or topographical data, and / or data as to property boundaries, may be obtained from the publicly available database, and wherein at least some data as to non-natural features and / or assets in the environment may be obtained from the party associated with the environment and / or from the human subject. The central server may be adapted to use at least some of the environment-specific data to generate and store the digital geospatial representation of the environment by: applying a trained convolutional neural network image classification model to satellite imagery or aerial imagery of the outdoor areas of the environment; said trained convolutional neural network image classification model outputting pixel-level terrain classifications which classify each of a plurality of pixels into a terrain classification; converting the classified pixels into vector geospatial polygons representing contiguous regions of the same terrain classification; and storing said geospatial polygons in a spatial database indexed by geographic coordinates for rapid intersection queries.

[0133] The digital geospatial representation may comprise one or more of: topographical and / or geographical features of the environment; non-natural features of the environment; or assets in the environment.

[0134] The digital geospatial representation may be generated using polygon geospatial data techniques comprising at least one of: GeoJSON format, Well-Known Text (WKT) format, or shapefile format, and wherein the geospatial polygon intersection analysis may be performed using spatial indexing selected from R-tree or Quad-tree indexing.

[0135] In use, the current location data may be overlaid by the central server onto the digital geospatial representation of the environment, and wherein the central server may be configured to enable or provide a visual map interface, said visual map interface displaying: the human subject's current location and recent movement trail derived from the movement data; boundaries of the active monitoring zones and non-active monitoring zones; environmental hazard indicators selected from extreme weather zones, flooded areas, or restricted zones; and locations of assets, infrastructure, and other workers in proximity of the human subject.

[0136] At least some of the environment-specific data may be obtained by the central server via one or more dedicated devices, comprising one or more of: GPS devices or RFID tags associated with mobile assets in the environment, adapted to periodically update the system as to their location in the environment; a satellite weather radar system; a temperature sensor; a light sensor; or a sensor for determining precipitation, wind, or lightning.

[0137] The one or more dedicated devices may be communicable with the central server without reliance on cell towers, wherein the one or more dedicated devices may be communicable with the system using one or more of: Internet of Things (loT) technology; a satellite connection comprising low Earth orbit and / or geostationary satellite connection; or a ground-based network comprising LoRaWAN, Bluetooth, or Wi-Fi.

[0138] At least some of the environment-specific data may be obtained, retrieved and / or assessed by the central server upon occurrence of a trigger; wherein the trigger may comprise a preliminary indication of an accident, incident or risk condition determined by the system based on the subject-specific data.

[0139] If the central server determines a preliminary indication of an accident, incident or risk condition, the central server may be configured to obtain, retrieve and / or assess said at least some of the environment-specific data to detect an environmental risk condition; and if an environmental risk condition is detected, generate the output signal.

[0140] The environmental risk condition may comprise one or more of: adverse current and / or predicted weather conditions; topographical or geographical features associated with risk; or assets or machinery associated with risk.

[0141] The expected movement data for the human subject may be dynamically generated by the central server in real-time as current location data is received, said dynamic generation comprising: querying the historical movement database for prior movement data of the human subject under the same or similar terrain characteristics and the same or similar weather conditions as those indicated by the environment-specific data; applying a machine learning regression model to obtain predicted values, being expected speed, direction, and dwell time, of the human subject in view of the environment-specific data; adjusting the predicted values based on environmental modulation factors selected from temperature adjustment, precipitation adjustment, terrain ruggedness adjustment, time-of-day adjustment, or asset proximity adjustment; and wherein the machine learning regression model may be continuously retrained with new movement data to adapt to changing behavioral patterns of the human subject over time.

[0142] Where there is insufficient historical movement data for the human subject, the expected movement data may be derived from population-based historical movement data from a plurality of other human subjects under the same or similar terrain characteristics and the same or similar weather conditions as those indicated by the environment-specific data,; wherein, when sufficient historical movement data for the human subject becomes available, the system may transition from using the population-based historical movement data to using the historical movement data for the human subject.

[0143] The historical movement data may comprise prior movement data of the human subject obtained over a predetermined period of time.

[0144] Obtaining or determining the expected movement data may comprise: identifying the current location of the human subject within the environment using GPS coordinates; performing a spatial query to identify geospatial polygon(s) containing the current location; retrieving data associated with the identified polygon(s), said data comprising one or more terrain characteristics including: terrain type, terrain ruggedness score, one or more typical activities performed in the identified polygon(s), or proximity to assets or hazards; retrieving current weather conditions at the current location from a high-resolution meteorological data source; querying the historical movement database for prior movement data of the human subject in the same polygon or adjacent polygons under weather conditions within a similarity threshold of the current weather conditions; computing statistical metrics from the historical movement data selected from mean speed, standard deviation of speed, or typical dwell time distribution; applying environmental modulation factors to adjust statistical metrics based on the current weather conditions; and defining the expected movement data as acceptable ranges around the adjusted statistical metrics.

[0145] The one or more terrain characteristics may be identified based on the digital geospatial representation of the environment.

[0146] The comparing of the current movement data against the expected movement data may comprise using the further subject-specific data to either: modify the expected movement data; or modify the calculation of the anomaly score and / or the predetermined threshold associated with the anomaly score.

[0147] The output alert signal generated by the central server may comprise an alert or alarm sent to the recipient device of at least one third party, being one or more of: emergency services; an employer or colleague of the human subject; a family member or friend of the human subject; or another designated third party.

[0148] The alert or alarm may comprise information including one or more of: the identity of the human subject; the current location of the subject; an indication or type of accident, incident or risk; an environmental risk condition; at least some of the subject-specific data; at least some of the environment-specific data; a risk factor calculated by the system based on the anomaly score, environmental hazard severity at the location, proximity to dangerous assets or terrain characteristics, and duration of the anomaly condition; or a suggested route to reach the location of the subject based on the terrain characteristics.

[0149] The human subject may be able to pre-emptively suspend anomaly detection by sending a suspension command from the user device to the central server, said suspension command comprising one or more of: a plarmed suspension duration; and a reason code selected from rest break, indoor work, or equipment troubleshooting; and wherein the central server may be adapted to suspend anomaly analysis for the planned suspension duration, wherein the central server may be adapted to automatically resume anomaly analysis upon expiration of the planned suspension duration unless the human subject sends a suspension extension command from the user device.

[0150] The human subject may be able to terminate or cancel the alert or alarm after it has been generated by the central server.

[0151] The central server may utilise at least one machine learning algorithm to perform one or more of: generating the terrain classification data from satellite imagery using a convolutional neural network; generating the expected movement data using a temporal convolutional network or recurrent neural network; performing the behavioral anomaly analysis using an isolation forest algorithm, gradient boosting classifier, or autoencoder; or performing environmental hazard risk prediction using an ensemble model combining the meteorological data and the terrain characteristic data; and wherein said machine learning algorithms may be continuously retrained with new data to improve accuracy over time, including retraining triggered by confirmed incident events to learn from false negatives.

[0152] According to another aspect of the invention, there is provided a computer-implemented method for environment-based accident, incident, or risk detection for human subjects in outdoor environments, the method comprising:

[0153] (a) receiving, at a central server from a user device carried or worn by a human subject in an outdoor environment:

[0154] current location data indicating a current location of the human subject; movement data indicating at least one of: speed, direction, or stationary duration of the human subject;

[0155] (b) accessing, by the central server, environment-specific data for the outdoor environment, including:

[0156] terrain characteristic data indicating terrain characteristics and comprising geospatial data and / or terrain classification data; and meteorological data indicating weather conditions;

[0157] (c) performing, by the central server, geospatial polygon intersection analysis to determine whether the current location falls within:

[0158] an active monitoring zone, where accident, incident or risk detection is enabled; or

[0159] a non-active monitoring zone, where accident, incident or risk detection is suspended or not enabled;

[0160] (d) if the current location is within an active monitoring zone:

[0161] retrieving historical movement data for the human subject under the same or similar terrain characteristics and the same or similar weather conditions as those indicated by the environment-specific data;

[0162] calculating expected movement data based on the historical movement data and the environment-specific data;

[0163] comparing the movement data against the expected movement data to generate an anomaly score;

[0164] if the anomaly score exceeds a threshold, generating an alert signal.

[0165] If the current location is within a non-active monitoring zone, the method may comprise:

[0166] suspending or disabling anomaly detection; and

[0167] instructing the user device to reduce transmission frequency; wherein the method does not rely on predefined static activity profile templates, but instead learns individualised behavioral baselines from the human subject's historical movement data.

[0168] According to another aspect of the invention, there is provided a system for environmentbased accident, incident, or risk detection, the system comprising:

[0169] (a) a user device configured to be carried or worn by a human subject in an outdoor environment, the user device comprising:

[0170] a location determination module;

[0171] a movement detection module;

[0172] a communication module configured to communicate with a remote server without reliance on cellular telecommunications networks, said communication module utilising at least one of:

[0173] a LoRaWAN transceiver,

[0174] a satellite communication transceiver comprising low Earth orbit or geostationary satellite transceiver, or

[0175] a short-range wireless protocol comprising Wi-Fi or Bluetooth for gateway relay;

[0176] a manual alert activation button; and

[0177] a microcontroller configured to package location data, movement data, and alert signals for transmission to the remote server;

[0178] (b) a remote server comprising at least one processor and memory, configured to perform the behavioral anomaly analysis of claim 1 ;

[0179] wherein the user device transmits data to the remote server at transmission intervals that are dynamically adjusted based on:

[0180] whether the human subject is in an active monitoring zone or a non-active monitoring zone, wherein if the human subject is in a non-active monitoring zone the transmission intervals are longer than if the human subject is in an active zone; and

[0181] the anomaly score, wherein if the anomaly score exceeds the predetermined threshold the transmission intervals are shorter than if the anomaly score does not exceed the predetermined threshold.

[0182] and wherein all anomaly detection processing is performed at the remote server to conserve power and / or extend battery life of the user device.

[0183] The environment-specific data may be obtained from at least two different sources, comprising:

[0184] a first source providing general geographical or topographical data comprising public mapping databases or satellite terrain models; and

[0185] a second source providing real-time localised environmental data comprising high-resolution weather API, on-site environmental sensors, or satellite-derived weather conditions;

[0186] wherein the digital geospatial representation may be generated by the central server based on environment-specific data from the first and second sources by:

[0187] applying a trained machine learning image classification model to satellite imagery of the outdoor environment to identify terrain classifications; for contiguous regions of the outdoor environment having the same terrain classification, assigning a geospatial polygon to said contiguous regions; superimposing user-defined data onto the geospatial polygons, said user- defined data including at least one of: data as to presence of buildings in the outdoor environment, data as to asset locations in the outdoor environment, data as to infrastructure locations in the outdoor environment, data as to designated work zones in the outdoor environment, or data as to locations of hazards in the outdoor environment;

[0188] wherein the central server may utilise the digital geospatial representation to: identify the terrain classification at the current location of the human subject; and

[0189] modulate the expected movement data based on terrain classification, wherein, for a first terrain classification indicating rugged terrain, an expected speed of the human subject is lesser than for a second terrain classification indicating flat terrain;

[0190] identify proximity of the human subject to at least one aspect of the user-defined data.

[0191] The location determination module may be selected from GPS, GNSS, or signal triangulation. The non-cellular-dependent protocols may include one or more of: LoRaWAN, satellite communication, or short-range wireless communication.

[0192] The meteorological data may be high-resolution meteorological data.

[0193] The terrain classification data may be derived from satellite imagery or aerial imagery.

[0194] The central server may be configured to store or maintain the digital geospatial representation. The non-active monitoring zones may include at least one of: buildings, shelters, or designated rest areas.

[0195] The user device may further comprise a manual alert activation interface comprising at least one of: a button, a touch interface, or a voice command interface; wherein the central server may be configured to receive, from the user device, a manual alert activation signal triggered by the human subject; wherein the system may be configured to generate the output alert signal to the at least one recipient device upon receipt of the manual alert activation signal.

[0196] If the current location falls within a non-active monitoring zone, the system may be adapted to suspend or disable the behavioral anomaly analysis and reduce data transmission intervals from the user device to conserve power at the user device.

[0197] The system may perform all of the behavioral anomaly analysis at the central server, wherein the behavioural anomaly analysis is not performed at the user device, to conserve power at the user device. The present invention provides a number of optional advantages over the prior art, including, in some embodiments, providing a system (and method) that:

[0198] - Does not rely on the subject being within cell tower coverage, meaning it is suitable for use by subjects in remote areas;

[0199] - Does not rely on the subject having sophisticated or expensive hardware or software installed, and is compatible with various platforms;

[0200] - Enables the output signal to be generated automatically and without requiring actuation by the subject when the system determines, based on the input data, that an accident, incident or risk scenario or condition is likely;

[0201] - Receives multiple data items as input data, including data as to the subject and data as to the environment; and uses these synergistically by cross-referencing the subjectspecific data against the environment-specific data to determine whether the subject’s behaviour or activity is within normal bounds for that environment (and / or location within the environment);

[0202] Optionally also receives further subject-specific data items, such as past characteristics of the subject, planned activities of the subject, and / or health conditions of the subject, to further inform the analysis of whether the subject’s current behaviour or activity is within normal bounds and thus whether an output signal is required or not;

[0203] Optionally employs a staggered approach, conducting an initial assessment of whether the subject’s behaviour or activity appears to be out of normal bounds and, if so, acquiring and / or assessing further environment-specific data to identify potential hazards or risk factors.

[0204] - Generates a detailed alert or alarm for sending to appropriate third parties, optionally including the identity of the subject, the type of scenario, location data, and other collected data.

[0205] - Is applicable to monitoring and protecting workers in:

[0206] o Agriculture: Farmers, farmhands, and agricultural workers operating in remote paddocks, fields, and rural properties;

[0207] o Construction: Construction workers on large outdoor sites, road crews, infrastructure projects in remote locations;

[0208] o Emergency Services: Firefighters, search and rescue personnel, park rangers operating in wilderness areas; o Military and Defense: Personnel conducting operations, training, or patrols in remote or hostile environments;

[0209] o Forestry and Mining: Workers in forested areas, open-pit mines, or exploration sites with limited infrastructure;

[0210] o Utilities and Infrastructure: Workers maintaining power lines, pipelines, telecommunications infrastructure in remote areas;

[0211] o Any outdoor work context where workers operate beyond the range of direct human supervision and cellular network coverage.

[0212] - Is manufacturable using commercially available components (GPS modules, LoRaWAN / satellite transceivers, servers, databases) and deployable as a cloud-based software-as-a-service platform with user devices distributed to workers.

[0213] - In some embodiments, does not require biometric data of the human subject (either at all or as a primary input), which may improve efficiency of the system;

[0214] - In some embodiments, detects anomalies without reference to generic activity templates, but rather by reference to the human subject’s own historical movement data under similar terrain characteristics and weather conditions;

[0215] - In some embodiments, utilises geospatial negative space methodologies to distinguish between active zones and non-active zones, which may improve system efficiency. At the very least, the present invention provides the public with a useful choice.

[0216] In some embodiments, the present invention may provide a server-based geospatially-contextualised behavioral anomaly detection system that:

[0217] Operates without continuous biometric sensor requirements;

[0218] - Learns individual behavioral patterns without predefined activity templates;

[0219] - Dynamically adjusts risk assessment based on high-resolution environmental context;

[0220] Functions reliably in zero-terrestrial-network environments;

[0221] - Distinguishes between active work zones and non-monitored spaces to prevent false alarms;

[0222] - Provides 'virtual lifeguard' monitoring at scale where human oversight is impossible.

[0223] BRIEF DESCRIPTION OF FIGURES

[0224] Further aspects and advantages of the invention will become apparent with reference to the accompanying Figures, which are given by way of example only and in which: FIGURE 1 is a schematic representing the system according to a first preferred exemplary embodiment of the invention; and

[0225] FIGURE 2 is a flowchart representing operation of the system according to a second preferred exemplary embodiment of the invention.

[0226] DETAIEED DESCRIPTION OF FIGURES

[0227] Example 1

[0228] Figure 1 is a schematic showing a representation of the system (100) according to a first preferred exemplary embodiment of the invention. (In other embodiments the system (100) may also comprise one or more of the devices discussed herein).

[0229] The system (100) comprises a server (not shown) having a processor (not shown) and a memory (not shown). The server may be of any suitable type and configuration known in the art; for instance, the server may be in a single location or may utilise a distributed architecture. The server(s) may be hosted by a third party / ies. The memory may be a transitory or non transitory computer readable storage medium storing program instructions for facilitating the system described herein when executed by the processor. The memory may comprise a volatile or nonvolatile memory type(s), such as random-access memory (RAM), read only memory (ROM), flash memory or electrically erasable programmable read-only memory (EEPROM). The processor may be of any suitable kind such as one or more central processing units, microprocessors, application specific instruction set processors (ASIPs), application specific integrated circuits (ASICs), tensor processing units (TPUs). The processor may comprise a single processor, a plurality of processors, or may comprise a distributed processor arrangement wherein the plurality of processors are physically separate from one another. As described elsewhere in this specification, the server may be a central server that is remote from the environment (103). All or most raw data may be received by the central server, and all or most processing performed at the central server. This may minimise the amount of processing required to be performed by the user device and other sensors or devices, and as such may conserve power and extend battery life.

[0230] The system (100) is, at least in some embodiments, adapted to utilise at least one algorithm, and / or artificial intelligence (Al), for one or more of the aspects, protocols or steps for which the system (100) is adapted, and which are discussed herein. The system (100) is adapted to receive, as an input(s), subject-specific data from at least one device (102) associated with a human subject (101) in the relevant environment (103), the subject-specific data comprising at least one subject-specific data item relating to movement and / or physiology of the subject (101).

[0231] In some embodiments, the system may comprise a plurality of devices (102) associated with a respective plurality of subjects (101) in the respective environments (103) in which the subjects (101) are located.

[0232] The at least one device ( 102) may be any suitable device adapted to track the subj ect’ s location, and / or movement and / or physiology (wherein movement and / or physiology may also be referred to as the subject’s behaviour), such as one or more of: a GPS tracking device; a GNSS device; a 3-axis accelerometer; a device adapted to effect signal triangulation using LoRaWan or Wi-Fi; a physiological sensor, such as a heart rate sensor, for example an optical (PPG) sensor. There may be more than one device (102) for obtaining the respective data items, and / or a single device may be adapted to obtain multiple different types of data item. The system may, upon receipt of the data, use the data in its raw form, or the system may in some cases further process the data; for instance, upon receipt of a succession of GPS locations over time, the system may be adapted to calculate the subject’s (101) speed. The data may be transmitted to the system directly by the device(s), and / or it may be transmitted via one or more intermediate devices, such as a mobile device of the subject. The mobile device may itself be adapted to act as a device associated with the subject in the sense of collecting some or all of the subject-specific data. The mobile device may also potentially be adapted to perform one or more of the steps, aspects or protocols for which the system (100) is adapted based on the data. Though not shown in this embodiment, the system (100) may also be adapted to receive or retrieve further subject-specific data, such as the subject’s (101) past subject-specific data, at least one planned activity of the subject (101), and health data and / or mental health data of the subject (101). This further subject-specific data may be obtained from various sources, depending on datatype. For instance, the subject’s past subject-specific data may be stored in the system (100), or in a database accessible by the system (100); the subject’s health / mental health data may likewise be stored on the system (100) or in an accessible database, and / or may be entered into the system ( 100) by the subj ect ( 101 ) and / or by an authorised third party; and data as to planned activities may be entered into the system by the subject (101) and / or an authorised third party such as their employer. Different items of subject-specific data and further subject-specific data may be received at different times. For instance, data transmitted by the at least one device (102) associated with the subject may be received by the system (100) periodically; and further subject-specific data such as planned activities, health / mental health, and past subject-specific data, may be inputted into / retrieved by the system (100) at the outset of, or prior to, the subject (101) being in the environment (103), and / or may be retrieved by the system (100) on an as-needed or dynamic basis in use.

[0233] The system (100) is further adapted to receive, as an input(s), environment-specific data comprising at least one environment-specific data item relating to the environment (103) that the subject (101) is in. More particularly, the environment-specific data may comprise at least one environment-specific data item relating to the location in the environment (103) where the subject (101) is located. The location in the environment (103) where the subject (101) is located may be part of the subject-specific data received by the system (100). Thus, the environment-specific data may comprise, for example, data as to the terrain characteristics and weather conditions at the location where the subject (101) is located.

[0234] The environment-specific data may be received from a number of sources, including the device (102) associated with the subject (101), a dedicated device (104) (which may be a third-party device outside of the system (100), or may be comprised in the system (100)), an existing database (107) in the system (100) or accessible by the system (100), and / or external sources such as satellite data (106) (where the satellite may itself be considered to be a dedicated device (104) or may be considered to be a separate device). For instance, data such as ambient light and / or temperature could be obtained from a sensor(s) on the device (102) associated with the subject (101); data such as topographical and / or geographical features of the environment (103), property boundaries, and / or non-natural features of the environment (103) such as buildings or assets such as fences and paths, could be inputted via a dedicated device (104), such as by the landowner or occupant (or even the subject (101) using their device (102)) scanning the environment (103) and its features into the system (100) using photo(s) or video(s) of the environment (103), and / or could be obtained from an external source(s) such as an external database(s) (107); data such as location of any mobile assets such as vehicles and machinery could likewise be inputted via a dedicated device (104) or in some cases the device (102) associated with the subject (101), and / or could be inputted by sensors associated with said mobile assets, such as GPS trackers or RFID tags adapted to periodically transmit to the system (100) the location of the assets; and data such as weather conditions could be obtained from external sources such as satellite data (106) and / or high-resolution weather models, and / or from dedicated devices (104) related to determining weather conditions, such as one or more of: a satellite weather radar system(s); a temperature sensor(s); a light sensor(s); a sensor(s) for determining precipitation, wind and / or lightning. The system may be adapted for use with 35.5 - 36GHz Ka Band and / or High (>400 MHz) and Low bandwidth.

[0235] By way of example only, types of environment-specific data include: property boundaries; nonnatural features such as fences, buildings, padlocks, roads / tracks / paths; assets such as machinery, vehicles or other assets; trees (alive and dead); livestock; topographical or geographical features such as flatlands, fields, steep terrain, bodies of water; weather conditions such as weather events or systems (current or predicted), soil moisture, precipitation, wind speed, wind direction, temperature conditions, light conditions.

[0236] Different environment-specific data items may be received at different times. For instance, topographical data and the like may be received at the outset of, or prior to, the subject (101) being in the environment (103); data obtained from the device (102) associated with the subject may be transmitted periodically to the system (100); and data obtained from sources such as satellites (106) and / or the dedicated device(s) (104) may be transmitted periodically to the system (100) or may be retrieved by the system on an as-needed or dynamic basis in use. In an exemplary embodiment, general geographical and / or topographical features of the environment, as well as certain non-natural features such as known roads and buildings, may be obtained from an external database(s) (107) such as Google Maps. The subject (101), or a third party such as the landowner or the subject’s (101) employer, may then input further features of the environment, such as current locations of assets, current locations (and optionally type) of livestock, planned locations of assets and / or livestock over time, particular types of foliage (grasslands, woodlands) on certain parts of the property, and any other features peculiar to the environment that may be known only to the subject (101) or a third party associated with the environment. Still further environment-specific data may be received by the system at a subsequent point, when the user is in the environment; for instance, periodically-transmitted temperature and ambient light data.

[0237] The system (100) may be adapted to use at least some of the inputted environment-specific data, such as topographical and / or geographical data, data as to non-natural features of the environment (103), and / or data as to assets in the environment (103), to generate and store a digital representation of the environment (103). This may involve superposing the different types of environment-specific data items onto one another. For instance, the general geographical or topographical data from an external database(s) (107) may have superposed onto it certain more specific environment-specific data items such as information as to location of assets, livestock, and different foliage types at different locations in the environment. The digital representation may be generated using polygon geospatial data techniques, for example, whereby a plurality of geospatial polygons (namely substantially closed, multi-sided shapes defined by connected coordinate pairs (vertices) to represent areas or regions) define active monitoring zones and non-active monitoring zones. Alternatively, any other technique deemed suitable by a person skilled in the art may be used. Subsequently, data such as the subject’s (101) location may be overlaid by the system (100) onto the representation of the environment (103). Furthermore, the system (100) may be adapted to use image categorisation or classification techniques, such as by using artificial intelligence (Al). These techniques may be employed, for example, at the point when the environment-specific data is inputted, such as to recognise objects (such as buildings, fences, machinery) in images or videos uploaded by the property owner or other party; and / or may be employed at a subsequent point, such as when a preliminary likelihood of an accident, incident or risk scenario is identified, whereupon the system (100) may parse or further assess the environment-specific data (optionally including by retrieving further environment-specific data) to determine a likely environmental risk condition and / or risk factor, for inclusion in the subsequent output signal. Alternatively, or additionally, these techniques may be used by the system (100) in relation to the digital representation of the environment, to parse the digital representation and recognise its various features and characteristics.

[0238] The system (100) is adapted to compare the at least one subject-specific data item (which may include comparing instances of that data item received over a predetermined preceding period of time, such as all instances of that data item received over substantially the past 1 minute, 2 minutes, 3 minutes, 4 minutes, 5 minutes, 6 minutes, 7 minutes, 8 minutes, 9 minutes, 10 minutes, 11 minutes, 12 minutes, 13 minutes, 14 minutes, 15 minutes, 16 minutes, 17 minutes, 18 minutes, 19 minutes, 20 minutes, 21 minutes, 22 minutes, 23 minutes, 24 minutes, 25 minutes, 26 minutes, 27 minutes, 28 minutes, 29 minutes, 30 minutes, 30 minutes, 32 minutes, 33 minutes, 34 minutes, 35 minutes, 36 minutes, 37 minutes, 38 minutes, 39 minutes, 40 minutes, 41 minutes, 42 minutes, 43 minutes, 44 minutes, 45 minutes, 46 minutes, 47 minutes, 48 minutes, 49 minutes, 50 minutes, 51 minutes, 52 minutes, 53 minutes, 54 minutes, 55 minutes, 56 minutes, 57 minutes, 58 minutes, 59 minutes, or 60 minutes) against a corresponding expected subject-specific data item for the environment (103).

[0239] In particular, in a preferred embodiment, the system (101) detects the location of the subject (101) in the environment (103) based on a received GPS signal (or other suitable signal) from the device (102) associated with the subject (101), identifies the features and / or characteristics of the environment (103) at that location using the digital representation of the environment (103) and / or using the obtained environment-specific data, and generates or retrieves a corresponding expected subject-specific data item forthat location in environment (103). Thus, the term “expected subject-specific data item for the environment” may be understood to mean an expected subject-specific data item for the subject’s (101) location in the environment (103). For instance, if the subject-specific data item relates to the subject’s (101) speed, the system will generate or retrieve an expected speed for that location, taking into account the features and / or characteristics at that location such as gradient and foliage. In this regard, the system (100) utilises a synergistic approach, in that it considers not only the sensed parameters (movement, physiology) of the subject (101) but also the location of the subject (101) and the characteristics of the environment (103) at that location. Thus, the subject’s (101) pace in flat and non-challenging terrain will be treated differently by the system (100) than the subject’s (101) pace in challenging (such as steep or dense) terrain; the system (100) will generate and compare against a corresponding expected pace that takes into account the respective terrain. The corresponding expected subject-specific data item for the environment (103) may be generated by at least one algorithm or by Al, and may be preprogrammed into the system (100), such as in a database of the system (100) or accessible by the system (100), or it may be generated on an as-needed or dynamic basis by the system (100) in use. Table 1 shows an exemplary algorithm in which a corresponding expected subject-specific data item for the environment (also referred to herein as “expected movement data”) is generated as part of the process.

[0240] Upon carrying out the comparison step, the system (100) determines whether the at least one subject-specific data item is outside of a predetermined range relative to the corresponding expected subject-specific data item for the environment (103). The predetermined range (103) may likewise be preprogrammed into the system (100) or may be generated on an as-needed or dynamic basis by the system (100) in use.

[0241] In some embodiments, the system (100) is adapted to modify either the comparison step or the predetermined range based on one or more of the further subject-specific data items. For instance, if it has been inputted into the system (100) that the subject (101) generally has a slower than average walking pace, this may be factored in by the system (100) either by altering the corresponding expected pace for a given environment (103), and / or by altering the size of the predetermined range, being the allowable deviation between detected and expected pace for the environment (103). As another example, if a plarmed activity (such as use of a vehicle, or a lunch break) has been inputted into the system (100), the system will factor this in when assessing the subject’s (101) speed (and potentially other parameters) at the relevant time. As such, in this further sense, the system (100) employs a flexible and synergistic approach for assessing inputted subject-specific data items.

[0242] If the system (100) determines that the at least one subject-specific data item is outside of the predetermined range relative to the corresponding expected subject-specific data item for the environment (103), the system (100) is adapted to generate an output signal.

[0243] The output signal may comprise an alert or alarm and may be adapted to be generated automatically, such that if the subject (101) is unconscious or incapacitated the signal is still transmitted. The signal may be sent to one or more recipient devices (108, 110, 112) of authorised third party recipients, such as emergency responders (the relevant type of which may be identified by the system (100) based on the identified type of incident (including accident), risk or hazard), the subject’s employer or colleagues, and / or family members or friends. The output signal may be in the form of an SMS, email, or any other suitable form of communication, and may contain information such as the subject’s (101) identity, the type of incident (including accident), risk or hazard identified based on the data, the location of the subject (101) in the environment (103), at least some of the subject-specific data and / or environment-specific data, a risk factor calculated by the system (100), and / or at least one suggested route to reach the location of the subject (101).

[0244] Alternatively, or additionally, the output signal may comprise a message to the subject (101) themselves, such as sent to a mobile device of the subject. The message may indicate that an alert or alarm will be sent to one or more third parties if the subject (101) does not respond within a given time frame. This may enable the subject (101) to prevent transmission of “false alarms” if, for example, the subject (101) has stopped moving because they have taken a rest stop. The subject (101) may also have the option of pausing or stopping the transmission of output signals altogether, and / or of terminating or cancelling an alert or alarm after it has been sent. The third party recipient(s) may also have the option of terminating or cancelling a received alert or alarm, such as if, based on the information contained in the alert or alarm or based on an independent communication from the subject, it is clear that a false alarm has occurred.

[0245] In some embodiments, the system (100) is adapted to perform a preliminary step of assessing at least some subject-specific data in the manner described above, to determine whether an accident, incident or risk scenario is likely to have occurred or be occurring; and if so, then further assess the environment-specific data (optionally including by retrieving further environment-specific data) to determine a likely environmental risk condition; and if an environmental risk condition is detected, generate the output signal. This enables the system (100) to identify any environmental factors that are likely to have caused the accident or incident, or that have presented or will present a hazard or risk. For instance, the system (100) may identify any potentially hazardous machinery or geographical features in the vicinity of the subject (103); or may identify any current or predicted adverse weather conditions. This information may be included in the output signal, being the alert or alarm.

[0246] The present invention has particular application to subjects who are in remote and / or difficult environments, such as agricultural workers, miners, or hikers. The invention does not rely on cell coverage, and so is more suited to such environments. The invention also enables automatic sending of alarms in case of a detected incident, accident or risk scenario, meaning an alarm will still be sent even if the subject is incapacitated. Furthermore, the invention synergistically compares the subject’s parameter(s) (movement, physiology) against their location within the environment and the environmental features and / or characteristics of that location, and determines a corresponding expected parameter(s) for that location in order to determine whether the subject’s detected parameter(s) are within or outside of acceptable bounds. As such, the system employs a synergistic methodology to increase the accuracy with which accidents, incidents or risk scenarios are identified. This is in contrast to conventional systems which tend to only consider the subject’s parameter(s), without regard to the characteristics of the environment or location where the subject is located and the effect that these environmental characteristics are likely to have on the subject’s parameters.

[0247] The system of the invention may be adapted to be operable without sophisticated or expensive hardware or software. The system may be compatible with various available platforms such as Azure and Amazon web services. Elements of the system may be communicable with other relevant elements of the system without reliance on cell towers or without needing to be within an area of cellphone coverage. The invention may rely on Internet of Things (loT) technology for facilitating communicability between elements of (or associated with) the system (such as the device associated with the subject, as well as the dedicated devices and any other devices) with other elements of the system (such as the server). Alternatively or additionally, communicability may be provided by a satellite connection(s), such as low Earth orbit and / or geostationary satellite connection(s), via a suitable communication protocol(s); and / or by a ground based network(s) such as LoraWan, Bluetooth, and / or Wi-Fi.

[0248] The invention may comprise a mobile app, such as programmed onto a mobile device of the subject. The mobile app may receive data from the at least one device associated with the subject, and transmit the data to the server. The mobile app may also facilitate inputs into the system (100) by the subject, and / or may facilitate receipt of messages or other data from third parties to the subject. The mobile device may itself collect some data, for instance may comprise one or more sensors such as GPS sensors or physiological sensors; and / or may be adapted to perform one or more of the steps, aspects or protocols for which the system (100) is adapted. In other embodiments, the one or more devices associated with the subject (such as sensors) may communicate with the server directly.

[0249] Figure 2 is a flowchart showing the system (200) according to a second preferred exemplary embodiment of the invention. The system (200) according to this embodiment is generally similar to the first embodiment, but with a more detailed sequence.

[0250] At 202A and 202B, certain subject-specific data (and / or further subject-specific data) and environment-specific data can be preprogrammed into the system (200), as described above, at the outset of or prior to the subject being in the environment. For instance, and without limitation, the geographical and / or topographical features of the environment, property boundaries, and such like can be preprogrammed into the system (200), as can information relating to the subject’s health conditions and / or planned activities. Other information, such as past subject-specific data of the subject for example, may already be in, or accessible by, the system (200).

[0251] At 204, the system (200) receives periodic subject-specific data from the at least one device associated with the subject. This can be data as to the subject’s location, movement, and / or the subject’s physiological parameters such as heart rate, respiratory rate, temperature, blood pressure, et cetera.

[0252] At 206, the system (200) assesses the received data. Specifically, at 206A, the system (200) determines the location of the subject in the environment, based on GPS (or other) readings from the at least one device associated with the subject. At 206B, the system (200) determines one or more features and / or characteristics of the environment at that location, based on the digital representation of the environment created by the system (200) based on the environment-specific data received at 206B (and potentially also received at other stages of the process), or alternatively directly based on the environment-specific data received at 206B. At 208, the system (200) retrieves (and / or receives) any further relevant data that may affect the system’s (200) assessment of the received subject-specific data. For instance, the system (200) may retrieve (and / or receive) any data relating to planned activities, the subject’s health conditions, or the subject’s past subject-specific data. This information may be used at step 206C and / or 206D.

[0253] At 206C, the system (200) determines an expected subject-specific data item, being of a type that corresponds to the received subject-specific data item received at step 204. For instance, if the received subject-specific data item is the subject’s pace, then the corresponding expected subject-specific data item will also relate to pace; or if the received subject-specific data item is the subject’s heartrate, then the corresponding expected subject-specific data item will also relate to heartrate. The expected subject-specific data item is arrived at with respect to the determined environmental features and / or characteristics at the subject’s location. For instance, steeper terrain will generally entail a slower expected pace than flat terrain. Optionally, at 206C any further relevant data obtained at 208 may be used to modify the expected subject-specific data item. For instance, if the subject is known to be a slower-than-average walker, then the expected pace may be decreased accordingly.

[0254] At 206D, the received subject-specific data item is compared against the corresponding expected subject-specific data item, to determine whether the received subject-specific data item is outside of a predetermined range (i.e. deviation) relative to the corresponding expected subject-specific data item for the environment. Optionally, at 206D any further relevant data obtained at 208 may be used to modify the predetermined range. For instance, if the subject is known to be a slower-than-average walker, then the predetermined range for pace or speed may be modified accordingly, such as by the predetermined range being broadened to account for the subject’s slower-than-average pace to prevent false alarms.

[0255] At 210, if it is determined that the received subject-specific data item is not outside of the predetermined range relative to the corresponding expected subject-specific data item for the environment, the system (200) takes no action and continues to monitor further periodically-received subject-specific data.

[0256] At 212, if it is determined that the received subject-specific data item is outside of the predetermined range relative to the corresponding expected subject-specific data item for the environment, the system (200) takes further action.

[0257] At 214, the system (200) retrieves further environment-specific data (and / or retrieves a subset of the existing environment-specific data), and at 216 the system (200) assesses this to identify any environmental risk factors or hazards that may be relevant to the subject’s deviation from expected parameters. This may include geographical or topographical features in the vicinity, hazardous assets or machinery in the vicinity, or present or predicted adverse weather conditions.

[0258] The system (200) may also calculate a risk score, based on the subject-specific data (in particular any anomalies detected in the data) and any identified environmental risk factors or hazards.

[0259] At 218, the system (200) sends an output signal, being an alert or alarm via SMS, email or another suitable medium, to the devices of the one or more designated third parties associated with the subject, the alert or alarm containing information such as the identity of the subject, the subject’s location (optionally as coordinates, and optionally shown on a map of the environment), the type of detected accident, incident or risk scenario, the risk score, any identified environmental risk factors or hazards, and some or all of the received subject-specific data and environment-specific data. The alert or alarm may also comprise one or more suggested routes to reach the subject. For instance, a route may be provided for access on foot or via vehicle, and / or a landing location may be suggested for access via helicopter. The route may be at least in part based on the route that the subject took to get to the location. The route may also be at least in part based on any known roads or pathways leading to, or in the vicinity of, the location, based on the environment-specific data and / or the digital representation of the environment.

[0260] Example 2

[0261] A third preferred exemplary embodiment of the invention will now be described.

[0262] The invention according to the third preferred exemplary embodiment is similar in many respects to the embodiments described under Example 1 above, and, unless the context requires otherwise, the above description will be understood to apply to this embodiment also. Likewise, unless the context otherwise requires, the description of the third preferred exemplary embodiment will be understood to apply also to the first and second preferred exemplary embodiments.

[0263] The third preferred exemplary embodiment has the following salient aspects. Firstly, the human subject’s biometric or physiological data is not required, either at all or as a primary input. Only the subject’s movement data (along with location) is required as a primary input, insofar as subject-specific data is concerned. This may provide efficiency gains in that it avoids the need for sensors on the subject, and avoids the need for continuous transmission, by the user device to the server, of data from said sensors, thereby conserving battery power of the user device. Secondly, the “corresponding expected subject-specific data item for the environment”, which is referred to in this third preferred exemplary embodiment as “expected movement data”, is obtained primarily on the basis of the human subject’s historical movement data. More particularly, the system selects, from a historical movement database, an instance of the subject’s prior movement data which occurred under similar environmental conditions (terrain characteristics and weather conditions) as those presently being experienced by the subject. Based on this, expected movement data for the subject is calculated, and is compared against the subject’s current movement data to determine an anomaly score.

[0264] Advantageously, this allows an accurate, personalised, real-time movement prediction to be made for the human subject, based on that same subject’s prior movement patterns and taking into account current environmental conditions. In other words, the system takes into account what is normal for this specific individual in this specific location under these specific conditions, in what can be referred to as individualised pattern learning with environmental modulation. This can be contrasted with rigid, preprogrammed “activity profiles” seen in some prior art systems, whereby particular activities have preprogrammed movement profiles associated with them without regard either to the subject’s individual movement patterns or baselines or to the terrain characteristics and weather conditions. For example, a prior art system may have a preprogrammed “farm work” activity profile which it would apply generically to all farmers. In contrast, in the present embodiment of the invention, for a given location (say, a paddock), the system uses historical movement data to ascertain that Worker 1 typically moves at 150m / hr in that location, while Worker 2 typically moves at 400m / hr in the same location. Accordingly, the expected movement data is tailored to each worker individually based on past behaviour.

[0265] Thirdly, in this embodiment, prior to undertaking behavioral anomaly analysis of the subject, the system determines whether the subject’s current location falls within an active monitoring zone or a non-active monitoring zone by performing geospatial polygon intersection analysis (also referred to as a “negative space” methodology). If the location is in an active monitoring zone, the system performs the behavioral anomaly analysis, while if the location is in a nonactive monitoring zone, the system suspends or disables the behavioral anomaly analysis. This may promote efficiency of the system in avoiding unnecessary analysis when the subject is in a safe location or not in a location of potential risk (such as a rest area). It may also prevent the user device from unnecessarily conveying data when in such a location, conserving power. It may also militate against false alarms.

[0266] The third preferred exemplary embodiment will now be discussed more particularly. According to this embodiment, a computerised system for environment-based accident, incident, or risk detection for human subjects in outdoor environments comprises at least one user device, and a central server, being remote from the at least one user device.

[0267] The at least one user device is configured to be carried or worn by a human subject, and comprises: a location determination module; a movement detection module comprising at least one of an accelerometer, gyroscope, or other motion sensor; and a communication module configured to transmit data via non-cellular-dependent protocols. The central server comprises at least one processor and memory. The central server is adapted to perform (or cause to be performed) all or most of the data processing (including artificial intelligence / machine learning functions) of the invention. The central server receives data from the user device, and performs the processing steps described herein. In this way, processing requirements at the user device side are minimised (limited to data collection, packaging and transmission), to limit power consumption by the user device. The central server is adapted to be communicable with the user device without reliance on terrestrial cellular networks. For example, the primary communication protocol may be LoRaWAN, satellite (LEO / GEO), or other non-cellular protocols.

[0268] The central server is configured to receive, from the at least one user device, subject-specific data comprising: current location data indicating a current location of the human subject in an outdoor environment; and movement data relating to the human subject, including at least one of: speed, direction, acceleration, or stationary duration. The term “current location data” will be understood to include both data from the present moment in time, and optionally also data from some time in the recent past, such as a predetermined time interval extending into the past. It will likewise be understood that the movement data may comprise data collected over said time interval.

[0269] The central server is further configured to receive or access environment-specific data indicating environmental conditions for the outdoor environment, and in particular for the location within the outdoor environment where the human subject is located. The environmentspecific data comprises terrain characteristic data comprising one or more of: geospatial data including topographical and geographical features; or terrain classification data which may be used for terrain classification, such as by identifying the density and / or type of vegetation to distinguish whether a region comprises, for example, bushland, forest, field or orchard. The environment-specific data further comprises meteorological data indicating localised weather conditions for the outdoor environment including at least one of: temperature, precipitation, wind, or visibility.

[0270] The environment-specific data may be received or accessed from a plurality of different sources, such as a first source providing general geographical or topographical data comprising public mapping databases or satellite terrain models, and a second source providing real-time localised environmental data comprising high-resolution weather API, on-site environmental sensors, or satellite-derived weather conditions. The central server is configured to generate a digital geospatial representation (also referred to as a geospatial polygon-based data structure presentable via visual map interface) of the outdoor environment based on the environment-specific data. The digital geospatial representation comprises a plurality of geospatial polygons (namely substantially closed, multisided shapes defined by connected coordinate pairs (vertices) to represent areas or regions) defining one or more active monitoring zones and one or more non-active monitoring zones. The digital geospatial representation is generated using polygon geospatial data techniques comprising at least one of: GeoJSON format, Well-Known Text (WKT) format, or shapefile format. Geospatial polygon intersection analysis is performed using spatial indexing selected from R-tree or Quad-tree indexing for computational efficiency. Active monitoring zones designate outdoor areas where accident, incident or risk detection monitoring is enabled. Nonactive monitoring zones, aka negative spaces, designate areas (such as buildings, rest areas, shelters) where accident, incident or risk detection monitoring is suspended or not enabled. The digital geospatial representation may also take into account proximity of potential hazards such as assets, machinery, water bodies, and other hazardous zones. By defining where monitoring should not occur, the system may significantly reduce false alarms and power consumption of the user device.

[0271] Where the environment-specific data is received or accessed from a plurality of different sources, the digital geospatial representation may be generated based on data from said plurality of different sources. Integration of multiple environment-specific data sources into the geospatial representation may enable context-specific behavioral expectation adjustment that adapts to localised environmental conditions at a resolution finer than general regional weather or terrain data.

[0272] The system may thereby integrate environment-specific data from a plurality of sources, such as satellite imagery, which may be used as terrain classification data; localised weather models providing high-resolution meteorological data providing micro-climate conditions; and topographical data such as elevation, slope, terrain ruggedness affecting movement expectations. The system may then process this data and combine it into a structured context model, being the digital geospatial representation or “digital twin”. This may allow the system to use environmental data to modulate behavioral expectations dynamically, not as a secondary alert trigger after biometric thresholds are exceeded. The central server is configured to determine, for the current location of the human subject, whether the location falls within an active monitoring zone or a non-active monitoring zone by performing geospatial polygon intersection analysis.

[0273] If the location falls within an active monitoring zone, the central server is configured to perform behavioral anomaly analysis, which is preferably performed wholly at the central server, not at the user device, to minimise power consumption at the user device. The behavioral anomaly analysis comprises, firstly, retrieving historical movement data for the human subject from a historical movement database, said historical movement data comprising prior movement data of the human subject under the same or similar terrain characteristics and the same or similar weather conditions as those indicated by the environment-specific data. The historical movement data may relate to the subject’s prior movement at the same location as the current location, or a different location with similar terrain characteristics (such as geology, topology, and / or intended use). Secondly, calculating expected movement data for the human subject at the current location based on: (i) the historical movement data, (ii) the terrain characteristics indicated by the environment-specific data, and (iii) the weather conditions indicated by the environment-specific data. Thirdly, comparing the movement data of the human subject against the expected movement data to generate an anomaly score.

[0274] The central server is configured to, if the anomaly score exceeds a predetermined threshold, generate an output alert signal to at least one recipient device.

[0275] If the location falls within a non-active monitoring zone, the central server is configured to suspend or disable behavioral anomaly analysis and reduce data transmission intervals from the user device. This prevents data being sent when the user is in a location that does not require safety monitoring, and thereby conserves power of the user device.

[0276] Table 1 provides an exemplary algorithm for effecting the third exemplary preferred embodiment of the invention.

[0277] Table 1: >

[0278] >

[0279] &

[0280] >

[0281]

[0282] In Step 1, based on the location of the human subject, the central server identifies whether the subject is in an active or non-active zone. The central server also begins processing environment-specific data. In Step 2, the central server begins the behavioural anomaly analysis by obtaining, from a historical movement database, the subject’s prior movement data under similar environmental conditions. In Step 3, the central server performs modulation of the prior movement data based on environmental modulation factors, to arrive at the expected movement data. In Step 4, the system generates an anomaly score by comparing the subject’s current movement data to the expected movement data. In Step 5, based on the anomaly score, the system determines which type of alert / alarm (if any) to generate.

[0283] The central server may employ at least one artificial intelligence (also referred to herein as machine learning) model or module. For example, the following Al models may be used: Model 1 : Terrain Classification Model (for terrain classification based on terrain classification data)

[0284] Architecture: ResNet-50 CNN

[0285] Input: RGB satellite tile (256x256 pixels, 0.5m resolution)

[0286] Output: 12 terrain categories + confidence score Update: Quarterly or event-triggered

[0287] Model 2: Individual Behavioral Baseline Model (for use in generating expected movement data based on historical movement data, by establishing a self-learning behavioral model for each individual without predefined activity templates)

[0288] - Architecture: Temporal Convolutional Network (TCN) with attention Training: Individual's historical movement (minimum 14 days initial) Output: Probability distribution of expected movement parameters

[0289] Privacy: One model per individual (no cross-user data)

[0290] Model 3 : Anomaly Detection Model (for generating an anomaly score)

[0291] - Architecture: Isolation Forest + gradient boosting classifier

[0292] Output: Anomaly score (0-100) + anomaly type classification

[0293] Update: Weekly retraining, immediate after confirmed incidents

[0294] A machine learning regression model may be applied to obtain predicted values, being expected speed, direction, and dwell time, of the human subject in view of the subject’s prior movement data and the environment-specific data. The predicted values are adjusted based on environmental modulation factors selected from temperature adjustment, precipitation adjustment, terrain ruggedness adjustment, time-of-day adjustment, or asset proximity adjustment. The machine learning regression model is continuously retrained with new movement data to adapt to changing behavioral patterns of the human subject over time. A machine learning model(s) may be applied to perform one or more of: generating the terrain classification data from satellite imagery using a convolutional neural network (such as via pixel-level terrain categorisation and vectorisation into geospatial polygons); generating the expected movement data using a temporal convolutional network or recurrent neural network; performing the behavioral anomaly analysis using an isolation forest algorithm, gradient boosting classifier, or autoencoder; or performing environmental hazard risk prediction using an ensemble model combining the meteorological data and the terrain characteristic data. Said machine learning algorithms are continuously retrained with new data to improve accuracy over time, including retraining triggered by confirmed incident events to learn from false negatives.

[0295] In the foregoing disclosure, it will be understood that, where reference is made to data being inputted into the system, this may be done via an appropriate device adapted to receive or detect input of data and transmit or enable transmission of the data. For instance a computer, mobile device, or sensor.

[0296] It will of course be realised that while the foregoing has been given by way of illustrative example of this invention, all modifications and variations thereto as would be apparent to persons skilled in the art are deemed to fall within the broad scope and ambit of this invention as is hereinbefore described.

[0297] The invention may further be said to consist in the individual parts, components, and features described herein, alone or in any combination of two or more of same.

[0298] If any reference numeral(s) is / are used in a claim or claims then such reference numeral(s) should not be considered as limiting the scope of that respective claim or claims(s) to any particular embodiment of the drawings.

[0299] It is acknowledged that the term ‘comprise’ may, under varying jurisdictions, be attributed with either an exclusive or an inclusive meaning. For the purpose of this specification, and unless otherwise noted, the term ‘comprise’ shall have an inclusive meaning - i.e. it will be taken to mean an inclusion of not only the listed components it directly references, but also other nonspecified components or elements. This rationale will also be used when the term ‘comprised’ or 'comprising' is used in relation to one or more steps in a method or process.

[0300] It is acknowledged that the terms “configured to” and “adapted to” may be used interchangeably herein.

Claims

CLAIMS:A computerised system for environment-based accident, incident, or risk detection for human subjects in outdoor environments, the system comprising:(a) at least one user device configured to be carried or worn by a human subject, the at least one user device comprising:a location determination module;a movement detection module comprising at least one of an accelerometer, gyroscope, or other motion sensor; anda communication module configured to transmit data via non-cellular- dependent protocols; and(b) a central server comprising at least one processor and memory, configured to: receive, from the at least one user device, subject-specific data comprising: current location data indicating a current location of the human subject in an outdoor environment; andmovement data relating to the human subject, including at least one of: speed, direction, acceleration, or stationary duration;receive or access environment-specific data indicating environmental conditions for the outdoor environment, said environment-specific data comprising:terrain characteristic data indicating terrain characteristics, said terrain characteristic data comprising one or more of:geospatial data including topographical and geographical features;terrain classification data;meteorological data indicating localised weather conditions for the outdoor environment including at least one of: temperature, precipitation, wind, or visibility; andgenerate a digital geospatial representation of the outdoor environment, comprising:a plurality of geospatial polygons defining one or more active monitoring zones and one or more non-active monitoring zones; wherein active monitoring zones designate outdoor areas where accident, incident or risk detection monitoring is enabled; and wherein non-active monitoring zones designate areas where accident, incident or risk detection monitoring is suspended or not enabled; determine, for the current location of the human subject, whether the location falls within an active monitoring zone or a non-active monitoring zone by performing geospatial polygon intersection analysis;if the location falls within an active monitoring zone, perform behavioral anomaly analysis comprising:retrieving historical movement data for the human subject from a historical movement database, said historical movement data comprising prior movement data of the human subject under the same or similar terrain characteristics and the same or similar weather conditions as those indicated by the environment-specific datacalculating expected movement data for the human subject at the current location based on: (i) the historical movement data, (ii) the terrain characteristics indicated by the environment-specific data, and (iii) the weather conditions indicated by the environment-specific data; comparing the movement data of the human subject against the expected movement data to generate an anomaly score;if the anomaly score exceeds a predetermined threshold, generate an output alert signal to at least one recipient device.2.The system of claim 1, wherein the subject-specific data comprises movement data and current location data, and does not require continuous physiological biometric data for the anomaly analysis, thereby reducing power consumption and hardware cost of the system compared to biometric -dependent systems.3.The system of claim 1, wherein the subject-specific data further comprises data relating to physiological parameters of the human subject selected from at least one of heart rate, respiratory rate, blood pressure, or body temperature, wherein said physiological parameters are used as supplementary context data and not as primary data for the anomaly analysis.4.The system of claim 1, wherein the at least one user device comprises one or more of: a GPS device; a GNSS device; a 3-axis accelerometer; a device adapted to effect signal triangulation using LoRaWAN or Wi-Fi; or a physiological sensor comprising an optical photoplethysmography (PPG) sensor.5.The system of claim 1, wherein the at least user device comprises a mobile device programmed with a mobile application configured to transmit data to the central server.6.The system of claim 1, wherein the at least one user device is communicable with the system without reliance on cell towers, wherein the at least one user device is communicable with the system using one or more of: Internet of Things (loT) technology; a satellite connection via low Earth orbit or geostationary satellite, via a suitable communication protocol; or a ground-based network comprising LoRaWAN, Bluetooth, or Wi-Fi.7.The system of claim 1, wherein the central server is configured to receive, as an input, further subject-specific data relating to one or more of: at least one planned activity of the human subject in the environment; past subject-specific data of the human subject; or health data and / or mental health data of the human subject.8.The system of claim 7, wherein said further subject-specific data relating to at least one planned activity of the human subject in the environment includes data relating to one or more of: the location of the at least one planned activity in the environment; the time of the planned activity; or machinery or assets involved in the planned activity.9.The system of claim 7, wherein said past subject-specific data of the human subject includes data as to the environmental conditions corresponding to said past subjectspecific data, enabling correlation of historical movement data with historical environmental conditions.10.The system of claim 1, wherein the environment-specific data further comprises data relating to one or more of: property boundaries; non-natural features of the environment; assets or machinery in the environment; light conditions in the environment; temperature conditions in the environment; or current and / or predicted weather conditions in the environment selected from precipitation, wind speed, or wind direction.11.The system of claim 1, wherein at least some of the environment-specific data is preprogrammed into the system prior to the subject being in the environment.12.The system of claim 1, wherein at least some of the environment-specific data is obtained from multiple sources of environment-specific data, comprising: a publicly available database or source; and data from a party associated with the environment and / or data from the human subject; wherein at least some geographical and / or topographical data, and / or data as to property boundaries, is obtained from the publicly available database, and wherein at least some data as to non-natural features and / or assets in the environment is obtained from the party associated with the environment and / or from the human subject.13.The system of claim 1, wherein the central server is adapted to use at least some of the environment-specific data to generate and store the digital geospatial representation of the environment by: applying a trained convolutional neural network image classification model to satellite imagery or aerial imagery of the outdoor areas of the environment; said trained convolutional neural network image classification model outputting pixel-level terrain classifications which classify each of a plurality of pixels into a terrain classification; converting the classified pixels into vector geospatial polygons representing contiguous regions of the same terrain classification; and storing said geospatial polygons in a spatial database indexed by geographic coordinates for rapid intersection queries.14.The system of claim 13, wherein the digital geospatial representation comprises one or more of: topographical and / or geographical features of the environment; non-natural features of the environment; or assets in the environment.15.The system of claim 13, wherein the digital geospatial representation is generated using polygon geospatial data techniques comprising at least one of: GeoJSON format, Well- Known Text (WKT) format, or shapefile format, and wherein the geospatial polygon intersection analysis is performed using spatial indexing selected from R-tree or Quadtree indexing.16.The system of claim 1, wherein, in use, the current location data is overlaid by the central server onto the digital geospatial representation of the environment, and wherein the central server is configured to enable or provide a visual map interface, said visual map interface displaying: the human subject's current location and recent movement trail derived from the movement data; boundaries of the active monitoring zones and non-active monitoring zones; environmental hazard indicators selected from extreme weather zones, flooded areas, or restricted zones; and locations of assets, infrastructure, and other workers in proximity of the human subject.17.The system of claim 1, wherein at least some of the environment-specific data is obtained by the central server via one or more dedicated devices, comprising one or more of: GPS devices or RFID tags associated with mobile assets in the environment, adapted to periodically update the system as to their location in the environment; a satellite weather radar system; a temperature sensor; a light sensor; or a sensor for determining precipitation, wind, or lightning.18.The system of claim 17, wherein the one or more dedicated devices are communicable with the central server without reliance on cell towers, wherein the one or more dedicated devices are communicable with the system using one or more of: Internet of Things (loT) technology; a satellite connection comprising low Earth orbit and / or geostationary satellite connection; or a ground-based network comprising LoRaWAN, Bluetooth, or Wi-Fi.19.The system of claim 1, wherein at least some of the environment-specific data is obtained, retrieved and / or assessed by the central server upon occurrence of a trigger; wherein the trigger comprises a preliminary indication of an accident, incident or risk condition determined by the system based on the subject-specific data.20.The system of claim 19, wherein if the central server determines a preliminary indication of an accident, incident or risk condition, the central server is configured to obtain, retrieve and / or assess said at least some of the environment-specific data todetect an environmental risk condition; and if an environmental risk condition is detected, generate the output signal.21.The system of claim 20, wherein the environmental risk condition comprises one or more of: adverse current and / or predicted weather conditions; topographical or geographical features associated with risk; or assets or machinery associated with risk.22.The system of claim 1, wherein the expected movement data for the human subject is dynamically generated by the central server in real-time as current location data is received, said dynamic generation comprising: querying the historical movement database for prior movement data of the human subj ect under the same or similar terrain characteristics and the same or similar weather conditions as those indicated by the environment-specific data; applying a machine learning regression model to obtain predicted values, being expected speed, direction, and dwell time, of the human subject in view of the environment-specific data; adjusting the predicted values based on environmental modulation factors selected from temperature adjustment, precipitation adjustment, terrain ruggedness adjustment, time-of-day adjustment, or asset proximity adjustment; and wherein the machine learning regression model is continuously retrained with new movement data to adapt to changing behavioral patterns of the human subject over time.23.The system of claim 22, wherein, where there is insufficient historical movement data for the human subject, the expected movement data is derived from population-based historical movement data from a plurality of other human subjects under the same or similar terrain characteristics and the same or similar weather conditions as those indicated by the environment-specific data, and wherein, when sufficient historical movement data for the human subject becomes available, the system transitions from using the population-based historical movement data to using the historical movement data for the human subject.24.The system of claim 1, wherein the historical movement data comprises prior movement data of the human subject obtained over a predetermined period of time.25.The system of claim 1, wherein obtaining or determining the expected movement data comprises: identifying the current location of the human subject within the environment using GPS coordinates; performing a spatial query to identify geospatial polygon(s) containing the current location; retrieving data associated with the identified polygon(s), said data comprising one or more terrain characteristics including: terrain type, terrain ruggedness score, one or more typical activities performed in the identified polygon(s), or proximity to assets or hazards; retrieving current weather conditions at the current location from a high-resolution meteorological data source; querying the historical movement database for prior movement data of the human subject in the same polygon or adjacent polygons under weather conditions within a similarity threshold of the current weather conditions; computing statistical metrics from the historicalmovement data selected from mean speed, standard deviation of speed, or typical dwell time distribution; applying environmental modulation factors to adjust statistical metrics based on the current weather conditions; and defining the expected movement data as acceptable ranges around the adjusted statistical metrics.26.The system of claim 25, wherein the one or more terrain characteristics are identified based on the digital geospatial representation of the environment.27.The system of claim 7, wherein said comparing of the current movement data against the expected movement data comprises using the further subject-specific data to either: modify the expected movement data; or modify the calculation of the anomaly score and / or the predetermined threshold associated with the anomaly score.28.The system of claim 1, wherein the output alert signal generated by the central server comprises an alert or alarm sent to the recipient device of at least one third party, being one or more of: emergency services; an employer or colleague of the human subject; a family member or friend of the human subject; or another designated third party.29.The system of claim 28, wherein the alert or alarm comprises information including one or more of: the identity of the human subject; the current location of the subject; an indication or type of accident, incident or risk; an environmental risk condition; at least some of the subject-specific data; at least some of the environment-specific data; a risk factor calculated by the system based on the anomaly score, environmental hazard severity at the location, proximity to dangerous assets or terrain characteristics, and duration of the anomaly condition; or a suggested route to reach the location of the subject based on the terrain characteristics.30.The system of claim 1, wherein the human subject can pre-emptively suspend anomaly detection by sending a suspension command from the user device to the central server, said suspension command comprising one or more of: a planned suspension duration; and a reason code selected from rest break, indoor work, or equipment troubleshooting; and wherein the central server is adapted to suspend anomaly analysis for the planned suspension duration, wherein the central server is adapted to automatically resume anomaly analysis upon expiration of the planned suspension duration unless the human subject sends a suspension extension command from the user device.31.The system of claim 28, wherein the human subject can terminate or cancel the alert or alarm after it has been generated by the central server.32.The system of claim 1, wherein the central server utilises at least one machine learning algorithm to perform one or more of: generating the terrain classification data fromsatellite imagery using a convolutional neural network; generating the expected movement data using a temporal convolutional network or recurrent neural network; performing the behavioral anomaly analysis using an isolation forest algorithm, gradient boosting classifier, or autoencoder; or performing environmental hazard risk prediction using an ensemble model combining the meteorological data and the terrain characteristic data; and wherein said machine learning algorithms are continuously retrained with new data to improve accuracy over time, including retraining triggered by confirmed incident events to learn from false negatives.33.A computer-implemented method for environment-based accident, incident, or risk detection for human subjects in outdoor environments, the method comprising:(a) receiving, at a central server from a user device carried or worn by a human subject in an outdoor environment:current location data indicating a current location of the human subject; movement data indicating at least one of: speed, direction, or stationary duration of the human subject;(b) accessing, by the central server, environment-specific data for the outdoor environment, including:terrain characteristic data indicating terrain characteristics and comprising geospatial data and / or terrain classification data; and meteorological data indicating weather conditions;(c) performing, by the central server, geospatial polygon intersection analysis to determine whether the current location falls within:an active monitoring zone, where accident, incident or risk detection is enabled; ora non-active monitoring zone, where accident, incident or risk detection is suspended or not enabled;(d) if the current location is within an active monitoring zone:retrieving historical movement data for the human subject under the same or similar terrain characteristics and the same or similar weather conditions as those indicated by the environment-specific data;calculating expected movement data based on the historical movement data and the environment-specific data;comparing the movement data against the expected movement data to generate an anomaly score;if the anomaly score exceeds a threshold, generating an alert signal.

34. The method of claim 33, wherein, if the current location is within a non-active monitoring zone, the method comprises:suspending or disabling anomaly detection; andinstructing the user device to reduce transmission frequency;wherein the method does not rely on predefined static activity profile templates, but instead learns individualised behavioral baselines from the human subject's historical movement data.35.A system for environment-based accident, incident, or risk detection, the system comprising:(a) a user device configured to be carried or worn by a human subject in an outdoor environment, the user device comprising:a location determination module;a movement detection module;a communication module configured to communicate with a remote server without reliance on cellular telecommunications networks, said communication module utilising at least one of:a LoRaWAN transceiver,a satellite communication transceiver comprising low Earth orbit or geostationary satellite transceiver, ora short-range wireless protocol comprising Wi-Fi or Bluetooth for gateway relay;a manual alert activation button; anda microcontroller configured to package location data, movement data, and alert signals for transmission to the remote server;(b) a remote server comprising at least one processor and memory, configured to perform the behavioral anomaly analysis of claim 1 ;wherein the user device transmits data to the remote server at transmission intervals that are dynamically adjusted based on:whether the human subject is in an active monitoring zone or a non-active monitoring zone, wherein if the human subject is in a non-active monitoring zone the transmission intervals are longer than if the human subject is in an active zone; andthe anomaly score, wherein if the anomaly score exceeds the predetermined threshold the transmission intervals are shorter than if the anomaly score does not exceed the predetermined threshold.and wherein all anomaly detection processing is performed at the remote server to conserve power and / or extend battery life of the user device.36.The system of claim 1, wherein:(a) the environment-specific data is obtained from at least two different sources, comprising:a first source providing general geographical or topographical data comprising public mapping databases or satellite terrain models; anda second source providing real-time localised environmental data comprising high-resolution weather API, on-site environmental sensors, or satellite-derived weather conditions;(b) the digital geospatial representation is generated by the central server based on environment-specific data from the first and second sources by:applying a trained machine learning image classification model to satellite imagery of the outdoor environment to identify terrain classifications; for contiguous regions of the outdoor environment having the same terrain classification, assigning a geospatial polygon to said contiguous regions; superimposing user-defined data onto the geospatial polygons, said user- defined data including at least one of: data as to presence of buildings in the outdoor environment, data as to asset locations in the outdoor environment, data as to infrastructure locations in the outdoor environment, data as to designated work zones in the outdoor environment, or data as to locations of hazards in the outdoor environment;(c) the central server utilises the digital geospatial representation to:identify the terrain classification at the current location of the human subject; andmodulate the expected movement data based on terrain classification, wherein, for a first terrain classification indicating rugged terrain, an expected speed of the human subject is lesser than for a second terrain classification indicating flat terrain;identify proximity of the human subject to at least one aspect of the user-defined data.

37. The system of claim 1, wherein:the location determination module is selected from GPS, GNSS, or signal triangulation; the non-cellular-dependent protocols include one or more of: LoRaWAN, satellite communication, or short-range wireless communication;the meteorological data is high-resolution meteorological data;the terrain classification data is derived from satellite imagery or aerial imagery; the central server is configured to store or maintain the digital geospatial representation; andthe non-active monitoring zones include at least one of: buildings, shelters, or designated rest areas.

38. The system of claim 1, wherein the user device further comprises a manual alert activation interface comprising at least one of: a button, a touch interface, or a voice command interface; wherein the central server is configured to receive, from the user device, a manual alert activation signal triggered by the human subject; wherein the system is configured to generate the output alert signal to the at least one recipient device upon receipt of the manual alert activation signal.

39. The system of claim 1, wherein, if the current location falls within a non-active monitoring zone, the system is adapted to suspend or disable the behavioral anomaly analysis and reduce data transmission intervals from the user device to conserve power at the user device.

0. The system of claim 1, wherein the system performs all of the behavioral anomaly analysis at the central server, wherein the behavioural anomaly analysis is not performed at the user device, to conserve power at the user device.