基于机器学习的人体动作深度信息识别方法研究

Research on a Recognition Method for Depth Information of Human Actions Based on Machine Learning

提出了基于机器学习的人体动作深度信息识别方法,构建人体动作的三维图像采集模型,建立人体动作三维重建图像的表面结构重构模型。结合模糊度特征提取方法对人体动作三维重建图像进行多尺度分解,采用三维空间结构重组的方法进行细节特征识别,建立图像的多维分割模型。采用机器学习算法进行细节特征分类识别,建立人体动作深度信息的提取和分类模型,在机器算法下实现人体动作的深度信息检测和多维识别。仿真结果表明,该方法准确度较高,特征分辨力较好,具有很好的人体动作信息检测和辨识能力。

A recognition method for depth information of human actions based on machine learning is presented,a 3D image acquisition model of human actions is constructed and the surface structure reconstruction model of 3D reconstruction image of human actions is established.Multi-scale decomposition on the 3D reconstruction image of human actions is carried out by combining with ambiguity feature extraction method,the detailed feature recognition on human actions is conducted by using the spatial structure reconstruction method and the multi-dimensional segmentation model of human action images is established.The classification and recognition on the detail features of human actions are carried out by using the machine learning algorithm,the extraction and classification models of depth information of human actions are established,and the depth information detection and multi-dimensional recognition of human actions are realized with machine algorithm.The simulation results show that the method has high accuracy,good feature resolution and good ability to detect and identify human action information.