不同视角下海量高分辨率视频图像数据挖掘方法

The Data Mining Method of Massive High Resolution Video Images from Different Perspectives

当前高分辨率视频图像数据挖掘方法容易受到外界环境的干扰,提取的视频图像特征不可靠,且不同视角下提取的特征值有很大差异,导致视频图像数据挖掘精度大大降低。为此,提出一种新的不同视角下海量高分辨率视频图像数据挖掘方法,通过Harris角点检测方法对待挖掘高分辨率视频图像数据时空特征进行提取。依据高分辨率视频图像数据时空特征,通过自相关矩阵建立相同事物不同视角下的递归图,将递归图看作一幅图像,通过计算像素点的梯度向量构建递归特征描述符。对相同事物不同视角下的关联性进行挖掘,将具有相同递归图梯度特征的高分辨率视频图像数据汇聚在一起,实现数据挖掘。实验结果表明,所提方法挖掘精度高。

The high resolution video image data mining method is easy to be interfered with the external environment,the video image feature extraction is not reliable,and features different from the perspective of the extracted values are very different,leading to the video image data mining accuracy greatly reduced. To this end,a massive high resolution video image data is different from the perspective of new mining methods,through the Harris corner detection method with high resolution video image data mining spatial feature extraction,based on high resolution video image data through the spatial and temporal characteristics,recursive graph autocorrelation matrix is set up the same thing from different angles of view. The recursive graph as an image,through the gradient vector calculation of pixels to construct recursive feature descriptors of the same things different from the perspective of relevance for mining,high resolution video image data with the same recursive graph gradient feature together to achieve data mining. The experimental results show that the proposed method has high precision.