MS
M. Smit
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
3 records found
1
Over the years, the pace at which data is generated keeps on increasing. As a consequence, the data itself no longer holds the highest value, but rather the information and context the data captures are. This principle also holds in the 3D environment modelling scene, as accurately depicting an environment holds more value than the number of models there are of it.
One of the major problems in 3D environments, especially when the environment represents a building, is the presence of glass. A lot of the data captured to model these 3D environments is captured using LiDAR laser scanning. This is where glass becomes a problem as glass is almost completely transparent to laser beams at the typical wavelengths used when using LiDAR laser scanning. As a consequence, glass can lead to problems with navigational routes as it is invisible in the environment but still blocks the path. It can also create false spaces in the captured environment as it can also partially act as a mirror reflecting the laser beam and showing these reflections in space as if they were captured in a straight line.
Alternative manners for capturing and identifying glass in environments captured with laser have been created over the years, but they often need a dedicated set-up, expensive equipment or a lot of data. These solutions are not always feasible for users of point cloud data.
Therefore, in this thesis a focus is put on how can a low entry solution be created for this problem, which leads to the main research question: How can the location of glass be deduced using only information acquired from 3D point clouds and a reference position?
To answer this question, this thesis focuses on the deduction of the locations of glass windows in the provided input. To find these a projection from 3D data to 2D is performed. In 2D image space, contours are then detected that match the criteria of window frames. These contours are then used to segregate parts of the 3D point cloud that should contain the window detected in the projection. After clustering these parts and the best matching cluster is deduced to be a window.
In this thesis, it is shown that using the proposed methodology it is possible to deduce the location of glass in a LiDAR point cloud using only an additional reference position, but there are some flaws with the simplified input of the method.
...
One of the major problems in 3D environments, especially when the environment represents a building, is the presence of glass. A lot of the data captured to model these 3D environments is captured using LiDAR laser scanning. This is where glass becomes a problem as glass is almost completely transparent to laser beams at the typical wavelengths used when using LiDAR laser scanning. As a consequence, glass can lead to problems with navigational routes as it is invisible in the environment but still blocks the path. It can also create false spaces in the captured environment as it can also partially act as a mirror reflecting the laser beam and showing these reflections in space as if they were captured in a straight line.
Alternative manners for capturing and identifying glass in environments captured with laser have been created over the years, but they often need a dedicated set-up, expensive equipment or a lot of data. These solutions are not always feasible for users of point cloud data.
Therefore, in this thesis a focus is put on how can a low entry solution be created for this problem, which leads to the main research question: How can the location of glass be deduced using only information acquired from 3D point clouds and a reference position?
To answer this question, this thesis focuses on the deduction of the locations of glass windows in the provided input. To find these a projection from 3D data to 2D is performed. In 2D image space, contours are then detected that match the criteria of window frames. These contours are then used to segregate parts of the 3D point cloud that should contain the window detected in the projection. After clustering these parts and the best matching cluster is deduced to be a window.
In this thesis, it is shown that using the proposed methodology it is possible to deduce the location of glass in a LiDAR point cloud using only an additional reference position, but there are some flaws with the simplified input of the method.
...
Over the years, the pace at which data is generated keeps on increasing. As a consequence, the data itself no longer holds the highest value, but rather the information and context the data captures are. This principle also holds in the 3D environment modelling scene, as accurately depicting an environment holds more value than the number of models there are of it.
One of the major problems in 3D environments, especially when the environment represents a building, is the presence of glass. A lot of the data captured to model these 3D environments is captured using LiDAR laser scanning. This is where glass becomes a problem as glass is almost completely transparent to laser beams at the typical wavelengths used when using LiDAR laser scanning. As a consequence, glass can lead to problems with navigational routes as it is invisible in the environment but still blocks the path. It can also create false spaces in the captured environment as it can also partially act as a mirror reflecting the laser beam and showing these reflections in space as if they were captured in a straight line.
Alternative manners for capturing and identifying glass in environments captured with laser have been created over the years, but they often need a dedicated set-up, expensive equipment or a lot of data. These solutions are not always feasible for users of point cloud data.
Therefore, in this thesis a focus is put on how can a low entry solution be created for this problem, which leads to the main research question: How can the location of glass be deduced using only information acquired from 3D point clouds and a reference position?
To answer this question, this thesis focuses on the deduction of the locations of glass windows in the provided input. To find these a projection from 3D data to 2D is performed. In 2D image space, contours are then detected that match the criteria of window frames. These contours are then used to segregate parts of the 3D point cloud that should contain the window detected in the projection. After clustering these parts and the best matching cluster is deduced to be a window.
In this thesis, it is shown that using the proposed methodology it is possible to deduce the location of glass in a LiDAR point cloud using only an additional reference position, but there are some flaws with the simplified input of the method.
One of the major problems in 3D environments, especially when the environment represents a building, is the presence of glass. A lot of the data captured to model these 3D environments is captured using LiDAR laser scanning. This is where glass becomes a problem as glass is almost completely transparent to laser beams at the typical wavelengths used when using LiDAR laser scanning. As a consequence, glass can lead to problems with navigational routes as it is invisible in the environment but still blocks the path. It can also create false spaces in the captured environment as it can also partially act as a mirror reflecting the laser beam and showing these reflections in space as if they were captured in a straight line.
Alternative manners for capturing and identifying glass in environments captured with laser have been created over the years, but they often need a dedicated set-up, expensive equipment or a lot of data. These solutions are not always feasible for users of point cloud data.
Therefore, in this thesis a focus is put on how can a low entry solution be created for this problem, which leads to the main research question: How can the location of glass be deduced using only information acquired from 3D point clouds and a reference position?
To answer this question, this thesis focuses on the deduction of the locations of glass windows in the provided input. To find these a projection from 3D data to 2D is performed. In 2D image space, contours are then detected that match the criteria of window frames. These contours are then used to segregate parts of the 3D point cloud that should contain the window detected in the projection. After clustering these parts and the best matching cluster is deduced to be a window.
In this thesis, it is shown that using the proposed methodology it is possible to deduce the location of glass in a LiDAR point cloud using only an additional reference position, but there are some flaws with the simplified input of the method.
SCIPoC: Semantic Classification of Indoor Point Cloud
A study into the possibilities of classifying indoor point cloud using a Deep Learning approach
Student report
(2020)
-
M. Smit, Z. Chen, M.A. Erbaşu, Y.A.L. Gaol, X. Li, E. Verbree, B.M. Meijers, J. Balado Frías, N. van der Vaart, R. Bunder
With the constantly evolving range of applications for technology the quality and amount of data constantly increases as well. In this growing data environment, there is a constant search to provide more value to all data that is available for as little effort as possible. Our research tries to add such additional value by diving into the concept of classifying point cloud by using deep learning, specifically in the indoor environment. This is done by first doing a neural network comparison and then doing a case study. In the neural network comparison, a look is taken into which of the neural networks that are capable of working with point clouds is best suited for our experiments in the indoor scene, based on the training speed, accuracy, ease of use concerning training on external datasets and setting up the network and space efficiency. After the comparison, we chose to continue with the PointCNN network during the case study. The case study is performed on data the NS (Nederlandse Spoorwegen) provided to us and all test results we got from our experiments can be visualized using the web application we developed along with this project. The purpose of the case study is to add extra value to the indoor LiDAR point cloud the NS has captured from Amersfoort Station by using deep learning to automatically classify assets present in their data. The value is in purposes, such as asset management, where the data does not need possibly hundreds of man-hours to be labelled. This saves a lot of time and also money each time a scan is made. In the case study we found through 4 different experiments that unbalanced data makes for bad results, but when a scene is labelled correctly very good results can be found in a local scene.
...
With the constantly evolving range of applications for technology the quality and amount of data constantly increases as well. In this growing data environment, there is a constant search to provide more value to all data that is available for as little effort as possible. Our research tries to add such additional value by diving into the concept of classifying point cloud by using deep learning, specifically in the indoor environment. This is done by first doing a neural network comparison and then doing a case study. In the neural network comparison, a look is taken into which of the neural networks that are capable of working with point clouds is best suited for our experiments in the indoor scene, based on the training speed, accuracy, ease of use concerning training on external datasets and setting up the network and space efficiency. After the comparison, we chose to continue with the PointCNN network during the case study. The case study is performed on data the NS (Nederlandse Spoorwegen) provided to us and all test results we got from our experiments can be visualized using the web application we developed along with this project. The purpose of the case study is to add extra value to the indoor LiDAR point cloud the NS has captured from Amersfoort Station by using deep learning to automatically classify assets present in their data. The value is in purposes, such as asset management, where the data does not need possibly hundreds of man-hours to be labelled. This saves a lot of time and also money each time a scan is made. In the case study we found through 4 different experiments that unbalanced data makes for bad results, but when a scene is labelled correctly very good results can be found in a local scene.
Bachelor thesis
(2019)
-
Tom Edixhoven, Hunter van Geffen, Bas Kruit, Mels Smit, Maurício Aniche, Otto Visser, Huijuan Wang
For a steel company it is advantageous to be able to easily track steel through the production process. At Tata Steel this is currently done with the Material Tracking Table. However, generating this table takes months. Therefore a new system had to be developed. This paper describes the building of such a new system, which generates this Material Tracking Table in less than 1 hour, as well as the related systems concerning the acquisition of the input data and the visualisation of the resulting output data.
...
For a steel company it is advantageous to be able to easily track steel through the production process. At Tata Steel this is currently done with the Material Tracking Table. However, generating this table takes months. Therefore a new system had to be developed. This paper describes the building of such a new system, which generates this Material Tracking Table in less than 1 hour, as well as the related systems concerning the acquisition of the input data and the visualisation of the resulting output data.