Portable 360-degree video-based fire and smoke detector and wireless alerting system
AssigneeFactory Mutual Insurance Company
InventorsJaap De Vries, Stanislav Kostka, Raphael Hussni Leonessa, Kyle Richard Burchesky
Life of this patent
Estimating the term…
Abstract
An aspect of the invention includes a fire and smoke detection system including a 360-degree camera, and a computing platform in communication with the 360-degree camera. The computing platform is configured to receive an image or video from the 360-degree camera, determine that tire or smoke is detected in the image or video received from the 360-degree camera, and venerate at least one alert in response to determining that fire or smoke is detected in the image or video received from the 360-degree camera. The fire and smoke detection system may be portable and not removably mounted.
Claims14
A fire and smoke detection system comprising:
- a 360-degree camera;
- and
- a computing platform in communication with the 360-degree camera;
- wherein the computing platform is configured to:
- receive an image or video from the 360-degree camera;
- determine that fire or smoke is detected in the image or video received from the 360-degree camera, wherein determining that fire or smoke is detected comprises using a convolutional neural network to determine that fire or smoke is detected in the image or video received from the 360-degree camera;
- and
- generate at least one alert in response to determining that fire or smoke is detected in the image or video received from the 360-degree camera,
- wherein the fire and smoke detection system is portable and not removably mounted, and
- wherein the convolutional neural network is a distortion-aware convolutional neural network trained using video or image data collected via fire testing using smoke or fire events selected from the group consisting of:
- rack storage fires, pool fires, wall fires, and smoke generators.
A method to detect fire and smoke using a portable system, the portable system comprising, a computing platform and a 360-degree camera in communication with the computing platform, the computing platform comprising one or more microprocessors and memory, the computing platform using the one or more microprocessors to execute software code stored in the memory, the method comprising:
- capturing an image or video by the 360-degree camera;
- transforming the image or video to an equirectangular image or equirectangular video;
- processing the equirectangular image or equirectangular video by the computing platform using convolutional neural network software code to determine whether fire appears in the equirectangular image or equirectangular video;
- processing the equirectangular image or equirectangular video by the computing platform using the convolutional neural network software code to determine whether smoke appears in the equirectangular image or equirectangular video;
- and
- generating at least one alert in response to determining that fire or smoke is in the image or video received from the 360-degree camera;
- wherein the convolutional neural network software code comprises a distortion-aware convolutional neural network trained using video or image data collected via fire testing using smoke or fire events selected from the group consisting of:
- rack storage fires, pool fires, wall fires, and smoke generators.
Dependencies are inferred from the claim text.