Body joint distributed information consistency estimation method based on interacting multiple models

Abstract

The invention discloses a body joint distributed information consistency estimation method based on interacting multiple models. The method comprises the following steps: initializing a skeleton joint position; estimating joint motion by use of a local sensor; constructing a motion model and an observation model by virtue of a human body joint, and realizing effective estimation on a joint state; estimating information consistency of a target joint between the sensors, i.e., each sensor transmits a joint information vector and an information matrix which are estimated by the sensor and corresponding information contribution and model probability to an adjacent communication sensor node and receives information of the surrounding sensors; by virtue of utilizing an information consistency algorithm and fusing the estimation results of the surrounding sensors, subsequent iteration is performed for multiple times, and convergence of the algorithm and the estimation result is realized; by virtue of constructing a distributed RGBD sensor network, and utilizing the information consistency algorithm, distributed estimation of human body joints is realized; no data fusion center exists in a network, so that robustness of a system for node information mistake and invalidity is improved, and expansion of the sensor network is relatively easily realized.

Description

technical field [0001] The invention relates to the field of information technology, in particular to a method for estimating the consistency of distributed information of joint points of a human body based on interactive multi-models. Background technique [0002] Human behavior recognition based on multiple RGBD cameras has attracted extensive attention from researchers, and has been applied to human behavior detection in operating rooms, factory workshops, automobile assembly, indoor monitoring and other environments, effectively solving the problem of human occlusion and possible human- The robot collision problem has important application value. [0003] At present, human behavior perception based on multiple RGBD sensors is still in a centralized stage, requiring one or more data fusion centers to fuse 3D data and human skeleton joint point data, which requires high computing power and robustness for data fusion centers , weak resistance to network instability and low...

AI Technical Summary

Problems solved by technology

[0003] At present, human behavior perception based on multiple RGBD sensors is still in a centralized stage, requiring one or more data fusion centers to fuse 3D data and human skeleton joint point data, which requires high computing power and robustness for data fusion centers , Weak resistance to network instability and low scalability

[0004] Among them, the related technology for the consensus estimation of the distributed information of the human body joints with interactive multi-models has not yet appeared.

Images