Method and system for recognizing human body tumble automatically based on stereoscopic vision

Abstract

The invention discloses a method and a system for recognizing human body tumble automatically based on stereoscopic vision. The method comprises the steps of 1, acquiring a depth image and a 3D space image of an action of a recognition object through a stereoscopic vision camera; 2, extracting curvature spatial characteristics of a human body contour from the depth image and the 3D space image; 3, analyzing and establishing a CSS characteristic pattern from the curvature spatial characteristics of the human body contour; 4, establishing a consistent action representation method by using bag-of-words CSS (BoWCSS) models; and 5, classifying all of the BoWCSS models by using extreme learning machine (ELM) so as to be used for differentiating a tumble action and other actions. Correspondingly, the invention further provides a system for recognizing human body tumble automatically based on the stereoscopic vision. The method and the system disclosed by the invention can improve the accuracy of human body tumble recognition under the condition of reducing the camera installation cost and the complexity.

Description

technical field [0001] The invention relates to the technical field of computer vision processing, in particular to an automatic human fall recognition method and system based on stereo vision. Background technique [0002] The existing human fall system based on visual sensor uses one or more video cameras to capture the picture of human body movement, and through the respective image processing algorithms, it verifies whether there is an image with fall characteristics, so as to identify the occurrence of human fall. Existing human fall recognition technologies are usually based on the following schemes: single-camera system and multi-camera system. Single-camera systems typically rely on image-space features extracted from silhouettes, such as bounding box aspect ratios. Multi-camera systems do not rely solely on these abstract features, such as determining the velocity of 3D objects constructed from multiple projections. [0003] However, the above two conventional tec...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to solve the problem existing in the prior art that the implementation technology is limited by camera defects, it is very difficult for a single camera system to extract accurate image features, the installation and configuration of a multi-camera system is very complicated, and with the external environment Changes, such as changes in the intensity of light, have a great impact on these systems, resulting in the problem that these existing systems are difficult to be applied in reality, and provide an automatic human fall recognition method and system based on stereo vision

Images