Image Matching with Large Viewing Angle Based on Local Features

Pengtu Zhao; Feipeng Da

doi:10.3788/AOS201939.0510002

Acta Optica Sinica, Volume. 39, Issue 5, 0510002(2019)

Image Matching with Large Viewing Angle Based on Local Features

Pengtu Zhao^1,2、* and Feipeng Da^1,2、*

¹ Key Laboratory of Measurement and Control of Complex System of Engineering, Ministry of Education, Southeast University, Nanjing, Jiangsu 210096, China

² School of Automation, Southeast University, Nanjing, Jiangsu 210096, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(9)

Fig. 1. Flow chart of feature extraction

Download full size

Fig. 2. Schematic of spatial transformation parameters

Download full size

Fig. 3. View transformation and mask maps. (a) Transformation map; (b) mask map

Download full size

Fig. 4. Multi-view descriptor

Download full size

Fig. 5. Matching results of different algorithms. (a)(c) ASIFT algorithm, image pair of Freiburg_center1-4 and Graffiti1-4; (b)(d) proposed algorithm, image pair of Freiburg_center1-4 and Graffiti1-4

Download full size

Fig. 6. Recall-1-Precision matching results of different image pairs by different algorithms under image transformation. (a) Boat1-2; (b) Bark1-2; (c) Graffiti1-2; (d) Graffiti1-4; (e) Wall1-2; (f) Wall1-4

Download full size

Fig. 7. Matching results of different algorithms on partial images

Download full size

Fig. 8. Matching result of proposed algorithm on Graffiti1-6

Download full size

Table 1. Time schedule for feature point detection
View table
Table 1. Time schedule for feature point detection
Matching algorithm Detection complexity Feature point detection time /s Feature point matching time /s
SIFT 1 1.614 0.685
KAZE 1 1.424 0.942
ORB 1 1.621 0.846
ASIFT 26 4.242 6.621
Proposed 26 3.428 2.226