摘要
针对散料自动装车中超声波料位检测方法不能反映料堆实际形态的缺点,提出了基于图像纹理识别的料位检测方法。首先,对原图像进行同态滤波及二值化预处理,以增强图像纹理并减小纹理特征提取计算量;然后,计算装车图像的共生矩阵纹理特征,并对所计算的特征进行主成分分析降维;最后,利用决策树分类算法对装车料位图像进行分块识别并拟合出直线料位。试验结果表明,所提方法料位识别平均偏差为6.5像素,料位识别率为96%,每帧图像处理时间约0.2s。算法基本满足散料装车料位实时检测的要求。
In order to solve the problem that the ultrasonic level detection cannot reflect the actual level of bulk material entrucking,a new method was presented based on image texture recognition.First,the image was enhanced with the homomorphic filter,and the binarization processing was applied to cut down the amount of calculation of the texture features extracting.Then,the co-occurrence matrix texture features were calculated,and principal component analysis method was used to reduce the dimensionality of the feature space.Finally,the image partitions were recognized by decision tree,and the level was linear fitted by the image segmentation result.Experimental results indicate that the mean error is as 6.5 pixels,and the recognition rate is as 96%.The processing time is as 0.2 seconds per frame.It can satisfy the real-time level detection of bulk material entrucking.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2013年第7期910-914,共5页
China Mechanical Engineering
基金
中央高校基本科研业务费专项资金资助项目(CHD2010ZY011)
关键词
机器视觉
散料装车
料位识别
纹理特征
machine vision
bulk material entrucking
level detection
texture feature
