Spatial Regularization Correlation Filtering Tracking via Deformable Diversity Similarity

Ning Mao; Dedong Yang; Yong Li; Yajun Han

doi:10.3788/AOS201939.0415002

Journals >Acta Optica Sinica >Volume 39 >Issue 4 >Page 0415002 > Article

Acta Optica Sinica
Vol. 39, Issue 4, 0415002 (2019)

Spatial Regularization Correlation Filtering Tracking via Deformable Diversity Similarity

Ning Mao, Dedong Yang^*, Yong Li, and Yajun Han

Author Affiliations

School of Artificial Intelligence, Hebei University of Technology, Tianjin 300130, China

show less

DOI: 10.3788/AOS201939.0415002 Cite this Article Set citation alerts

Ning Mao, Dedong Yang, Yong Li, Yajun Han. Spatial Regularization Correlation Filtering Tracking via Deformable Diversity Similarity[J]. Acta Optica Sinica, 2019, 39(4): 0415002 Copy Citation Text

show less

Attribute	Algorithm
Attribute	DDST	SRDCF	Staple	MEEM	MUSTer	Siamfc3s	DCFNet	KCF	DSST	TGPR
Accuracy	0.825	0.793	0.784	0.780	0.774	0.771	0.761	0.696	0.680	0.643
Success rate	0.625	0.599	0.581	0.529	0.577	0.582	0.580	0.477	0.513	0.458

Table 1. Results of 10 algorithms on OTB-100 dataset

Attribute	Algorithm
Attribute	DDST	SRDCF	Staple	MEEM	MUSTer	Siamfc3s	DCFNet	KCF	DSST	TGPR
IV	0.821	0.792	0.791	0.740	0.782	0.736	0.722	0.719	0.721	0.633
OPR	0.796	0.748	0.738	0.794	0.744	0.756	0.755	0.677	0.644	0.642
OCC	0.779	0.735	0.726	0.739	0.734	0.722	0.755	0.630	0.597	0.594
SV	0.803	0.751	0.727	0.735	0.710	0.735	0.738	0.633	0.638	0.599
DEF	0.768	0.734	0.748	0.754	0.689	0.690	0.671	0.617	0.542	0.630
MB	0.790	0.767	0.707	0.727	0.678	0.705	0.665	0.601	0.567	0.529
FM	0.778	0.769	0.697	0.750	0.683	0.743	0.671	0.621	0.552	0.533
OV	0.701	0.597	0.661	0.678	0.591	0.669	0.727	0.501	0.481	0.493
IPR	0.766	0.745	0.770	0.792	0.773	0.742	0.738	0.701	0.691	0.659
BC	0.846	0.775	0.766	0.742	0.784	0.690	0.741	0.713	0.704	0.593
LR	0.774	0.765	0.695	0.808	0.747	0.900	0.726	0.671	0.649	0.622

Table 2. Tracking accuracy of 10 algorithms on 11 attribute sequences

Attribute	Algorithm
Attribute	DDST	SRDCF	Staple	MEEM	MUSTer	Siamfc3s	DCFNet	KCF	DSST	TGPR
IV	0.641	0.613	0.598	0.517	0.600	0.568	0.581	0.479	0.558	0.452
OPR	0.590	0.551	0.534	0.525	0.537	0.558	0.575	0.453	0.470	0.455
OCC	0.592	0.559	0.548	0.503	0.554	0.543	0.573	0.443	0.453	0.429
SV	0.604	0.562	0.525	0.469	0.512	0.552	0.565	0.394	0.468	0.404
DEF	0.562	0.544	0.554	0.489	0.524	0.506	0.497	0.436	0.420	0.455
MB	0.620	0.594	0.546	0.554	0.544	0.550	0.544	0.459	0.468	0.429
FM	0.605	0.597	0.537	0.541	0.533	0.568	0.541	0.459	0.447	0.420
OV	0.539	0.460	0.481	0.483	0.469	0.506	0.557	0.393	0.385	0.373
IPR	0.564	0.544	0.552	0.528	0.551	0.557	0.557	0.469	0.502	0.462
BC	0.635	0.583	0.574	0.517	0.581	0.523	0.569	0.498	0.522	0.428
LR	0.526	0.514	0.396	0.382	0.415	0.618	0.503	0.290	0.370	0.344

Table 3. Tracking success rate of 10 algorithms on 11 attribute sequences

Sequence	Frame	Algorithm
Sequence	Frame	TGPR	KCF	Siamfc3s	SRDCF	Staple	MUSTer	DDST	DCFNet
Lemming	261	/	/	3.6851	/	/	3.2016	2.1213	/
	353	/	/	56.7350	/	/	14.2302	15.9531	/
	556	/	/	6.7107	/	/	12.1357	13.5093	/
Box	27	/	/	/	/	/	3.6401	3.3541	2.7174
	311	/	/	/	/	/	3.1623	0.7071	3.0192
	469	/	/	/	/	/	45.0694	4.5000	5.5862
Human3	31	49.9710	56.7671	15.7467	3.9051	4.6144	3.5355	3.9051	/
	57	94.9710	100.6504	209.9629	16.6208	17.0386	15.7003	27.1662	/
	95	163.0616	170.7513	260.5964	84.1576	86.0142	166.1246	1.0000	/

Table 4. Center errors of Lemming, Box and Human3 sequence with different algorithms

Frame	Algorithm
Frame	DDST	MUSTer	KCF
10	62.3939	53.1507	55.5428
42	7.0178	222.6398	226.2576
89	15.9765	10.8096	420.0583

Table 5. Center error of MotorRolling sequence

Sequence	Frame	Algorithm
Sequence	Frame	MUSTer	TGPR	Siamfc3s	MEEM	Staple	DCFNet	DDST	KCF	DSST
Board	67	239.8101	284.4558	294.7367	95.4303	6.8969	2.8056	1.4142	/	/
	553	230.2411	272.9528	288.4955	44.6796	11.2097	7.8479	22.5000	/	/
	623	225.7977	266.7260	287.3400	7.8743	25.8323	24.5449	4.9497	/	/
Couple	16	/	27.6993	/	/	/	1.6144	1.1180	4.0311	32.8824
	37	/	111.6756	/	/	/	0.6909	0.5000	35.2172	98.5203
	122	/	177.3186	/	/	/	16.5059	8.0623	77.1314	176.5708
Football	32	/	/	/	/	2.0616	0.6016	2.0616	1.0000	1.0000
	61	/	/	/	/	0.5000	1.7121	0.5000	3.0414	1.1180
	348	/	/	/	/	13.2857	53.8005	11.2805	49.2976	57.0285

Table 6. Center errors of Board, Couple and Football sequence with different algorithms

Ning Mao, Dedong Yang, Yong Li, Yajun Han. Spatial Regularization Correlation Filtering Tracking via Deformable Diversity Similarity[J]. Acta Optica Sinica, 2019, 39(4): 0415002

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information