A multi-stage fall detection method based on video

Abstract

The invention discloses a video-based multi-stage fall detection algorithm, which comprises the following steps: (1) moving target separation: generating a binarized image from an input video frame sequence image through a moving target separation method; (2) image optimization Processing: performing morphological processing and rectangular frame optimization on the binarized image generated in step (1) to generate a target binarized image; (3) image feature extraction: performing feature extraction on the generated target binarized image; (4) Fall judgment: perform a fall judgment algorithm to judge the motion state of the target according to the features of the extracted target binarized image; the fall judgment algorithm includes the following steps: 1) discrimination of occlusion occurrence; 2) discrimination result according to occlusion occurrence , execute the corresponding decision algorithm for falling with occlusion or falling without occlusion. This method can distinguish whether the target is occluded by environmental objects, and use the corresponding fall judgment algorithm to quickly and accurately judge the fall event according to the different occlusion degrees.

Description

technical field [0001] The invention relates to image pattern recognition, in particular to a video-based multi-stage fall detection algorithm. Background technique [0002] Accidental falls are one of the important factors that threaten the healthy life of the elderly. Due to the increasing trend of population aging in recent years, the timely prevention and treatment of accidental falls has attracted the attention of many researchers. At present, there are three main technical solutions in the field of fall recognition, namely wearable sensor devices, environmental sensor devices, and computer vision devices. Among them, the wearable sensor device is to wear the sensor on the tester to collect information during the exercise, and judge whether the exercise is a fall through a certain algorithm; the environmental sensor is to install audio sensors on walls, ceilings and other environments to collect space. The sound wave information of the internal target is used to detect...

AI Technical Summary

Problems solved by technology

This method combines multiple features, so the description of the fall event is more sufficient. However, the threshold method is a fall discrimination algorithm, and its single threshold is not enough to fully adapt to different environments such as whether the target is occluded or not occluded.

In the Chinese invention patent CN103186902A, an adaptive fall detection method is proposed. It does not need to manually set the threshold and continuously updates the threshold from the information collected from the scene, which has better environmental adaptability. However, in terms of feature extraction, its main collection The downward shift of the center of gravity is not enough to fully explain the fall behavior, and for the video fall detection method, the impact of the occlusion of environmental objects on the judgment effect of fall detection is not analyzed

Images