基于综合特征的Bp-adaboost工业仪表图像分类方法  

Industrial instrument image classification based on comprehensive characteristic and Bp-adaboost

作　　者：吴学礼[1,2] 赵萌[1] 孟凡华[2] 刘浩南[1] 

机构地区：[1]河北科技大学电气工程学院,河北石家庄050018 [2]燕山大学电气工程学院,河北秦皇岛066000

出　　处：《河北科技大学学报》2013年第5期451-456,共6页Journal of Hebei University of Science and Technology

基　　金：国家自然科学基金(60974018;61273188);河北省自然科学基金(F2012208075)

摘　　要：针对工业场景图像背景复杂,受多种因素影响、利用单个特征完成工业仪表图像分类不能达到满意效果的问题,提出了一种综合利用图像的颜色和纹理特征,通过Bp_adaboost的方法对工业仪表图像进行分类的方法。首先基于HSV空间进行低阶颜色矩特征提取;然后基于灰度共生矩阵进行纹理特征提取;最后用17维综合特征向量对工业仪表图像进行Bp_adaboost分类学习和测试。实验结果表明,该方法对液位控制系统工业仪表与液位容器设备图像能取得较好的分类结果。The industrial scene images,affected by various factors,always have complex background,so a single feature cannot achieve satisfactory result for classification.In this paper,a method using the image color and texture features through the Bp_adaboost method to classify the industrial instrument image is proposed.It first extracts the low order color moment features in HSV space,and then extracts the texture features based on the gray level co-occurrence matrix.Finally,the17-dimensional integrated feature vector for industrial instrumentation image is adapted to the Bp_adaboost classification learning and testing.Experimental results show that the method can achieve better classification results for the industrial level control system instrumentation equipments and level container images.

关 键 词：颜色矩 灰度共生矩阵 分类器 BP神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]