Adham Mohamed, Ahmad Mostafa, Mohamed Mahmoud, Mohamed Saed

Supervised By: Professor Alaa Hamdy, Eng. Hager Sobeah, Eng. Mahmoud ElSahhar

Publishing Date

29th of December 2020


Concrete pillars are designed to withstand different environmental conditions like earthquakes and to carry tons of loads. Buildings have multiple reinforced concrete columns to carry them and protect them. but, if one of them got damaged, the whole building would lose its integrity causing the building to fall apart costing lives and wasting huge amounts of money. The main idea of the project is identifying defects in concrete by scanning the pillars of the building where the defects like (voids or fractures) might exist or not, to do so we would need to scan the inside of the concrete to acquire these defects that are why engineers and users use the GPR (Ground penetration radar) in the collecting of the images because these defects are impeded within the concrete, after the acquiring phase then the user or the engineer will use his prediction to determine the defect which has a high human error factor because it`s only based over the experience of the user which higher the risk factor even more. Image processing and deep learning techniques will be used to detect and classify those defects according to their types using the VGG-16 algorithm.

1.1 Purpose of this document

The purpose of this document is to demonstrate a system that would be able to fully scan a pillar of concrete and give you a piece of full information about the pillar`s defects using GPR (Ground Penetration Radar), VGG16 module, image processing, and web development.

1.2 Scope of this document

This document targets Quality Control Concrete Management Specialist, Civil Engineers, Structural Engineers, and any user who has used or has a background in the GPR(Ground Penetration Radar), they are going to use our website to detect defects in pillars concrete using B-scan images, which will save them a lot of time, not having to be a certified expert and lowering the human error.

1.3 System Overview

The project is providing a web application system that will depend on the B-scan type images that have been taken by GPR (Ground Penetration Radar) as the user will start the examination process by uploading the images then the system will start the computing process as the images will go through some image processing pipelines which will be automated after that the images that have been resulted from the pre-processing will be stored in a local data storage, then the system will send the positive images to the machine learning process as the system will use VGG-16 model to train data and the negative will be kept in the database, the results will be shown as histogram and mapping to the classes of the defects then it will be stored in the database as it will be extracted from the web application whenever the user needs to display the information.

1.4 System Scope

In our system we will be able to state the difference between many classes of different mediums other than concrete and if the inserted image is in fact concrete, the system will be able to classify and detect its common defects using our trained model also it will detect moisture using image processing . Our system shall:-

1-Sate the difference between many negative classes (which aren’t concrete like walls and other mediums ) and locate the positive images (concrete).

2-Classify and detect common defects like voids, fractures and cracks in the concrete of the pillar.

3-Detect moisture level using image processing techniques.

4-Retrieve a report that describe the defects after scanning them and uploading them to the model.