Course detail

Reconstruction and Analysis of 3D Scenes

FSI-SR0-A Acad. year: 2027/2028 Winter semester

The subject concerns the reconstruction and analysis of 3D scenes based on point clouds. This research area is important in reverse engineering, robotics, geography, and autonomous traffic systems. The first part of the lectures deals with the data acquisition types and algorithms, such as registration, edge detection, and feature extraction. We make the measurements in the laboratory using an optical scanner (ATOS) and other hand scanners. We will process the point cloud with software GOM inspect and Rhinoceros. We will also work with 3D printer Prusa.
We will use Arduino Engineering Kit to test the proposed algorithms.
Students prepare their own implementation of some of the algorithms in the last lectures of the semester.

Language of instruction

English

Number of ECTS credits

5

Learning outcomes of the course unit

Upon successful completion of the course, students will be able to:

  1. 3D Data Acquisition Methods: Describe and compare basic active and passive methods of acquiring three-dimensional data, particularly Structure from Motion, Time of Flight, and laser scanning.

  2. Point Cloud Registration: Explain the principles of PCA, SVD, and ICP methods and apply selected algorithms to point cloud registration and processing.

  3. 3D Scene Analysis: Implement and apply selected algorithms for point cloud analysis, particularly RANSAC and methods for detecting geometric objects and characteristic features of a scene.

  4. 3D Scanning: Perform 3D scanning of a selected object using an available 3D scanner and process the acquired spatial data.

  5. 3D Modelling and Printing: Process and modify acquired data using appropriate software, create a geometric model, and prepare it for 3D printing.

  6. Practical Applications of 3D Data: Apply selected spatial data processing techniques to solve a basic practical problem using the Arduino/Raspberry.

  7. Independent Project: Design, implement, and evaluate a solution to a selected problem in 3D scene reconstruction or analysis, and present and defend the chosen approach and obtained results.

Entry knowledge

elementary knowledge of mathematical analysis and algebra (matrix, derivative), computer graphics, the recommended is the knowledge of programming language (C, C++, Pascal, atd) or programm (e.g. Matlab).

Rules for evaluation and completion of the course

The course is completed by a colloquium. Assessment is based on students' active participation in computer-based exercises, continuous evaluation of their work, and their ability to understand the methods presented and apply them to practical problems involving real 3D data.

During the semester, students work individually or in groups on assigned practical tasks and projects. Assessment also includes either a final individual project or a series of projects completed throughout the semester.

The assessment focuses primarily on understanding the methods used, their correct application, the student's individual contribution to solving the assigned tasks, and the ability to explain and interpret the obtained results.

The colloquium takes the form of a discussion of the completed tasks and projects, during which students demonstrate their understanding of the applied methods and the results achieved.

Assessment is qualitative, without a point-based grading system. If the requirements are not sufficiently met, students may complete or revise the assigned tasks or projects and submit them for reassessment in accordance with the rules governing the colloquium.

Attendance at practical exercises is compulsory. Attendance at lectures is recommended, as their content is directly related to the practical exercises.

Aims

The aim of the course is to introduce students to the principles of three-dimensional scene reconstruction and analysis, methods of acquiring and processing spatial data, and their practical applications. Emphasis is placed on applying algorithms to real 3D data and developing students' ability to independently solve practical problems.

The main benefit of the course is to provide students with an understanding of point clouds, methods of acquiring them through 3D scanning, and their subsequent processing using algorithms for edge and object detection, point cloud registration, and spatial data analysis. Students will also become familiar with practical applications of the acquired data, particularly in creating models for 3D printing and analysing three-dimensional scenes.

Study aids

All study materials and teaching resources, including lecture presentations, practical assignments, project specifications, and additional supporting materials, are available to students through the course's e-learning platform.

The study programmes with the given course

Programme N-MAI-A: Mathematical Engineering, Master's, compulsory-optional

Type of course unit

 

Lecture

26 hours, optionally

Syllabus

1st week: The data acquisition (Terrain, Mobile, Airborne) in dependance of the application. Passive (Structure from Motion) and active methods (Time of Flight, laser)
2nd week: Point cloud registration (methods PCA, SVD, ICP).
3rd week: RANSAC – algorithm and its usage, feature extraction
4-5th week: Laboratory measurements on the optical scanner ATOS and hand scanners.
6th week: Point cloud processing in software (GOM Inspect, Rhinoceros, etc.)
7th week: 3D printing – principles, settings, problems.
8th week: Arduino Engineering Kit – vehicle Rover can navigate between given reference points, move objects with a forklift.
9-12th week: Consultations
13th week: Presentation of the seminar work.

Computer-assisted exercise

26 hours, compulsory

Syllabus

1st week: The data acquisition (Terrain, Mobile, Airborne) in dependance of the application, Matlab programming.
2nd week: Point cloud registration – MATLAB or other programming languages.
3rd week: RANSAC programming.
4-5th Laboratory measurements on the optical scanner ATOS and using hand scanners.
6th week: Point cloud processing in software (GOM Inspect, Voxelizer, Rhinoceros). Creating own models for 3D printing.
7th week: 3D printing of models (Voxelizer 3D printer)
8th week: Arduino Engineering Kit – vehicle rover programming.
9-12th week: Seminar work with consultations.
13th week: Presentation of the seminar work.