基于层间信息熵的混凝土3D打印构件精细分割

Fine Segmentation of Concrete 3D-Printed Elements Based on Information Entropy Between Layers

面向复杂环境下固废基混凝土3D打印构件的智能检测需求,引入机器视觉理论,提出一种基于层间信息熵的目标精细分割算法,实现3D打印构件的特征分析与智能检测。首先,考虑到混凝土3D打印的复杂环境,构建一种视觉特征增强的预处理方法,通过Gamma灰度变换和直方图均衡化算法调整对比度,增强图像特征细节,并结合自适应中值滤波去除图像中的随机噪声;随后,针对构件的分层叠加特性,定义了层间信息熵指标,设计了一种基于层间信息熵和双阈值优化的打印构件精细分割方法,实现了复杂环境下对3D构件的分层检测与精细分割;最后,采集真实混凝土3D打印构件的目标影像,对所提算法进行有效性验证。实验结果表明,所提算法的准确率值平均提高了12.44%,F_(1)值平均提高了30.79%,极大地提高了目标分割精度,为进一步实现3D打印构件的精确测量与路径优化奠定基础。

Because of the demand for intelligent detection of solid waste-based concrete three-dimensional(3D)-printed components in complex environments, this paper introduces the machine vision theory and proposes a target fine-segmentation algorithm based on the interlayer information entropy to realize the feature analysis and intelligent detection of 3D-printed components. First, considering the complex environment of concrete 3D printing, a preprocessing method for visual feature enhancement was constructed, the contrast was adjusted, and the image feature details were enhanced using Gamma grayscale transformation and histogram equalization algorithm. It was combined with adaptive median filtering to remove the random noise in images. Then,considering the layered superposition characteristics of the components, the interlayer information entropy index was defined, and a fine-segmentation method of printing components based on the interlayer information entropy and double threshold optimization was designed to realize the complex environment hierarchical detection and fine segmentation of 3D components. Finally, the target images of real concrete 3D-printed components were collected to verify the effectiveness of the proposed algorithm. Experimental results show that the proposed algorithm increases the accuracy by 12.44% and the F_(1) value by 30.79% on average, considerably improving target segmentation accuracy. It lays the foundation for further realizing accurate measurement and path optimization of 3D-printed components.