Anonymized video analysis methods and systems

摘要

Methods and systems for anonymized video analysis are described. In one embodiment, a first silhouette image of a person in a living unit may be accessed. The first silhouette image may be based on a first video signal recorded by a first video camera. A second silhouette image of the person in the living unit may be accessed. The second silhouette image may be of a different view of the person than the first silhouette image. The second silhouette image may be based on a second video signal recorded by a second video camera. A three-dimensional model of the person in voxel space may be generated based on the first silhouette image, the second silhouette image, and viewing conditions of the first video camera and the second video camera. In some embodiments, information on falls, gait parameters, and other movements of the person living unit are determined. Additional methods and systems are disclosed.

主权项

1. A method comprising:
accessing a first silhouette image of a person in a living unit, the first silhouette image based on a first video signal recorded by a first video camera; accessing a second silhouette image of the person in the living unit, the second silhouette image being of a different view of the person than the first silhouette image, the second silhouette image based on a second video signal recorded by a second video camera, the second video camera being a different video camera than the first video camera, the first video camera and the second video camera being positioned to record an area of the living unit from a different position in the living unit; generating a three-dimensional model of the person in voxel space based on the first silhouette image, the second silhouette image, and viewing conditions of the first video camera and the second video camera; accessing a third silhouette image of the person in the living unit, the third silhouette image based on the first video signal recorded by the first video camera, the third silhouette image associated with a different frame of the first video signal than the first silhouette image; accessing a fourth silhouette image of the person in the living unit, the fourth silhouette image based on the second video signal recorded by the second video camera, the fourth silhouette image associated with a different frame of the second video signal than the second silhouette image; analyzing a projection of a bottom section of the three-dimensional model of the person onto a ground plane of the living unit during the time period; identifying a plurality of voxel intersections during a time period, a particular voxel intersection including an overlapping portion of the bottom section of the three-dimensional model of the person onto the ground plane on multiple frames, the third silhouette image being after the first silhouette image, the fourth silhouette image being after the second silhouette image;
calculating a voxel volume of the three-dimensional model of the person; applying fuzzy set theory to the voxel volume of the three-dimensional model of the person based on an expected voxel volume of the person; and generating a fuzzified version of the three-dimensional model of the person in the voxel space based on the first silhouette image, the second silhouette image, the viewing conditions of the first video camera and the second video camera, and application of fuzzy set theory.