Path: Top -> Journal -> Telkomnika -> 2019 -> Vol 17, No 1, February 2019

Stereo matching based on absolute differences for multiple objects detection

Journal from gdlhub / 2019-10-18 14:12:46
Oleh : Rostam Affendi Hamzah, Melvin Gan Yeou Wei, Nik Syahrim Nik Anwar, Telkomnika
Dibuat : 2019-05-14, dengan 1 file

Keyword : absolute differences, bilateral filter, computer vision, guided filter, stereo matching
Url : http://journal.uad.ac.id/index.php/TELKOMNIKA/article/view/9185
Sumber pengambilan dokumen : WEB

This article presents a new algorithm for object detection using stereo camera system. The problem to get an accurate object detion using stereo camera is the imprecise of matching process between two scenes with the same viewpoint. Hence, this article aims to reduce the incorrect matching pixel with four stages. This new algorithm is the combination of continuous process of matching cost computation, aggregation, optimization and filtering. The first stage is matching cost computation to acquire preliminary result using an absolute differences method. Then the second stage known as aggregation step uses a guided filter with fixed window support size. After that, the optimization stage uses winner-takes-all (WTA) approach which selects the smallest matching differences value and normalized it to the disparity level. The last stage in the framework uses a bilateral filter. It is effectively further decrease the error on the disparity map which contains information of object detection and locations. The proposed work produces low errors (i.e., 12.11% and 14.01% nonocc and all errors) based on the KITTI dataset and capable to perform much better compared with before the proposed framework and competitive with some newly available methods.

Beri Komentar ?#(0) | Bookmark

PropertiNilai Properti
ID Publishergdlhub
OrganisasiTelkomnika
Nama KontakHerti Yani, S.Kom
AlamatJln. Jenderal Sudirman
KotaJambi
DaerahJambi
NegaraIndonesia
Telepon0741-35095
Fax0741-35093
E-mail Administratorelibrarystikom@gmail.com
E-mail CKOelibrarystikom@gmail.com

Print ...

Kontributor...

  • , Editor: sustriani

Download...