| CPC G06T 7/001 (2013.01) [G06T 2207/30184 (2013.01)] | 20 Claims |
|
1. A computer-implemented method comprising:
obtaining, by a property monitoring system, data indicating that precipitation is forecasted to occur at a particular time at a property;
in response to obtaining the data indicating that the precipitation is forecasted to occur at the particular time, capturing, by a camera of the property monitoring system installed inside the property before the particular time, a first image of a ceiling of the property;
capturing, by the camera, a second image of the ceiling of the property at a predetermined time after the first image is captured, wherein the second image is a different image than the first image;
detecting, by the property monitoring system and using differences between the first image and the second image, a candidate water leak at the property, wherein the detecting comprises:
generating a difference image by subtracting the first image from the second image;
detecting the candidate water leak using the difference image;
detecting, using a machine learning model, a window region in the first image and the second image that corresponds to a window installed on the ceiling of the property;
determining whether the candidate water leak is in the window region in the first image and the second image; and
in response to determining that the candidate water leak at the property is not in the window region in the first image and the second image and detecting the candidate water leak at the property:
determining a likelihood that there is a water leak at a region that is different from the window region at the property using the first image and the second image; and
performing, by the property monitoring system, one or more actions for the property using the likelihood that there is a water leak at the property.
|