光照不均条件下的钢管图像分割算法研究

The research of steel image segmentation algorithm under conditions of uneven illumination

针对传统的全局阈值分割算法在复杂背景下分割不足的缺点以及传统局部动态阈值分割灰度不连续的缺点,在传统局部动态阈值算法的基础上进行了改进。该算法先求取每一局部子图像的阈值,再利用子图像的阈值求取局部动态阈值,最后用求得的局部动态阈值来求取所有图像像素的阈值。用此方法来分割大量无序堆放的钢管图像,实验表明,这种改进算法和几种传统算法比较,能够更精确地分割受光不均及多阴影的复杂背景图像,为后续统计钢管数量的工作奠定了基础。使得图像中钢管的识别率有了明显的提高,接近96%。

In view of the faults of the segmentation is insufficient using traditional global threshold segmentation al- gorithm under complex background and the shortcoming of traditional local dynamic threshold segmentation grey discon- tinuous, based on the traditional local dynamic threshold algorithm. The improved algorithm to calculate each local sub -image threshold, and then using the sub-image threshold to calculate the local dynamic threshold, Finally, local dy- namic threshold was obtained to be used to calculate the threshold values for all image pixels. Using this method to segment a large number of disorderly stacked steel pipe image, the experimental results show that the improved algo- rithm can more accurately segment images under uneven light and multi shadow of complex background compared with several traditional algorithms, which lay the foundation for subsequent work of steel pipe statistics, making that the im- age recognition rate of steel has been significantly improved, close to 96%.