RECONSTRUCTION OF HEAD PHANTOM̉̉S RANGE IMAGE ACQUIRED BY KINECT CAMERA USING NURBS
In radiotherapy procedure, patient position during treatment time can be monitored from internal patient body for example using EPID (Electronic Portal Imaging Device), but the problem is that this method gives additional radiation exposure to patient. Other method is monitoring from patient surface...
Saved in:
Main Author: | |
---|---|
Format: | Theses |
Language: | Indonesia |
Online Access: | https://digilib.itb.ac.id/gdl/view/22115 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Institut Teknologi Bandung |
Language: | Indonesia |
id |
id-itb.:22115 |
---|---|
spelling |
id-itb.:221152017-11-17T15:18:07ZRECONSTRUCTION OF HEAD PHANTOMÃâÃâS RANGE IMAGE ACQUIRED BY KINECT CAMERA USING NURBS PUSPA LESTARI (NIM : 20214059), FAUZIA Indonesia Theses INSTITUT TEKNOLOGI BANDUNG https://digilib.itb.ac.id/gdl/view/22115 In radiotherapy procedure, patient position during treatment time can be monitored from internal patient body for example using EPID (Electronic Portal Imaging Device), but the problem is that this method gives additional radiation exposure to patient. Other method is monitoring from patient surface for example by using surface marker or range image camera. On the other hand, there is range image camera which has been popular in gaming called Kinect camera. Kinect camera has infrared source and range image camera which can produce point distribution. But, to detect patient position we should detect feature points from smooth surface. For that purpose, we have developed an automated approach for reconstructing mathematical surfaces of head phantom’s range images acquired by Kinect camera. Three hundred frames of head phantom’s range images were acquired by the Kinect camera which was placed 1 meter from the nose of a phantom with 45 degree angle with respect to a vertical line. Distance measured was then corrected by using a linear interpolation method to decrease errors between reference distance and measured distance, which were caused by lens distortion. The next step was to localize a region of interest (ROI) using a template matching technique in gray scale images taken by the Kinect camera. Since the distances measured at the same pixels of the range images may fluctuate with frame, a temporal filtering technique with Kalman filter and moving average filter was applied to reduce the fluctuation and noise. Then, the range images were resampled with an interval of 0.1 mm, followed by spatial filtering using bilateral filter to smooth while preserving edge of the image. The last step was the reconstruction of mathematical surfaces using NURBS with assumption that all of point’s weights have the same value. The parameters of NURBS were optimized by evaluating roughness and edge slope ratio of reconstructed mathematical surfaces with changing the degree of B-spline functions and an interval between control points. The roughness of the reconstructed surface was calculated by using standard deviation ratio of 10 by 10 ROI in original image and NURBS surface. The higher NURBS degree results had the lower roughness which means that the smoother surfaces were obtained. To quantify edge preservability, the ratio between edge slope of NURBS and original image (reffered to edge slope ratio: ESR) was calculated. Higher ESRs would represent the better edge preserving on the surface. The result shows that surface with a sampling interval 10 points, NURBS degree equal to 3 provided better surface with a roughness 0.0962 and a highest ESR of 0.6919. We concluded that the proposed approach for reconstructing mathematical surfaces can provide smooth surfaces, which may detect patient positioning errors by using the Kinect camera. text |
institution |
Institut Teknologi Bandung |
building |
Institut Teknologi Bandung Library |
continent |
Asia |
country |
Indonesia Indonesia |
content_provider |
Institut Teknologi Bandung |
collection |
Digital ITB |
language |
Indonesia |
description |
In radiotherapy procedure, patient position during treatment time can be monitored from internal patient body for example using EPID (Electronic Portal Imaging Device), but the problem is that this method gives additional radiation exposure to patient. Other method is monitoring from patient surface for example by using surface marker or range image camera. On the other hand, there is range image camera which has been popular in gaming called Kinect camera. Kinect camera has infrared source and range image camera which can produce point distribution. But, to detect patient position we should detect feature points from smooth surface. For that purpose, we have developed an automated approach for reconstructing mathematical surfaces of head phantom’s range images acquired by Kinect camera. Three hundred frames of head phantom’s range images were acquired by the Kinect camera which was placed 1 meter from the nose of a phantom with 45 degree angle with respect to a vertical line. Distance measured was then corrected by using a linear interpolation method to decrease errors between reference distance and measured distance, which were caused by lens distortion. The next step was to localize a region of interest (ROI) using a template matching technique in gray scale images taken by the Kinect camera. Since the distances measured at the same pixels of the range images may fluctuate with frame, a temporal filtering technique with Kalman filter and moving average filter was applied to reduce the fluctuation and noise. Then, the range images were resampled with an interval of 0.1 mm, followed by spatial filtering using bilateral filter to smooth while preserving edge of the image. The last step was the reconstruction of mathematical surfaces using NURBS with assumption that all of point’s weights have the same value. The parameters of NURBS were optimized by evaluating roughness and edge slope ratio of reconstructed mathematical surfaces with changing the degree of B-spline functions and an interval between control points. The roughness of the reconstructed surface was calculated by using standard deviation ratio of 10 by 10 ROI in original image and NURBS surface. The higher NURBS degree results had the lower roughness which means that the smoother surfaces were obtained. To quantify edge preservability, the ratio between edge slope of NURBS and original image (reffered to edge slope ratio: ESR) was calculated. Higher ESRs would represent the better edge preserving on the surface. The result shows that surface with a sampling interval 10 points, NURBS degree equal to 3 provided better surface with a roughness 0.0962 and a highest ESR of 0.6919. We concluded that the proposed approach for reconstructing mathematical surfaces can provide smooth surfaces, which may detect patient positioning errors by using the Kinect camera. |
format |
Theses |
author |
PUSPA LESTARI (NIM : 20214059), FAUZIA |
spellingShingle |
PUSPA LESTARI (NIM : 20214059), FAUZIA RECONSTRUCTION OF HEAD PHANTOM̉̉S RANGE IMAGE ACQUIRED BY KINECT CAMERA USING NURBS |
author_facet |
PUSPA LESTARI (NIM : 20214059), FAUZIA |
author_sort |
PUSPA LESTARI (NIM : 20214059), FAUZIA |
title |
RECONSTRUCTION OF HEAD PHANTOM̉̉S RANGE IMAGE ACQUIRED BY KINECT CAMERA USING NURBS |
title_short |
RECONSTRUCTION OF HEAD PHANTOM̉̉S RANGE IMAGE ACQUIRED BY KINECT CAMERA USING NURBS |
title_full |
RECONSTRUCTION OF HEAD PHANTOM̉̉S RANGE IMAGE ACQUIRED BY KINECT CAMERA USING NURBS |
title_fullStr |
RECONSTRUCTION OF HEAD PHANTOM̉̉S RANGE IMAGE ACQUIRED BY KINECT CAMERA USING NURBS |
title_full_unstemmed |
RECONSTRUCTION OF HEAD PHANTOM̉̉S RANGE IMAGE ACQUIRED BY KINECT CAMERA USING NURBS |
title_sort |
reconstruction of head phantomãâãâs range image acquired by kinect camera using nurbs |
url |
https://digilib.itb.ac.id/gdl/view/22115 |
_version_ |
1822019699994525696 |