基于Zernike矩和ELM的舞蹈视频图像中人体动作识别研究

Research on human motion recognition in dance video images based on zernike moment and ELM

为了有效地识别舞蹈动作进而对舞者的舞蹈动作进行纠正,提出了一种基于Zernike矩和ELM算法的舞蹈视频图像中人体动作识别方法。首先,通过对图像进行预处理,将图像转化为二值化图像便于特征分析;其次,通过三维Zernike矩对预处理后的图像进行特征提取,建立码书;再次,根据码书提供的相关性信息,通过ELM算法对舞蹈动作进行学习,输出分类信息;最后,分别对NADA-KTH数据库中的静态动作图像和weizmann数据库提供的动态芭蕾舞视频图像进行了实验,分别从静态和动态的角度验证了所提出方法的有效性,初步得出其与识别率的正相关关系。

In order to effectively recognize the dance movement and then correct the dancer's dance movement,a human motion recognition method based on Zernike moment and ELM algorithm is proposed.Firstly,by preprocessing the image,transforming the image into a binarized image facilitates feature analysis.Secondly,feature extraction is performed on the preprocessed image by three-dimensional Zernike moment to establish a codebook;again,according to the correlation provided by the codebook Information,through the ELM algorithm to learn the dance movements,output classification information;Finally,the static action images in the NADA-KTH database and the dynamic ballet video images provided by the weizmann database were tested separately,from static and dynamic perspectives.The effectiveness of the proposed method is investigated.The positive correlation between the proposed method and the recognition rate is obtained.