1. UKM Learning and Research Repository
2. Master Thesis / Tesis Master
3. Faculty of Engineering and Built Environment / Fakulti Kejuruteraan dan Alam Bina

Title:	Laser pointer detection for visual feedback in gesture assisted teleconsultation
Authors:	Naireen Imtiaz (P73129)
Supervisor:	Mohd. Marzuki Mustafa, Prof. Dato' Ir. Dr.
Keywords:	Teleconsultation Laser communication systems
Issue Date:	17-Apr-2017
Description:	Teleconsultation is one of the important practices of telemedicine which utilizes the opinion of a medical expert from a remote place to guide the primary healthcare provider in giving treatment to the patient. One of the ways to improve the teleconsultation system is to equip the remote specialist with a laser pointer which can be used as means of gesture. As such, accurate detection of the laser pointer’s point and tracking are crucial in implementing such a system as they rely on visual feedback for control. The main issue in laser point detection and tracking is due to camera saturation that makes the laser pixels appear white and are not always distinguishable from glare and other bright regions of the image frames. Typically, such a problem can be solved effectively by using additional hardware which are costly and would benefit from an improved preliminary detection method. Consequently, this research proposes a software based solution with an aim of developing an efficient algorithm that is able to detect and track the laser point without the use of additional hardware and done in a less controlled environment. Development of the laser pointer feedback algorithm is divided into two parts a) detecting laser point in static images and b) tracking the laser point in consecutive video frames. To make the algorithm robust, the laser point detection is done on images without using calibration. This process involves three steps which are i) obtaining the region of interest (ROI) via template matching approach using laser pointer point images taken at a predetermined distance. ii) narrowing down the number of ROI by using intensity thresholding and iii) applying Linear Regression to establish cross sectional intensity profile similarity for classification of ROI as laser. To track the laser pointer in video frames, windowing method is used with the detection algorithm. To improve tracking and increase robustness, model based Kalman filter is applied and trend based double exponential smoothing is also used to correct noisy sensing and predict laser path values overcome missing detection. A hybrid of Kalman filter and double exponential smoothing is applied to accommodate features of both model based as well as trend based tracking. The tracking time is optimized using partitioning method and limiting ROI detection to single step intensity thresholding. Obtained results showed that the developed algorithm gave better results when compared with the commonly used method of thresholding and blob evaluation. For profile matching, various similarity measures were tested and the findings suggested that linear regression gave the best result. The tracking hybrid algorithm can improve the path identification in noisy data. Kalman filter is able to provide best path prediction results. In conclusion, this research has successfully developed an effective feedback algorithm for tele-consultation setup that involves laser gesture. The laser pointer detection was accurate which enabled the tracking algorithm to be implemented on an actual setup. Furthermore, the algorithm does not require any calibration or hardware enhancements.,Certification of Master's/Doctoral Thesis" is not available
Pages:	107
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Engineering and Built Environment / Fakulti Kejuruteraan dan Alam Bina

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.