复杂背景下镁熔液第一气泡图像检测方法研究

Study on Image Detection Method of the First Bubble on Magnesium Melt under Complex Background

基于第一气泡法衍生的镁熔液测氢仪精度强烈依赖镁熔液表面析出第一气泡的精确判别。从图像处理方法的角度进行自动识别。镁熔液表面由于受环境振动、梯度温度因素的影响,图像背景缓慢起伏变化,信噪比和对比度都较低,而且气泡目标像素较少,提出一种基于小波变换和自适应阀值分割的算法,并结合数学形态学的开运算,有效地检测出第一气泡,而且应用该算法对第一气泡析出前后的序列图像进行了验证。实验结果表明,采用该算法可以得到满意的结果,目标检测概率为97.3%。

The precision of hydrogen content detection instrument in magnesium melt based on the first bubble method intensively relies on precise recognition of the first bubble precipitated on the surface in magnesium melt.This paper had recognized the bubble automatically by image processing technology.The character of the image represents lower SNR（signal-to-noise ratio） and contrast when the background varied gradually because of the influence of vibration and gradient temperature,and the pixels of the target are low.The detection algorithm of the first bubble was presented that depended on wavelet transform and adaptive gray-level threshold split-run combining with the open algorithm of mathematical morphology,at last the bubble was detected.Based on this algorithm,the verification was constructed for the image sequences of the fore and after of the first bubble precipitated.The result of experiment showed that satisfactory results could be achieved through the detection method mentioned above,the detection probability of the target is 97.3%.