基于机器学习的人体动作深度信息识别方法研究

Research on Recognition Method of Human Action Depth Information Based on Machine Learning

为了实现人体动作的准确检测识别,提出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人体动作细节特征识别,建立人体动作图像的多维分割模型;采用机器学习算法进行人体动作的细节特征分类识别,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,采用该方法进行人体动作深度信息识别的准确度较高,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。

In order to realize the accurate detection and recognition of human action, a recognition method of human action depth information based on machine learning is proposed, the 3 D image acquisition model of human action is constructed, the surface structure reconstruction model of 3 D reconstruction image of human action is constructed, the 3 D reconstruction image of human action is decomposed by multi-scale decomposition combined with ambiguity feature extraction method, and the detail feature recognition of human action is carried out by using the method of 3 D spatial structure reconstruction. The multi-dimensional segmentation model of human action image is established, the machine learning algorithm is used to classify and recognize the detail features of human action, the extraction and classification model of human action depth information is established, and the depth information detection and multi-dimensional recognition of human action are realized under 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.