Laser & Optoelectronics Progress, Volume. 56, Issue 14, 141502(2019)
Research and Implementation of Binocular Distance Measurement System Based on Improved Scale-Invariant Feature Transform Algorithm with Parallel Acceleration
Figures & Tables(8)
Fig. 1. Principle of binocular distance measurement. Solid lines represent light path of left camera and dotted lines represent light path of right camera
Table 1. Comparison of time consumption, complexity, and error for four Gaussian blur calculation methods
View tableTable 1. Comparison of time consumption, complexity, and error for four Gaussian blur calculation methods
Method Time consumption/ms Complexity Error /% n = 11n = 21Original Gaussian blur 2452 8990 O (n ×n ×W ×H )0 Mean blur 589 2437 O (n ×n ×W ×H )0.04 Separation mean blur 230 435 O (2×n ×W ×H )0.04 Integral mean blur 32 34 O (W ×H )0.04
Table 2. Comparison of time consumption between proposed SIFT algorithm and previous work
View tableTable 2. Comparison of time consumption between proposed SIFT algorithm and previous work
Item Proposed algorithm Previous work in Ref. [23] Convolution core width 31×31 19×19 Image size /(pixel×pixel) 640×480 640×480 GPU hardware GT940M GTX560 Theoretical floating-point computing power /TFlops 1 1 Time consumption /ms 100 634
Tools
Get Citation
Copy Citation Text
Zhiqiang Zhang, Wenhua Shi. Research and Implementation of Binocular Distance Measurement System Based on Improved Scale-Invariant Feature Transform Algorithm with Parallel Acceleration[J]. Laser & Optoelectronics Progress, 2019, 56(14): 141502
Paper Information
Category: Machine Vision
Received: Jan. 25, 2019
Accepted: Feb. 21, 2019
Published Online: Jul. 12, 2019
The Author Email: Zhang Zhiqiang (zhangzhiqiang08@gmail.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20