基于SIFT-GMLBP的机器人视觉图像特征提取算法

为了改善环境变化较大时机器人在对图像特征提取效果欠佳的问题,对局部二值模式(LBP)进行了改进,根据图像中心像素点邻域之间的相互关系划分网格进行编码,提出了SIFT-MLBP相结合的图像特征提取算法。使用SIFT算法得到图像特征的关键点后,以区域中每个像素点为中心构建网格化结构,计算之间的相邻象素的局部差异,并对对比度不同的像素编码分配权重。结合Gabor变换对基于模式的特征向量进行提取,建立SIFT-GMLBP特征向量,采用原补码互相映射的方式降低特征向量维数。实验证明,SIFT-GMLBP算法具有良好的特征匹配效果,匹配正确率达到95%以上,运行时间降低0.05S。该方法对外部环境的变化具有较强的鲁棒性,能够提高移动机器人在复杂环境中对图像识别的速度和精度。

Visual Image Feature Extraction Algorithm forMoving Robot Base on SIFT-GMLBP

In order to improve problems of poor image feature extraction in larger environmental changes, the LBP is improved. Encoded by meshing according to the relationship between neighborhoods of the center pixel and the image feature extraction algorithm of SIFT-MLBP is proposed. In each pixel as the center to construct the grid structure and calculate local differences between the adjacent pixels after get the key point of the image features by u_sing SIFT algorithm, and then assign weights to the pixel coding of different contrast. The feature vectors based on the mode are extracted and establish SIFT-GMLBP feature vector by combining Gabor transform. Reduce the dimension of feature vectors using the original code and complement mapping. Experimental results show that the proposed algorithm has a good matching result on Visual Image Feature Extraction, the recognition time is shortening to 0. 05S, and the recognition rate is improved to more than 95%. It is validated that the algorithm is strongly robust to the environment change, and is able to meet the requirements of Speed and accuracy in the Image recognition for mobile robots.