Image Segmentation of Bricks in Masonry Wall Using a Fusion of Machine Learning Algorithms

Abstract

Autonomous mortar raking requires a computer vision system which is able to provide accurate segmentation masks of close-range images of brick walls. The goal is to detect and ultimately remove the mortar, leaving the bricks intact, thus automating this constructionrelated task. This paper proposes such a vision system based on the combination of machine learning algorithms. The proposed system fuses the individual segmentation outputs of eight classifiers by means of a weighted voting scheme and then performing a threshold operation to generate the final binary segmentation. A novel feature of this approach is the fusion of several segmentations using a low-cost commercial offthe-shelf hardware setup. The close-range brick wall segmentation capabilities of the system are demonstrated on a total of about 9 million data points.

Weakly Supervised SVM-Enhanced SAM Pipeline for Stone-by-Stone Segmentation of the Masonry of the Loire Valley Castles

“…The dataset was comprised of 2814 crops of 224 × 224 px from 107 images with a variety of brick colors, angles, illumination, and resolution. The joints of the brick walls studied in the previous work [ 12 , 13 , 14 ] are clearly distinguishable from bricks, with distinguishing colors, unlike the limestone facades of renaissance castles used in the present study, which have very homogeneous joints and stones.…”

“…The dataset includes 245 images of 256 × 256 px manually labeled for training and testing. In another context, for bricks segmentation, Kajatin [ 12 ] proposed the analysis and fusion of eight classifiers (kNN, Bayes, QDA, SVM, decision tree, random forest, AdaBoost, U-Net) for the segmentation of closed range photos of reddish bricks. The dataset used was composed of 27 manually labeled photos of 848 × 480 px.…”

Masonry Structure Analysis, Completion and Style Transfer Using a Deep Neural Network

“…Several wall image delineation techniques are presented in the literature for various wall types. Many of them are working on 2D images of walls, relying on various image features and modeling approaches, such as a combination of objectoriented and pixel-based image processing technology [3], a color-based automated algorithm based on an improved marker-controlled watershed transform [16], a Hough transform-based delineation algorithm [14], a machine learningbased algorithm [9], and a deep neural-based network for stone-by-stone segmentation [7]. As for using 3D point clouds instead of images, [20] used a Continuous Wavelet Transform (CWT) applied on a 2.5D depth map to obtain the outlines of the bricks.…”