Smoke and fire recognition, fire forecasting, and monitoring

Abstract

A method and system to receive, one or more first images, one or more second images, one or more ambient weather related information, and one or more land related information, wherein the one or more land related information comprise vegetation features, terra firma topography, elevation, slope and aspect, of one or more regions of interest, of a geographical region; to map automatically, one or more risk areas of the one or more regions of interest; to recognize automatically, one or more smoke or fire related signals; and to predict computationally, existence of a fire causing smoke, a fire, a fire-growth and spread.

What is claimed is: 
     
         1 . A system comprising a server executing via an edge computing and cloud-computing platform comprising a communication device, a database, a memory, and a processor coupled to the memory, wherein the processor is configured to:
 receive periodically, a first image, ambient weather-related information, land related information and data on past fire history of a region of interest of a geographical region;   map automatically, using a deep-neural network that includes CNNs and RNNs, the region of interest into a plurality of risk areas using the first image;   receive periodically, a second image of the plurality of risk areas;   extract, from the second image, using a convolutional neural network, a convolutional output comprising at least one of a smoke, a flame, a spark, and an ember; and   recognize automatically, from a geolocation of the second image, using the deep-neural network, an existence of a fire by identifying a smoke signal or a fire related signal and the ambient weather-related information.   
     
     
         2 . The system of  claim 1 , wherein the plurality of risk areas comprise an arid vegetation land, a semi-arid vegetation land, an arid vegetation land with active human interference and an arid vegetation land with dry lightning. 
     
     
         3 . The system of  claim 1 , wherein the smoke signal or the fire related signal comprise a smoke, a flame, a spark, an ember, a rapid rise in surface temperature, increase in surface heat and increase in CO 2  level. 
     
     
         4 . The system of  claim 1 , is further operable to:
 annotate the second image using a visual combustible object tagging tool to specify bounding boxes around edges of the smoke signal or the fire related signal;   process, the second image, through a plurality of filters to create map edges of the smoke signal or the fire related signal;   perform combustible object recognition, within given scenes by slicing and indexing the second image using a N-dimensional array mapping method; and   identify, by time-correlation analysis, cause, precise geolocation, and timestamp of start of the fire using a RNN model.   
     
     
         5 . The system of  claim 4 , wherein the plurality of filters comprise a horizontal line filter, a vertical line filter, and a diagonal line filter. 
     
     
         6 . The system of  claim 1 , is further operable to:
 generate, with a non-linear regression model operably coupled to the deep-neural network, a time-series pattern representation of the smoke signal or the fire related signal;   classify, with a classifier operably coupled to the deep-neural network, the smoke signal, or the fire related signal based on the convolutional output; and   predict fire-growth, and fire-spread by time series correlation, of the classified smoke signal or the classified fire related signal, the time-series pattern representation, the ambient weather-related information, and the land related information.   
     
     
         7 . The system of  claim 6 , wherein the system is further operable to provide information for situational awareness and decision making based on the fire-growth and the fire-spread prediction. 
     
     
         8 . The system of  claim 1 , wherein the plurality of risk areas comprise an arid vegetation land, a semi-arid vegetation land, an arid vegetation land with active human interference and an arid vegetation land with dry lightning. 
     
     
         9 . The system of  claim 1 , wherein the cloud-computing platform executes information processed via a fog network and an edge device. 
     
     
         10 . The system of  claim 1 , wherein the first image, and the second image comprise still-images, videos, and infrared images. 
     
     
         11 . The system of  claim 1 , wherein the system communicates via a high-speed broadband internet network comprising 5G network and a next generation Wi-Fi network. 
     
     
         12 . The system of  claim 1 , wherein the system is further operable to generate and send a warning signal on possible start of the fire. 
     
     
         13 . A system comprising:
 a self-steering unmanned aerial device, comprising a first processor, a first memory, a first communication device, a first database, and a server executing via a cloud-computing platform that comprises a second processor comprising a graphical processing unit (GPU) and a central processing unit (CPU), a second memory, a second communication device, a second database, wherein the system is operable to:   obtain periodically, a first image, of a region of interest, of a geographical region, using the self-steering unmanned aerial device;   obtain, ambient weather-related information of the region of interest;   obtain, land related information, wherein the land related information comprise vegetation features, topography, elevation, slope, and aspect for the geographical region;   obtain data on past fire history of the region of interest;   map automatically, using a deep-neural network, a plurality of risk areas, of the region of interest, from the first image of the region of interest;   obtain periodically, a second image of the plurality of risk areas, mapped from the first image, of the region of interest, using the self-steering unmanned aerial device;   extract, from the second image, using a convolutional neural network, a convolutional output comprising at least one of a smoke, a flame, a spark, and an ember; and   recognize automatically, from a geolocation of the second image, using the deep-neural network, an existence of fire by identifying a smoke signal or a fire related signal, and the ambient weather-related information; and   predicting automatically, existence of a fire, a fire-growth and spread based on the smoke signal or the fire related signal, the ambient weather-related information of the region of interest and the land related information.   
     
     
         14 . The system of  claim 13 , wherein the system is operable to:
 generate and send a warning signal on possible start of the fire; and   provide information for situational awareness and decision making based on the fire-growth and fire-spread prediction.   
     
     
         15 . The system of  claim 13 , wherein the self-steering unmanned aerial device further comprises a geolocation device and plurality of sensors comprising a high-definition camera, an audio sensor, a heat sensor, a temperature sensor, a wind speed sensor, a smell sensor, a smoke sensor, a CO 2  sensor, a wind direction sensor, a humidity sensor, an atmospheric pressure sensor, a solar radiation sensor and a lightning detector. 
     
     
         16 . The system of  claim 13 , wherein the self-steering unmanned aerial device is controlled by a ground-based controller comprising an Internet of Things (IoT) based device management system. 
     
     
         17 . The system of  claim 13 , wherein the cloud-computing platform executes information processed via a fog network and an edge device. 
     
     
         18 . A non-transitory computer storage medium storing a sequence of instructions, which when executed by a processor, causes:
 receiving periodically, a first image, ambient weather-related information, land related information and data on past fire history of a region of interest of a geographical region, wherein the land related information comprise vegetation features, topography, elevation, slope, and aspect for the geographical region;   receiving data on past fire history of the region of interest;   mapping automatically, using a deep-neural network, the region of interest into a plurality of risk areas using the first image, wherein the plurality of risk areas comprise an arid vegetation land, a semi-arid vegetation land, an arid vegetation land with active human interference and an arid vegetation land with dry lightning;   receiving periodically, a second image of the plurality of risk areas;   extracting, from the second image, using a convolutional neural network, a convolutional output comprising at least one of a smoke, a flame, a spark, and an ember;   recognizing automatically, from a geolocation of the second image, using the deep-neural network, an existence of fire by identifying a smoke signal or a fire related signal and the ambient weather-related information; and   predicting automatically, existence of a fire, a fire-growth and spread based on the smoke signal or the fire related signal, the ambient weather-related information of the region of interest and the land related information.   
     
     
         19 . The non-transitory computer storage medium of  claim 18 , wherein the mapping, the recognizing and the predicting is performed via a cloud-computing platform using information processed via a fog network and an edge device. 
     
     
         20 . The non-transitory computer storage medium of  claim 18 , wherein the sequence of instructions comprises machine-learning algorithms further comprising an evolutionary algorithm.