Real-time detection method for falling of pedestrian in complex environment

Abstract

The invention discloses a real-time detection method for falling of a pedestrian in a complex environment, and relates to a detection method for pedestrian state monitoring. The method comprises the following steps: preprocessing a collected video, converting each frame of a video stream into a picture, and performing normalization processing; performing pedestrian detection: adjusting the size ofa detection frame according to the change of the distance between a pedestrian and a camera; performing target tracking, namely performing target tracking, feature extraction and similarity calculation by adopting a sort algorithm; predicting the current position through a Kalman filter, and associating the detection box with the target position through a Hungary algorithm; performing tumble judgment: judging whether the pedestrian in the target area tumbles or not, wherein when the pedestrian stands, the length-width ratio of the pedestrian is smaller than or equal to 0.4, and at the momentwhen the pedestrian tumbles, the length-width ratio is increased to 0.7-1.2. The invention aims to solve some problems existing in a pedestrian falling detection and judgment task, and constructs a pedestrian falling real-time detection method which is simple and clear in structure, diverse in application scene, high in precision and high in robustness.

Description

technical field [0001] The invention relates to a detection method for pedestrian state monitoring, in particular to the technical field of real-time detection of pedestrian falls in complex environments. Background technique [0002] In today's society, the probability of accidental death due to falls of the elderly is increasing, which has aroused widespread concern, and a lot of research work has been carried out on the fall detection of the elderly. At present, the existing automatic pedestrian fall detection systems are mainly divided into three categories: scene-based equipment, wearable device-based and computer vision-based automatic fall detection systems. [0003] The automatic fall detection system based on scene equipment analyzes and judges the data of human body movement collected by multiple sensor devices in a specific scene. Although it can not affect normal life and has a high accuracy rate, it is difficult due to the high cost of such equipment. be promot...

AI Technical Summary

Problems solved by technology

[0005] However, there are still some defects and deficiencies in the fall detection algorithm based on computer vision, including: 1. The traditional pedestrian detection and tracking method based on computer vision mainly uses artificial feature values ​​for global feature detection, and these features are easily affected by light changes. , scene change, occluder occlusion and other complex environments, the robustness of the detection and tracking algorithm is poor, and the ideal effect cannot be obtained

2. The target tracking algorithm using correlation filtering can improve the robustness of moving object detection in complex environments, but this type of algorithm will cause tracking drift problems due to the scale transformation of the target object itself

Images