A fall detection method based on video joints and hybrid classifier

Abstract

The invention discloses a fall detection method based on video articulation points and a hybrid classifier. Traditional video-based fall detection algorithms rely on manual extraction of fall features, and use linear discriminant classifiers to detect falls. The model is simple but the accuracy is low. The present invention is as follows: 1. Extract each frame image of the detected video segment. 2. Obtain human joint data matrix; 3. Establish multiple behavioral matrices. 4. Calculating time characteristic parameters and spatial characteristic parameters. 5. Primary classification; 6. Secondary classification. The method for extracting joint points of human skeletons adopted in the present invention solves the problem that the traditional method of extracting the aspect ratio and projected area of ​​the human body cannot accurately estimate the posture of the human body. The invention adopts a fixed-size sliding window to construct a behavior matrix, which can simultaneously model the time and space axes, and fully express the characteristics of the falling behavior.

Description

technical field [0001] The invention belongs to the technical field of fall detection, and in particular relates to a fall detection method based on a video joint point and a hybrid classifier. Background technique [0002] Scholars at home and abroad have done a lot of research on falls of the elderly. There are three main fall detection methods: wearable sensor-based fall detection, environmental sensor-based fall detection, and video image-based detection. Fall detection based on wearable sensors mainly detects falls by collecting data collected by wearable sensors and setting thresholds to detect falls. Inaccurate threshold settings will affect the final detection results. The fall detection based on the environmental sensor mainly judges the fall through the data obtained by the pressure sensor on the ground or the sound detection equipment. If the environmental noise is too large, the data may be abnormal. Video-based fall detection mainly uses ordinary cameras or dep...

AI Technical Summary

Problems solved by technology

However, the traditional video-based fall detection algorithm relies on manually extracting fall features, and detects falls with the help of a linear discriminant classifier. The model is simple but the accuracy is low. The existing deep learning-based detection algorithm model is complex, and it is difficult to reduce the time-consuming detection.

Images