张浩, 李向春, 杨倩, 吴承璇, 雷卓. 水下气泡光学图像识别方法[J]. 红外与激光工程, 2019, 48(3): 326001-0326001(7). DOI: 10.3788/IRLA201948.0326001
引用本文: 张浩, 李向春, 杨倩, 吴承璇, 雷卓. 水下气泡光学图像识别方法[J]. 红外与激光工程, 2019, 48(3): 326001-0326001(7). DOI: 10.3788/IRLA201948.0326001
Zhang Hao, Li Xiangchun, Yang Qian, Wu Chengxuan, Lei Zhuo. Optical image recognition of underwater bubbles[J]. Infrared and Laser Engineering, 2019, 48(3): 326001-0326001(7). DOI: 10.3788/IRLA201948.0326001
Citation: Zhang Hao, Li Xiangchun, Yang Qian, Wu Chengxuan, Lei Zhuo. Optical image recognition of underwater bubbles[J]. Infrared and Laser Engineering, 2019, 48(3): 326001-0326001(7). DOI: 10.3788/IRLA201948.0326001

水下气泡光学图像识别方法

Optical image recognition of underwater bubbles

  • 摘要: 针对水中气泡与固体悬浮微粒不易区分的问题,提出了一种基于Zernike矩与灰度计算的水下光学气泡图像识别方法。该方法主要分为图像划分、图像预处理和特征提取三个步骤。首先,获取水下悬浮微粒的图像,从中划分出单个气泡并选取部分样本;为了更好地提取与表示气泡轮廓与灰度特征,然后采用图像预处理方法增强气泡边缘特征,选择并构建气泡特征库;最后,采用Zernike矩计算悬浮微粒特征的相似度,区分圆形微粒与非圆形微粒,之后计算微粒中心与灰度变化趋势,辨别气泡与固体悬浮微粒。实验结果表明,在测试数据集上的气泡识别准确率达到94%。该方法不仅能够辨别圆形与非圆形微粒,而且能够融合灰度梯度计算方法以获取更好的结果。该方法从形状与灰度两个方面提取与辨别目标的特征信息,提高了气泡识别精度,具有较高的精确性与适用性。

     

    Abstract: A new method of bubble recognition using optical underwater imaging was presented by employing Zernike moments and gray gradient, to differentiate bubbles from solid particles. This method included 3 parts: image division, image pre-processing and feature extraction for bubble recognition. Firstly, images of the suspended particles were obtained from underwater particle database, in which a particular bubble was divided and selected manually from the whole. Secondly, image pre-processing was employed to enhance single bubble images, to extract and represent bubble silhouette and gray level. Thus, the database of bubble features were selected and formed. Finally, the shape descriptor, Zernike moments, was utilized to measure the similarity with features of other suspended particles to differentiate circle particles from the irregular ones. Subsequently, the center of circle particle and the trend of gray gradient were computed, so as to distinguish the bubbles from solid particles. The experimental results show that, the accuracy of bubble recognition is up to 94%. It is concluded that this method not only recognizes bubbles from irregular suspensions, but also improves gray gradients calculation for enhanced results. By extracting and distinguishing object features through the prospects of both shape and gray, this method enhances the accuracy of bubble recognition, with higher precision and broader suitability.

     

/

返回文章
返回