基于Gabor和灰度共生矩阵混合特征叶片泵装配质量检测

本文提出一种Gabor和灰度共生矩阵相结合的特征来检测叶片泵中叶片装配质量的方法。首先构建叶片图像数据集,用5种尺度的和4种方向的Gabor滤波器对图像滤波,根据滤波后的图像计算得到幅值特征图,然后提取幅值特征图的灰度共生矩阵特征,最后融合归一化各个幅值特征图提取到的特征,利用主成分分析法降维,并用这些特征向量训练支持向量机(SVM)分类器,实现对叶片装配质量的评估。将本文提出的混合特征与LBP特征、灰度共生矩阵分别进行了比较得到的分类效果约提高了约10%。基于Gabor和灰度共生矩阵混合特征的叶片装配质量检测准确率提升到了93%。实验结果表明Gabor特征和灰度共生矩阵结合后能够很好从多尺度、多方向上提取图像的纹理特征,并应用于图像分类取得了良好的效果,在一些图像识别上有很宽广的应用前景。

Vane pump assembly quality detection based on gabor and gray level co-occurrence matrix hybrid characteristics

This paper presents a method to detect blade assembly quality in a vane pump by features of the combination of Gabor and gray level co-occurrence matrix. Firstly, the blade image dataset is constructed, and the image is filtered by five-scale and four-direction Gabor filters. The amplitude feature map is calculated on the basis of the filtered image, and then the gray level co-occurrence matrix feature of the amplitude feature map is extracted. Finally, we combine and normalize the features extracted from each feature map, reduce the dimensions by principal component analysis, train the support vector machine (SVM) classifier with these feature vectors to achieve the evaluation of blade assembly quality. Respectively comparing mixing features proposed in this paper with the LBP features and gray level co-occurrence matrix, the classification effect is improved by about 10%. The classification effect of blade assembly quality is improved to 93% based on the combination of Gabor and gray level co-occurrence matrix. The experimental results show that the combination of Gabor features and gray level co-occurrence matrix can extract texture features by multi-scale and multi-directions, and obtain good results by applying to image classification. It has wide application prospects in some image recognition.