基于图像处理的冷轧薄板板形识别

Cold-rolled strip plate-profile recognition based on image processing method

为了解决冷轧薄板板形识别问题,采用基于图像处理的方法,对图像进行直方图均衡化,使处理后的图像对比度、图像边界的清晰度有了很大的提高,并在此基础上再进行形态学增强,改善图像效果.对于处理后的图像利用canny算子提取其边缘,利用图像的均值、方差和对比度的统计特征作为BP神经网络分类器的输入进行缺陷特征分类.根据上述方法进行了板形识别系统的硬件、软件设计,实际应用表明,该方法可以有效地识别出常见的板形缺陷.

A cold-rolled strip plate-profile recognition method was presented to improve the detection accuracy based on image processing. The contrast of original flatness strip image and the edge definition were increased through gray-level histogram equalization. The image morphology filtering method was used to improve the image quality, and the edge detection was processed using canny operator. The back propagation （BP） neural network with three input parameters： mean, variances and contrast of image were proposed to classify the plate-profile defection. Application in an iron sted plant showed the effectiveness of the proposed method, and the hardware and software system in this application were also presented.