基于特征融合的金属铸造工件表面裂纹检测算法研究

针对复杂结构的金属铸造工件表面因成像复杂引发干扰，裂纹提取判别困难的检测问题，本文提出一种结合了颜色形态特征融合图像分割和纹理特征裂纹判定的金属铸造工件表面裂纹检测算法。算法通过GAMMA变换增强裂纹并弱化背景，根据裂纹目标的颜色特征与几何形状特征相融合，量化特征并滤波特征值分割提取裂纹目标，基于灰度共生矩阵对候选裂纹区域提取纹理特征，使用支持向量机分类器进行训练并识别裂纹。金属工件表面裂纹检测实验表明，该算法在图像分割方面能更加完整准确的提取裂纹，在真伪裂纹的识别中准确率、精确率、召回率和F1得分分别为94.47%、92.51%、96.67%和93.74%。相较于传统检测算法，该算法克服了上述干扰影响，在准确率等方面具有优势，且具有较快的识别速度。

Research on surface crack detection algorithm of metal casting workpiece based on feature fusion

In order to solve the problem that the surface of metal casting workpiece with complex structure causes interference due to complex imaging, and it is difficult to detect cracks, this paper proposes a surface crack detection algorithm of metal casting workpiece based on color morphology feature fusion image segmentation and texture feature crack determination. The algorithm enhanced the crack and weakened the background by GAMMA transformation, fused the color feature and geometric shape feature of the crack target, quantified the feature and filtered the feature value to extract the crack target, extracted the texture feature of the candidate crack region based on the gray co-occurrence matrix, and used the support vector machine classifier to train and identify the crack. The experiments of metal workpiece surface crack detection show that the algorithm can extract cracks more completely and accurately in the aspect of image segmentation, and the accuracy, precision, recall and F1 scores of the real and fake cracks are 94.47%, 92.51%, 96.67% and 93.74%, respectively. Compared with the traditional detection algorithm, the proposed algorithm overcomes the above interference, has advantages in accuracy, and has faster recognition speed.