水下微孔气泡群的图像分割及特征检测

Image Segmentation and Features Detection of Submerged Microporous Bubbles

摘要: 目前，检测水下大气泡群特征的主要方法是用数字图像处理技术处理高速摄像机采集到的气泡群图像。本文针对高速摄像机采集到的斜上方投光的水下小孔气泡群图像，研究了基于改进分水岭算法的气泡群图像分割和基于区域生长算法的气泡群特征检测方法，解决了图像中粘连、重叠气泡的分割及特征检测，较常用的阈值分割方法检测气泡群特征更准确；同时，解决了斜上方投光拍摄的气泡群图像噪声多、气泡亮度不均匀造成的传统分水岭算法产生过度分割的问题。

Abstract: At present, the main method for features detection of submerged bubbles is Digital image pro-cessing of bubbles images photoed by high speed camera. In this paper, we develop a bubbles segmentation algorithm based on improved watershed algorithm as well as an algorithm for bub-bles features detection based on regional growth method, which solves the segmentation and fea-tures detection for adhesive and overlap bubbles for the bubbles image photoed by high speed cam-era with oblique-upper irradiation. It is more accurate to detect the features of bubbles than tradi-tional threshold segmentation. The algorithm in this paper also eliminates ultra-segmentation of classical watershed algorithm caused by noise and uneven brightness of bubbles.