Quaternion time-space convolutional neural network-based human body behavior identification method

Abstract

The invention belongs to the field of computer vision and particularly relates to a quaternion time-space convolutional neural network-based human body behavior identification method. The method comprises the following specific implementation steps of (1) inputting a to-be-identified action video set; (2) performing image preprocessing, and extracting a key region image of a human body motion; (3) constructing a quaternion time-space convolutional neural network; (4) training the network by adopting a BP algorithm, and outputting a training result; and (5) inputting a video test set, and outputting a test result. According to the method, the region image of the human body motion is extracted by utilizing a codebook model, so that the human body motion can be detected under the condition of a complex background; and the quaternion time-space convolutional neural network directly takes a color image as an input, so that the problem of image feature deficiency in a process of converting the color image into a grayscale image or performing processing by channels in a conventional convolutional neural network is solved, the performance of network feature extraction is improved, and the human body behavior identification is more accurate.

Description

technical field [0001] The invention belongs to the field of computer vision, and further relates to a human behavior recognition method based on a quaternion space-time convolutional neural network in target recognition. The invention can be used in applications such as human-computer interaction and intelligent monitoring. Background technique [0002] Human behavior recognition is an important research direction of computer vision, pattern recognition, image processing and artificial intelligence. It has great application value and theoretical significance in the fields of human-computer interaction, intelligent monitoring and medical treatment. It mainly analyzes and processes moving image sequences containing people, extracts features, and classifies motions to realize the recognition and understanding of individual human actions, interactions between people and between people and the external environment. [0003] Compared with object classification based on still imag...

AI Technical Summary

Problems solved by technology

However, the current convolutional neural network model only processes grayscale image sequences or RGB three-channels separately, without considering the correlation and integrity of the three channels, which destroys the color characteristics of the real environment and affects the recognition method in the actual environment. robustness

Images