Characterisation of variable focus liquid lens camera system for depth estimation of a moving object

Soon, Adrian Bee Tiong (2021) Characterisation of variable focus liquid lens camera system for depth estimation of a moving object. Doctoral thesis, Universiti Tun Hussein Onn Malaysia.


Download (401kB) | Preview
[img] Text (Copyright Declaration)
Restricted to Repository staff only

Download (255kB) | Request a copy
[img] Text (Full Text)
Restricted to Registered users only

Download (28MB) | Request a copy


Depth estimation of an object or a scene are used for the purpose of motion detection, obstacle detection, positioning, depth mapping, or 3D shape recovery. These capabilities can be applied in home, industry, medical, education and other areas of applications. There are different types of depth sensor based on different technology, which suit different kinds of applications. Depth sensors can be divided into active sensor that emits out energy signal and passive sensor that does not require emission of energy signal. Camera-based depth sensor such as stereo camera and monocular camera are passive sensor. Hence, they do not have external or mutual interference problem, no emission hazard, better object detectability, while having the advantage of visual information. Compared to monocular camera, depth sensing with stereo camera vision has longer depth range. However, stereo camera faces challenges from occlusion, radiometric distortion, depth discontinuity, homogenous regions, false boundary problem, and reflection issues. Depth estimation with monocular camera uses images acquired at different focus settings. This can be achieved by varying the lens’ position or the lens’ optical power. Past works on depth sensing with variable focus mechanically actuates the lens position. The moving of the lens position results in change of field of view or magnification in the images, a phenomenon known as lens breathing. Image stacks acquired with linear actuator lens needs to be aligned before being processed, which adds on the complexity of image alignment, processing time, and dependence on the accuracy of image alignment. The developed liquid lens monocular camera system for depth estimation showed successful depth estimation with depth from focus technique without the need for image alignment. Lens breathing is avoided by varying the thickness of the lens to change the focal length without affecting the field of view. This research characterises the liquid lens monocular camera for depth estimation of a moving object that utilizes liquid lens to eliminate lens breathing. The response time of the liquid lens monocular camera system to complete a successful image acquisition at each lens’ voltage change was vi 0.274 s. A function describing the relationship between the liquid lens’ voltage, liquid lens’ temperature and object distance is presented, based on experimental setup for object at 1 m to 8 m distance. In the second research studies, an object�based focus measure method based on the mean of sum of modified Laplacian (SML) of the edge and texture features of an object image area is presented. In the third research work, an automated depth estimation using liquid lens camera system is proposed. Based on the experiment for object distance range of 1 m to 8 m with depth resolution of 1 m and 1.5 m, the root-mean-square error (RMSE) for depth estimation of static object was 21%. Depth estimation of moving object shows standard deviation of the steady-state error of 0.78 m and the RMSE was 1.2 m. The estimated speed of the moving object was 0.47 m/s. Based on the results, the method accurately estimated depth for static object distance of 1 m to 5 m and for moving object was 1 m to 4 m.

Item Type: Thesis (Doctoral)
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7800-8360 Electronics
Divisions: Faculty of Electrical and Electronic Engineering > Department of Electrical Engineering
Depositing User: Mrs. Sabarina Che Mat
Date Deposited: 03 Feb 2022 02:21
Last Modified: 03 Feb 2022 02:21

Actions (login required)

View Item View Item