带钢表面缺陷在线检测图像噪声滤除算法研究

Research of Image Noise Filtering Algorithm Online Strip Surface Defect Inspection System

在对带钢表面缺陷进行视觉检测的过程中,光照条件、金属碎末飞溅等外界干扰会产生图像噪声,这种噪声会直接影响到后续的带钢缺陷检测。因而对图像进行预处理十分必要。提出了一种针对带钢图像表面噪声预处理的算法,该算法通过分析带钢表面噪声分布的特点,对其进行标识,进而提出了改进的标准中值算法,对所有噪声点邻域内的中值和均值的差值与设定阈值进行比较,实现对噪声的有效滤除。通过大量的实验证明该算法有效可行,在算法的复杂度上有较大的降低,在噪声滤除效果上有明显的改善。

In the process of strip surface defect detection by the way of vision method, image noise will be produced as a result of metal fragments ,light condition and so on,which will directly affect the subsequent image measurement accuracy. Thus, the image preprocessing is very important. Improving standard median filtering method,a new preprocessing algorithm used in strip sufrace defect detection is proposed. In this algorithm, the noise distribution characteristics is studied firstly, then the differentials of mid-value and mean value in the neighborhood of all noise point is compared with the threshold. The algorithm is proved effective and feasible by a large number of experiments, meanwhile,complexity of the algorithm reduced greatly and noise filtering effect improved significantly.