结合邻域信息的带钢表面明暗域缺陷图像配准

Strip Steel Surface Bright/Dark Field Defect Image Registration Combined with Neighborhood Information

为解决带钢表面明暗域缺陷图像检测方式采集到的同一位置处的图片之间存在位移及旋转问题,同时考虑到因板带振动及现场生产环境造成的噪声影响,提出采用结合邻域内像素间空间与灰度信息的互信息配准方法.在互信息配准计算中图像中每个像素的灰度值由其邻域内像素的灰度值按照距离及灰度变化关系分配不同的权值共同得到.实验证明,该方法的配准精度完全可满足带钢缺陷检测需求,有效减弱了噪声对配准精度的影响,为带钢表面缺陷识别及质量评价奠定了基础.

To get rid of the displacement and rotation as shown in the two strip steel surface defect images at the same position,which are obtained by the detection system for both bright and dark fields,and denoise the images detected during strip vibration and in-situ production process,a new mutual-information-based registration of images was proposed combining the pixel spaces in neighborhood with grey level data. During the computation of the mutual-information-based registration of images,the grey level of every pixel is determined together with the grey level of a pixel in its neighborhood,which distributes the different weights according to the relationship between distance and grey level. Experimental results showed that the accuracy of the new method of registration can meet fully the requirements for strip steel defect detection and reduce effectively the impact of noise on registration accuracy,thus laying a foundation to identify the strip surface defects and evaluate the strip quality.