Abstract
Abnormal falls in public places have significant safety hazards and can easily lead to serious consequences, such as trampling by people. Vision-driven fall event detection has the huge advantage of being non-invasive. However, in actual scenes, the fall behavior is rich in diversity, resulting in strong instability in detection. Based on the study of the stability of human body dynamics, the article proposes a new model of human posture representation of fall behavior, called the “five-point inverted pendulum model”, and uses an improved two-branch multi-stage convolutional neural network (M-CNN) to extract and construct the inverted pendulum structure of human posture in real-world complex scenes. Furthermore, we consider the continuity of the fall event in time series, use multimedia analytics to observe the time series changes of human inverted pendulum structure, and construct a spatio-temporal evolution map of human posture movement. Finally, based on the integrated results of computer vision and multimedia analytics, we reveal the visual characteristics of the spatio-temporal evolution of human posture under the potentially unstable state, and explore two key features of human fall behavior: motion rotational energy and generalized force of motion. The experimental results in actual scenes show that the method has strong robustness, wide universality, and high detection accuracy.
Highlights
Understanding and recognizing human behavior remains a challenging and important task in the field of computer vision
In order to cope with the two problems above, we study the intrinsic dynamics of human fall behavior from the human body structure
The study consists of four steps, as shown in Introducing two-branch multi-stage Convolutional neural networks (CNN) to extract the Part Affinity Fields of human skeleton structure and establishing a description model of human body instability—five-point inverted pendulum model; Based on the five-point inverted pendulum model, constructing the vectors of human body motion using spatio-temporal information coupling; Using the principle of human body dynamics and analyzing the characteristics of human instability caused by abnormal behavior of falling and its quantitative theory; Realizing the detection and identification of human fall behavior
Summary
Understanding and recognizing human behavior remains a challenging and important task in the field of computer vision. Our research can help staff detect the behaviors in real time, automatically and intelligently in a large amount of video data from the video surveillance system and make correct emergency responses in time It can play a positive role in the intelligent development of public safety. When the view of camera and the human body fall in the same line or direction, it is called the posture of same direction (SD posture), the morphology change in the fall process is not obvious These two problems inevitably lead to a significant decline in the robustness of identifying human abnormal fall behavior in the wild environment. The abnormal fall behavior detection method proposed in this article can dynamically analyze and judge motion continuously, with high precision and strong robustness, which is very suitable for fall recognition in complex scenes.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call