Fall detection method and device

Abstract

The invention discloses a fall detection method and device, and the method comprises the following steps: receiving a left lens image and a right lens image shot by a binocular camera, and enabling the left lens image and the right lens image to have a target person; extracting 2D human body skeletons of the target person in the left lens picture and the right lens picture, wherein the 2D human body skeletons comprise sole key points and upper body key points; constructing a 3D human body skeleton of the target person based on the 2D human body skeleton of the target person in the left lens picture and the right lens picture and camera parameters calibrated in advance; and calculating a relative relationship between the key points of the upper body in the 3D human body skeleton of the target person and the ground plane in the to-be-detected area in real time, and judging whether falling occurs according to a calculation result. The 3D skeleton and the ground plane parameters of the person are constructed, whether the key point touches the ground or not is judged by calculating the distance between the key point of the upper body and the ground plane, the method is stable and efficient, and the extremely high falling detection rate is achieved.

Description

technical field [0001] The invention relates to the technical fields of computer science and deep learning, in particular to a fall detection method and device. Background technique [0002] On May 11, 2021, the results of the seventh national census showed that China’s population aged 60 and over accounted for more than 18%, and the degree of population aging is further deepening. As the age of the elderly increases, the physical function of the elderly gradually declines, and their health The risks will gradually increase. Especially quite a lot of old people live alone at home. For the old people living alone, if they fall down, if they cannot be found in time and take corresponding rescue measures, it may often cause serious physical injuries such as fractures, hemorrhage, nerve damage, and paralysis. If the fall behavior can be detected in time when the old man falls, the old man can be treated effectively in the first time, and the serious injury caused by the old man...

AI Technical Summary

Problems solved by technology

[0003] The current main methods are basically divided into two categories: 1. Human skeleton key point detection using monocular (single camera) deep learning, which can generate a 2D human skeleton, and then evaluate the posture of the human body or recognize actions, detect falls, etc., but 2D The human skeleton lacks depth information, action recognition will produce ambiguity, and the detection of falls is not accurate enough, especially for fake fall movements; 2. Use wearable sensing devices to perceive multiple directions of the human body in three-dimensional space through sensors Due to poor carrying comfort and other reasons, the elderly just do not carry it when sleeping or taking a bath at night, or forget to carry it due to other reasons, requiring users to wear relevant devices and equipment. , can not really effectively alarm when a fall occurs

Images