On-line sequential extreme learning machine-based incremental human behavior recognition method

Abstract

The invention discloses an on-line sequential extreme learning machine-based incremental human behavior recognition method. According to the method, a human body can be captured by a video camera on the basis of an activity range of everyone. The method comprises the following steps of: (1) extracting a spatio-temporal interest point in a video by adopting a third-dimensional (3D) Harris corner point detector; (2) calculating a descriptor of the detected spatio-temporal interest point by utilizing a 3D SIFT descriptor; (3) generating a video dictionary by adopting a K-means clustering algorithm, and establishing a bag-of-words model of a video image; (4) training an on-line sequential extreme learning machine classifier by using the obtained bag-of-words model of the video image; and (5) performing human behavior recognition by utilizing the on-line sequential extreme learning machine classifier, and performing on-line learning. According to the method, an accurate human behavior recognition result can be obtained within a short training time under the condition of a few training samples, and the method is insensitive to environmental scenario changes, environmental lighting changes, detection object changes and human form changes to a certain extent.

Description

technical field [0001] The invention relates to a method for recognizing human body behavior by using machine vision, and belongs to the technical field of pattern recognition. Background technique [0002] Video-based human behavior recognition is widely used in robotics, human-computer interaction, video-based intelligent monitoring, motion analysis, content-based video retrieval and other fields. It is a research hotspot in computer vision and has broad application prospects and potential. Because of its economic and social value, it has been highly concerned by the majority of scientific research workers and related businesses. [0003] In terms of robotics, the study of human gait characteristics provides a theoretical basis for the gait planning of biped robots; video-based intelligent monitoring plays an inestimable role in building a harmonious society, maintaining social stability, and criminal technology detection. Understanding human behavior is the key; in terms...

AI Technical Summary

Problems solved by technology

[0004] However, due to the complex and changeable real natural environment (such as complex background, environmental lighting changes), individual differences of people (height, shape, clothes, etc.), different viewing angles for obtaining video images, and different ways and speeds for people to complete a certain action , making video-based human action recognition a very challenging problem

"An Efficient Dense and Scale-Invariant Spatio-Temporal Interest Point Detector" (an effective high-density and scale Invariant spatio-temporal interest point detector) Apply Kmeans clustering algorithm to cluster motion feature vectors extracted from video sets to construct visual vocabulary, use support vector machine (SVM) classifier to classify and recognize human behavior, the establishment of SVM classifier needs Large number of training samples and long training time

The classification model trained by the offline classifier is only suitable for a specific environment or a specific person. Due to the characteristics of the offline classifier training, the detected object needs to be consistent with the training object, so when the size and posture of the person in the video image, the environment When it is seriously inconsistent with the training samples, accurate behavior recognition cannot be performed, and the portability is not good

In addition, offline classifiers generally learn the classification samples one by one, the learning speed is slow, and the real-time performance is not good.

Images