Human skeleton tracking system and method based on multiple Kinects

Abstract

The present invention provides a human skeleton tracking system and method based on multiple Kinects. The system comprises: a first terminal, where the first terminal comprises multiple Kinects, and is configured to collect human body images and change the human body images into point cloud data; a data transmission module; a second terminal, wherein the second terminal comprises a data receivingmodule and is configured to receive the point cloud data; a data conversion module, wherein the data conversion module is configured to unify a camera coordinate system of multiple Kinects and a shotmark point coordinate system to obtain an initial transformation relationship of each two Kinects; a data fusion module, wherein initial pairing is performed on each two Kinects to obtain multiple matching methods for one Kinect, the optimal pairing of each Kinect is obtained through the least squares iteration, and the point cloud data of multiple Kinects are fused according to the optimal transformation relationship of each two Kinects; and a skeleton extraction module, wherein human reconstruction is performed according to the point cloud data fused by the data fusion module. According to the system and method provided by the present invention, the misalignment of point cloud data registration can be reduced.

Description

technical field [0001] The present invention relates to the field of vision technology, and more specifically, to a multi-Kinect-based human skeleton tracking system and method. Background technique [0002] With the vigorous development of human-computer interaction technology, somatosensory devices have been widely used in shopping malls, entertainment, medical and other fields. A single Kinect has a small interactive range and limited accuracy. In order to meet the greater needs of actual users and researchers, existing somatosensory devices still have many deficiencies, such as: the collection distance of a single device is short, the range of human-computer interaction is small, the collection accuracy of the device itself, and Problems such as user self-occlusion and external noise. [0003] The current somatosensory devices mainly have problems such as limited interaction range, user self-occlusion and insufficient interaction accuracy. In order to solve these prob...

AI Technical Summary

Problems solved by technology

[0005] In order to increase the accuracy of the skeleton data in the existing multi-Kinect human-computer interaction technology, the geometric triangulation method is used to accurately place the specific positions and directions of the three Kinect cameras, so as to solve the problem of the lack of accuracy of a single Kinect and the For problems such as self-occlusion, the position and direction of each Kinect must be accurately placed in terms of physical geometry. Such a scene is difficult to build, and their position information must be determined again for reuse.

In addition, the existing multi-Kinect human-computer interaction technology has not dealt with the possible external interference and noise

Furthermore, even if the existing multi-Kinect human-computer interaction technology is placed accurately, there will still be a large number of dislocations in the registration of point cloud data, and the registration accuracy will be low.

Images