一种用于工件识别的LBP-HOG特征融合方法  

A method on work piece image recognition based on LBP-HOG feature fusion

作　　者：吴益红[1] 许钢[1] 江娟娟[1] 毕运锋 

机构地区：[1]安徽工程大学安徽检测技术与节能装置省级实验室,安徽芜湖241000

出　　处：《南阳师范学院学报》2016年第9期33-38,共6页Journal of Nanyang Normal University

基　　金：国家自然科学基金(61271377);安徽工程大学优秀青年人才基金一般项目(2013RZR009)

摘　　要：针对目前机器人在工业现场对工件目标识别过程中出现的低识别率问题,提出一种基于GA寻优的LBP-HOG特征融合方法,结合SVM对工件图像进行分类识别.首先,分别运用了基本LBP算子、LBP等价模式以及LBP旋转不变模式,结合不同去噪方式,评价并选择最优的LBP算子,然后,利用LBP和HOG算子分别提取工件特征,并以组合向量方式融合两类特征.最后,利用GA寻优两类特征的融合权重,通过反复评价SVM对融合特征的识别效果,更新权重,最终得到最优权重和最优识别精度.实验结果表明,单独使用LBP和HOG的图像识别率分别为80%和84%,而GA寻优后的LBPHOG组合模型,准确率提高到了96%.Aiming at the problem of the low recognition rate at the industrial robot recognizing work piece target m the industrial field, this paper proposes a combined method named LBP-HOG based on GA intelligent optimization algorithm, combined with SVM to extract and recognize the features of the workpiece images. Firstly, basic LBP pattern, LBP uniform pattern and LBP rotation invariant pattern are used combining with various noise attenuation and the best LBP method is selected. Secondly, LBP and HOG are used to get the features of the work pieces separately, and then the features are fused in the way of combined vector approach. Finally, the best classification results and the optimal weight are acquired by evaluating classification effect, improving the classifi- cation accuracy through repeating the SVM classification model and updating the weight. The experimental results indicate that image recognition accuracy on GA optimization of LBP -HOG combination model is up to 96%, a good improved results is better than LBP and HOG image recognition rate of 80% and 84% respectively.

关 键 词：局部二值模式 方向梯度直方图 特征融合 支持向量机 智能寻优 工件分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]