基于D-S证据理论的信息融合图像识别

Image information fusion recognition based on D-S evidence theory

针对图像识别的不确定问题,提出一种基于Dempster-Shafer(D-S)证据理论的信息融合图像识别算法.用灰度-相位共生矩阵和灰度-梯度共生矩阵提取图像的纹理特征参数,对纹理特征参数进行转化得到待识别图像在其他类图像上的信度函数分配;利用D-S联合规则得到融合后的信度函数分配,从而准确识别图像;通过单一矩阵图像识别结果与融合识别结果比较,说明D-S数据融合在识别图像方面的优越性.

In view of the uncertainty of image recognition, an information fusion image recognition algorithm is presented based on the Dempster-Shafer （D-S） evidence theory. The image texture feature parameters are extracted by gray level phase co-occurrence matrix and gray level-gradient co-occurrence matrix respectively. The texture feature parameters are converted to obtain the belief function assignment of the image to be recognized on the other types of image. The fusion belief function assignment can be achieved by using D-S joint rules to identify the image accurately. The superiority of the D-S data fusion in image recognition is illustrated by comparison of the single matrix image recognition with fusion recognition results.