Hyperspectral Target Tracking Based on Deep Spectral-Ternary Concatenated Features

Xiaofang Pei; Haorui Zhang; Huilin Xia; Siqi Li; Weikun Xie; Dong Zhao

doi:10.3788/LOP240617

Laser & Optoelectronics Progress, Volume. 61, Issue 24, 2428006(2024)

Hyperspectral Target Tracking Based on Deep Spectral-Ternary Concatenated Features

Xiaofang Pei^1,2, Haorui Zhang^1,2, Huilin Xia³, Siqi Li⁴, Weikun Xie⁵, and Dong Zhao^1,2、*

¹School of Electronics & Information Engineering, Nanjing University of Information Science & Technology, Nanjing 210044, Jiangsu , China

²School of Electronics and Information Engineering, Wuxi University, Wuxi 214105, Jiangsu , China

³School of Measurement-Control Technology and Communications Engineering, Harbin University of Science and Technology, Harbin 150000, Heilongjiang , China

⁴School of Computer Science and Technology, Harbin University of Science and Technology, Harbin 150000, Heilongjiang , China

⁵The 58th Research Institute of China Electronics Technology Group Corporation, Wuxi 214072, Jiangsu , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(12)

Fig. 1. Overall flowchart of the HVT-DSTC algorithm

Download full size

Fig. 2. Dimension reduction algorithm experimental results. (a) Original hyperspectral images; (b) spectral curves; (c) dimension reduction result

Download full size

Fig. 3. Deep feature maps (14×14×64)

Download full size

Fig. 4. Process of obtaining STC features

Download full size

Fig. 5. SILTP feature maps

Download full size

Fig. 6. STC feature maps

Download full size

Fig. 7. DSTC feature maps

Download full size

Fig. 8. Qualitative analysis result. (a) Toy; (b) car; (c) hand; (d) paper

Download full size

Fig. 9. Tracking success and precision rates for various algorithms across all test sequences, illumination variations, and scale variations challenges. (a)~(b) All test sequences; (c)~(d) illumination variations sequences; (e)~(f) scale variations sequences

Download full size

Table 1. Precision rates of various algorithms underdifferent challenges

View table

Table 1. Precision rates of various algorithms underdifferent challenges

Sequence	Precision	Precision_IV	Precision_SC
Ours	0.847	0.860	0.862
SiamBAG	0.856	0.800	0.8481
DC-HVT	0.817	0.809	0.820
HOMG	0.840	0.831	0.847
MHT	0.845	0.781	0.8482
SiamOHOT	0.848	0.778	0.8480
SiamCar	0.830	0.773	0.820
SiamBan	0.807	0.750	0.818
SiamFC	0.769	0.716	0.748

Table 2. Success rates of various algorithms underdifferent challenges

View table

Table 2. Success rates of various algorithms underdifferent challenges

Sequence	Success	Success_IV	Success_SC
Ours	0.623	0.607	0.629
SiamBAG	0.632	0.534	0.622
DC-HVT	0.600	0.598	0.612
HOMG	0.598	0.518	0.592
MHT	0.584	0.508	0.582
SiamOHOT	0.636	0.519	0.628
SiamCar	0.611	0.530	0.595
SiamBan	0.568	0.493	0.581
SiamFC	0.502	0.402	0.485

Table 3. Ablation study precision and success rates
View table
Table 3. Ablation study precision and success rates
Method PCA BM D-f DTC DSTC Precision Success
i √ × × × √ 0.835 0.617
ii √ × √ × × 0.714 0.513
iii × √ √ × × 0.732 0.529
iv √ × × √ × 0.802 0.598
v × √ × √ × 0.820 0.611
Ours × √ × × √ 0.847 0.623

Tools

Get Citation

Copy Citation Text

Xiaofang Pei, Haorui Zhang, Huilin Xia, Siqi Li, Weikun Xie, Dong Zhao. Hyperspectral Target Tracking Based on Deep Spectral-Ternary Concatenated Features[J]. Laser & Optoelectronics Progress, 2024, 61(24): 2428006

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites