3D head reconstruction solution
Our client is a creative design agency that provides branding, advertising, graphic design, and virtual modeling services for individuals and corporate customers. The agency has its own in-house souvenir production.
Building automation system
Japan is a country with increased seismic activity. It is important to ensure the stability of structures during construction. The goal of this project was to automate the marking of reinforcing bars using an AI solution.
Project background
As we know, Japan is a country with increased seismic activity. It means it’s very important to ensure the stability of structures during construction. The main goal of our client’s project was to automate the marking of reinforcing bars that provide construction stability using an AI solution.
As a rule, in order to demonstrate the availability of the required number of reinforcing bars, builders have to attach colored markers to each section of the reinforcing mesh. Markers are attached horizontally and vertically.
Plastic markers often fall off. In addition, the marking process has to be repeated hundreds of times. Together with colored markers, builders attach a ruler for a photo to measure the distance between reinforcing bars.
It is physically difficult and time-consuming to carry out this work on marking reinforcing bars and calculating the distance.
- September 2021 – December 2021
- Japan
- Construction
- Open CV
- sklearn
- PyQt5
- Python
- QtDesigner
- Select or take a photo A user selects a photo from the gallery or takes a photo using the camera.
- Detect the outer layer The user can press Detect and get the result of the outer layer reinforcement detection.
- Calculate the distance The user can click Calculate and the distance between the rebars will be calculated.
Business needs
As a rule, in order to demonstrate the availability of the required number of reinforcing bars, builders have to attach colored markers to each section of the reinforcing mesh. Markers are attached horizontally and vertically.
Plastic markers often fall off. In addition, the marking process has to be repeated hundreds of times. Together with colored markers, builders attach a ruler for a photo to measure the distance between reinforcing bars.
It is physically difficult and time-consuming to carry out this work on marking reinforcing bars and calculating the distance. Our application helps to automate the control over the location of reinforcing rods on construction sites.
The main challenges:
- It was hard to find an approach to select the first layer from the total mass of reinforcing bars and calculate the distance between them.
- Through analysis, we decided that we needed to use something that would serve as a scale. So, a special tape was made, which served as a filter and a measurement scale for computer vision.
- It was not easy to make the tape with the desired transparency so that the front reinforcing bars were clearly visible and those in the background were less visible.
- We had to conduct a number of experiments to work out an algorithm that took the app from idea to reality.
- Our client needed a working solution within 3 months.
Solution
To create this automated system, the CHI Software development team took the following steps:
-
We used image pre-processing techniques based on classical CV to reduce noise and to remove unnecessary objects from the image, to highlight main rebars, and separate them from inappropriate rebars
-
We used lines Detection with Hough Transform to detect rebars on images with high quality and fast inference
-
We used different color spaces and morphological image processing to detect tape without deep learning models
-
We wrote a lot of logic for post-processing to filter and select only appropriate rebars
-
We created automatic distance calculations between rebars
Client values
- Our application helps to automate the control over the location of reinforcing bars on construction sites.
- We have successfully implemented an application that helps to reduce the time spent and the effort that builders make to check the reliability of reinforcement structures.
- We delivered a working app within 3 months.
Employee testimonial
As an ML/CV Engineer, I was responsible for tasks related to Computer Vision. The main goal of the project was to automate the process of detecting rebars in the image of the construction and to estimate the distance between rebars. To create such an automated system, the following steps were taken: - we used image pre-processing techniques based on a classical CV to reduce noise and remove unnecessary objects from the image, highlight main rebars, and separate them from inappropriate rebars; - we used lines Detection with Hough Transform to detect rebars on images with high quality and fast inference; - we used different color spaces and morphological image processing to detect tape without deep learning models; - we wrote a lot of logic for post-processing to filter and select only appropriate rebars; - we created automatic distance calculations between rebars. We delivered a working app within 3 months. Our team consisted of 5 experienced engineers and a PM. We successfully implemented an application that helps to reduce the time and effort that builders need to check the reliability of reinforcement structures.
Vladyslav Marchenko ML/CV Engineer