基于环状分区和LVQ神经网络的混凝土CT图像分割方法

Segmentation Technology of Concrete CT Image Processing Based on Circular Partition and LVQ Neural Network

CT技术是目前了解混凝土细观结构的主要方法,受扫描和视窗成图精度影响,CT扫描切片灰度图像往往无明显双峰、亮度不均匀,导致混凝土骨料、砂浆、孔隙难以分割识别。为提高图像分割时骨料、砂浆、孔隙的识别率,针对混凝土圆柱形试样扫描图像,建立了一种基于环状分区和LVQ神经网络相结合的图像分割算法,通过与其他分割方法进行效果对比表明,该方法具有识别准确度高、人工干预少、后期处理简洁等优点,能够为混凝土细观结构模型建立提供更为准确的数据信息。

CT technology is the main method to understand the meso-structure of concrete.Due to the influence of scanning and window mapping precision,the gray image of CT scanning slice often has no apparent bimodal and uneven brightness,which leads to the difficulty of segmentation and identification of concrete aggregates,mortar and pores.In order to improve the accuracy of aggregate,mortar and pores,aiming at scanning image of concrete cylindrical specimen,an image segmentation algorithm based on circular partition and LVQ neural network is designed.Compared with other segmentation methods,this method has the advantages of high recognition accuracy,less manual intervention and simple post-processing,which can provide more accurate data information for concrete meso-structure modeling.