基于彩色-深度传感器的电容层析成像图像重建方法

Image Reconstruction Method of Capacitance Tomography Based on RGB-D Sensor

工业过程层析成像使用非干扰传感器获取过程容器、反应器等内部状态的二维或三维图像,因在工业应用中获取的电容层析成像资料和信息量有限,难以对影像进行精准、稳定的重构。为此,设计基于彩色-深度(Red Green Blue-Depth,RGB-D)传感器的电容层析成像图像重建方法。利用RGB-D传感器采集电容层析成像图像,采用非局部均值-权重法剔除图像中的噪声后,将图像输入到随机森林分类器中,提取图像的轮廓特征。通过快速投影Landweber算法对轮廓特征求解后,利用电容物质分布的重组完成电容层析成像图像的重建。实验结果显示:所提方法的峰值信噪比数值在35 dB附近波动,结构相似性数值在0.89~1.03之间,重建耗时在4.1 s以下,具有较好的重建效果、质量和结构相似性,能够有效提高重建效率。

For industrial process tomography,non-interference sensors are used to obtain two-dimensional or three-dimensional images of the internal state of process vessels,reactors,etc.Due to the limited amount of image data and information obtained in industrial appli-cations,it is difficult to reconstruct the images accurately and stably.Therefore,an image reconstruction method of electrical capacitance tomography based on Red Green Blue-Depth(RGB-D)sensor is designed.RGB-D sensor is used to collect electrical capacitance tomo-graphy image,non local mean weight method is used to eliminate noise in the image,after that,the image is input into a random forest classifier to extract the contour features of the image.After the contour feature is solved by the fast projection Landweber algorithm,the reconstruction of the electrical capacitance tomography image is completed by the reorganization of the capacitance material distribution.The experimental results show that the peak signal to noise ratio of the proposed method fluctuates around 35 dB,the structural similarity value is between 0.89 and 1.03,and the reconstruction time is less than 4.1 s.It has good reconstruction effect,quality and structural similarity,and can effectively improve the reconstruction efficiency.