Depth image human body joint positioning method based on convolution nerve network

Abstract

The invention discloses a depth image human body joint positioning method based on a convolution nerve network. The method is characterized by comprising a training process and an identification process. The training process comprises the following steps: 1, inputting a training sample; 2, initializing a deep convolution nerve network and its parameters, wherein the parameters comprise a weight and a bias of each layer edge; and 3, by use of a forward algorithm and a backward algorithm, learning the parameters of the constructed convolution nerve network. The identification process comprises the following steps: 4, inputting a test sample; and 5, performing regression on the input test sample by use of the trained convolution nerve network to find positions of human body joints. According to the invention, by use of the deep convolution nerve network and large data, multiple challenges such as shielding, noise and the like can be resisted, and the accuracy is quite high; and at the same time, by means of parallel calculation, the effect of accurately positioning the human body joints in real time can be realized.

Description

technical field [0001] The invention relates to the fields of computer vision, pattern recognition and human-computer interaction, in particular to a convolutional neural network-based deep image human joint positioning method. Background technique [0002] Body pose estimation and motion capture is an important research direction in the field of computer vision. Its applications include home entertainment, human-computer interaction, motion recognition, security systems, remote monitoring, intelligent monitoring, and even patient health care. However, human pose estimation in ordinary RGB images or videos is a very challenging task. Because it cannot be robust to natural environmental factors such as color, lighting, and occlusion, coupled with too many degrees of freedom in human posture and different observation angles, this problem is naturally even more difficult. [0003] The depth image is a two-dimensional grayscale image, but unlike the traditional grayscale image...

AI Technical Summary

Problems solved by technology

[0006] 1) The depth image has the defects of low resolution and large mechanical noise

Making hand-designed features to locate human joints cannot achieve better results

[0007] 2) The positioning of human body joints is very difficult to achieve accurate and robust positioning due to the different camera placement angles, different distances between the camera and the characters, and different occlusion degrees of the characters themselves.

[0009] 4) It is difficult to integrate the positioning and tracking of the bones and joints of the human body

At present, the position and posture positioning of characters are all based on a single depth image, because it is difficult to express the motion consistency of bones and joints in the time domain

[0010] The above difficulties make the goal of accurate and robust human joint positioning still have a certain gap. Therefore, it is necessary to solve the above difficulties

Images