A machine vision method for detecting falls in an elderly person

Abstract

The invention provides a machine vision method for detecting the fall of an old person, which comprises the following steps: a camera shoots a moving image of a human body in real time, and the imageis transmitted to a microcomputer in real time; an ICNet model is used to realize image semantic segmentation, and the corresponding label of each pixel in the image is obtained, and the region of a human body in the image is obtained. the human body region segmented by ICNet is judged conditionally; the binary gray image generated in the third step is inputted, and the ellipse contour of human body in the image is quickly generated by using the ellipse fitting algorithm based on the least square method, and the ellipse azimuth angle and the ratio of the major axis to the minor axis are calculated; using a KNN algorithm, the ellipse azimuth and the ratio of major axis to minor axis are used as inputs to classify the current state of the elderly, and it is judged whether the elderly fall ornot. The method is based on the depth learning method to analyze the collected image information, according to the semantic segmentation and KNN binary classification model to monitor the activitiesof the elderly in real time, and effectively detect the fall action.

Description

technical field [0001] The invention relates to a machine vision method for detecting falls of an elderly person. Background technique [0002] In recent years, with the intensification of population aging, the number of falls among the elderly has increased. Falls are one of the biggest risks for older adults living alone. Falls kill at least thousands of older adults each year. According to the survey, falls are the sixth leading cause of death for people over the age of 65 and the second leading cause of death for special populations aged 65-75. Sometimes an accidental fall in the elderly can cause serious injury to the spinal cord and hip area. In this case, elderly people who fall cannot call others for help, and may even remain on the ground for hours after the fall, unable to receive timely and effective treatment. [0003] In order to avoid the occurrence of such incidents in a timely and effective manner, some people use various non-visual sensors, such as accel...

AI Technical Summary

Problems solved by technology

However, the methods based on machine vision mainly have the following difficulties: 1. How to distinguish which objects (people) in the scene are falling and which are not; 2. How to distinguish whether the person on the ground is lying or falling; 3. Since the background is always In the change, how to subtract the background without being affected by the change of lighting and scene, and effectively extract the outline of the human body

Images