3D Segmentation & Anomaly Detection for Construction Use-Case (on-campus)

Students will develop a tool that takes multiple pictures at required angles and feed them for 3-D reconstruction. Once the image is reconstructed, students will develop post-processing techniques for segmentation of area-of-interest and anomaly detection in 3-D space. Students will be provided with a sandbox where camera images from the android application are fed, and 3-D reconstruction output can be availed using API(s). Once 3-D images are reconstructed, post-processing with be researched to identify anomalies with photogrammetric accuracies. The results will be compared with the physical models.

Objectives

  1. Camera Android Application. Students will start with the camera app that was developed by previous OSU capstone students and improve the camera/gimble handling.
  2. Students will develop a frontend to bind it to the cloud backend for task handling, image processing, previews, etc.
  3. Students will develop algorithms to identify the anomalies in 3-D rendered models and segment the area of interest.
  4. Optional: Students will develop a unity application to draw the 3-D model (and handling) and visualize it.

Motivations

Remote inspections are key to improving productivity and travel costs. Furthermore, as the experienced workforce is retiring, we must maintain inspection quality despite the shortage of inspectors and skilled workers. Accurate 3-D models with fine-grained measurement abilities go a long way in improving productivity and worker safety.

Qualifications

Minimum Qualifications:
  1. AI Development (2): Familiarity with Vision-based AI algorithms, Computer Vision, OpenCV, Desired/Optional Open3D, Python
  2. Cloud/UX Development(1): Cloud/Frontend programming
  3. Android Development (1): App development for camera automation with Gimble 
  4. Options AR Development: Visualize the 3-D model using HoloLens 

Preferred Qualifications:

CS, EECS


Details

Project Partner:

Rahul Khanna

NDA/IPA:

No Agreement Required

Number Groups:

4

Project Status:

Accepting Applicants

Keywords:
3-D Reconstruction Computer Vision 3D-CNN Metashape
Card Image Capstone