Laser & Optoelectronics Progress, Volume. 59, Issue 16, 1611007(2022)
Lightweight Infrared Small-Target Detection Algorithm
Figures & Tables(7)
Fig. 3. Comparison of different feature fusion methods. (a) Simple fusion structure; （b） structure of ACM
Fig. 4. Data set images of different background examples. (a) Sky background; (b) architectural background; (c) ground background; (d) sea background
Fig. 5. Comparison of detection results of different algorithms. (a) Improved Tophat; (b) PSTNN; (c) Yolo-v4; (d) proposed algorithm
Table 1. Backbone structure of network
View tableTable 1. Backbone structure of network
Input Operator Expand size Out size SE NL Step 4162*1 con2d， 3*3 8 - HS 2 2082*8 bneck1， 3*3 32 16 √ RE 2 1042*16 bneck2， 3*3 72 24 - RE 2 522*24 bneck3， 3*3 88 24 - RE 1 522*24 bneck4， 3*3 96 40 √ HS 2 262*40 bneck5， 3*3 120 40 √ HS 1
Table 2. Comparison of detection indicators of different algorithms
View tableTable 2. Comparison of detection indicators of different algorithms
Algorithm Detection rate /% False alarm rate /% mIoU Board latency /ms Improved Tophat 93.5 47.2 0.35 100 PSTNN 94.4 19.4 0.51 500 Yolo-V4 91.6 9.3 / 150 Proposed+simple fusion 92.6 5.5 0.65 45 Proposed +ACM 93.5 4.7 0.68 50
Tools
Get Citation
Copy Citation Text
Qi Zhang, Hongtai Zhu, Hu Cheng, Jun Zhang, Ye Zhang. Lightweight Infrared Small-Target Detection Algorithm[J]. Laser & Optoelectronics Progress, 2022, 59(16): 1611007
Paper Information
Category: Imaging Systems
Received: Nov. 24, 2021
Accepted: Feb. 17, 2022
Published Online: Aug. 8, 2022
The Author Email: Hongtai Zhu (627242573@qq.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20