Prawira, Winal and Sulistiyanti, Sri Ratna and Setyawan, FX. Arinto and Nasrullah, Emir (2017) The Detection of 3D Object Using a Method of a Harris Corner Detector and Lucas-Kanade Tracker Based on Stereo Image. Proceeding ICECOS 2017, 2 (1). pp. 163-166. ISSN 978-1-5386-2780-8-0

[img]
Preview
Text
ICECOS 2017.pdf

Download (766kB) | Preview
Official URL: https://ieeexplore.ieee.org/document/8167126

Abstract

This research proposes the use of Harris Corner Detector and Lucas-Kanade Tracker methods for the detection of 3D objects based on stereo image. The test image obtained from the results of capturing of the camera to the object of the form of tubes, balls, cubes, and 2D images. This research is the early step in the development of the ability of a computer vision to be able to mimic the performance of eye organs in humans in detecting an object. The detection step of the proposed method begins by determining the feature point on the image of the taking results of two cameras using the Harris Corner Detector. After the feature point of the two images obtained, then performed tracking feature point using the Lucas-Kanade Tracker method. In this research, the distance between the cameras used 10, 20, and 50 cm. The Effectiveness of the detection result of the proposed method is measured using the recall and precision parameter values obtained in the merged of the image. The proposed method gives a Recall value above 90% and a precision value above 50% for a distance of the cameras 10cm and 20cm for ball and tube objects. In the box object, the Recall value is 60% for a distance between the cameras 50cm and below 25% for a distance of the cameras 10cm and 20cm. The precision value for detection of the box object is very low, i.e. less than 25%.

Item Type: Article
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering
Depositing User: Dr. Sri Ratna Sulistiyanti
Date Deposited: 16 Jun 2020 07:02
Last Modified: 16 Jun 2020 07:02
URI: http://repository.lppm.unila.ac.id/id/eprint/22486

Actions (login required)

View Item View Item