結(jié)合GLCM與三階張量建模的在線目標(biāo)跟蹤

金廣智; 石林鎖; 崔智高; 劉浩; 牟偉杰

doi:10.11999/JEIT151108

結(jié)合GLCM與三階張量建模的在線目標(biāo)跟蹤

doi: 10.11999/JEIT151108

基金項(xiàng)目:

國家自然科學(xué)基金(61501470)

計(jì)量
- 文章訪問數(shù): 1765
- HTML全文瀏覽量: 129
- PDF下載量: 317
- 被引次數(shù): 0
出版歷程
- 收稿日期: 2015-09-29
- 修回日期: 2016-03-03
- 刊出日期: 2016-07-19

Online Object Tracking Based on Gray-level Co-occurrence Matrix and Third-order Tensor

Funds:

The National Natural Science Foundation of China (61501470)

摘要

摘要: 為提高目標(biāo)跟蹤算法對(duì)多種目標(biāo)表觀變化場(chǎng)景的自適應(yīng)能力和跟蹤精度，論文提出一種結(jié)合灰度共生(GLCM)與三階張量建模的目標(biāo)優(yōu)化跟蹤算法。該算法首先提取目標(biāo)區(qū)域的灰度信息，通過GLCM的高區(qū)分度特征對(duì)目標(biāo)進(jìn)行二元超分描述，并結(jié)合三階張量理論融合目標(biāo)區(qū)域的多視圖信息，建立起目標(biāo)的三階張量表觀模型。然后利用線性空間理論對(duì)表觀模型進(jìn)行雙線性展開，通過在線模型特征值描述與雙線性空間的增量特征更新，明顯降低模型更新時(shí)的運(yùn)算量。跟蹤環(huán)節(jié)，建立二級(jí)聯(lián)合跟蹤機(jī)制，結(jié)合當(dāng)前時(shí)刻信息通過在線權(quán)重估計(jì)構(gòu)建動(dòng)態(tài)觀測(cè)模型，以真實(shí)目標(biāo)視圖為基準(zhǔn)建立靜態(tài)觀測(cè)模型對(duì)跟蹤估計(jì)動(dòng)態(tài)調(diào)整，以避免誤差累積出現(xiàn)跟蹤漂移，最終實(shí)現(xiàn)對(duì)目標(biāo)的穩(wěn)定跟蹤。通過與典型算法進(jìn)行多場(chǎng)景試驗(yàn)對(duì)比，表明該算法能夠有效應(yīng)對(duì)多種復(fù)雜場(chǎng)景下的運(yùn)動(dòng)目標(biāo)跟蹤，平均跟蹤誤差均小于9像素。
- 目標(biāo)跟蹤 /
- 灰度共生 /
- 三階張量 /
- 線性空間 /
- 在線跟蹤
Abstract: In order to improve the stability and accuracy of the object tracking under different conditions, an online object tracking algorithm based on Gray-Level Co-occurrence Matrix (GLCM) and third-order tensor is proposed. First, the algorithm extracts the gray-level information of target area to describe the two high discrimination features of target by GLCM, the dynamic information about target changing is fused by third-order tensor theory, and the third-order tensor appearance model of the object is constructed. Then, it uses bilinear space theory to expand the appearance model, and implements the incremental learning. Updating of model by online models characteristic value description, thus computation of the model updating is greatly reduced. Meanwhile, the static observation model and adaptive observation model are constructed, and secondary combined stable tracking of object is achieved by dynamic matching of two observation models. Experimental results indicate that the proposed algorithm can effectively deal with the moving object tracking on a variety of challenging scenes, and the average tracking error is less than 9 pixels.
- Object tracking /
- Gray-Level Co-occurrence Matrix (GLCM) /
- Third-order tensor /
- Linear space /
- Online tracking

HTML全文

參考文獻(xiàn)(20)

YANG H, SHAO L, ZHENG F, et al. Recent advances and trends in visual tracking: A review[J]. Neurocomputing, 2011, 74(18): 3823-3831.

HO J, LEE K C, YANG M H, et al. Visual tracking using learned linear subspaces[C]. IEEE Conference on Computer Vision and Pattern Recognition, Rhode Island, USA, 2012: 782-789.

LEE K C and KRIEGMAN D. Online learning of probabilistic appearance manifolds for video-based recognition and tracking[C]. IEEE Conference on Computer Vision and Pattern Recognition, Portland, USA, 2013: 852-859.

ROSS D, LIM J, LIN R S, et al. Incremental learning for robust visual tracking[J]. International Journal of Computer Vision, 2013, 77(3): 125-141.

LATHAUWER L, MOOR B, and VANDEWALLE J. On the best rank-1 and approximation fhigherorder tensors[J]. SIAM Journal of Matrix Analysis and Applications, 2000, 21(4): 1324-1342.

LI X, HU W, ZHANG Z, et al. Visual tracking via incremental Log-Euclidean Riemannian sub-space learning [C]. IEEE Conference on Computer Vision and Pattern Recognition, Portland, USA, 2013: 578-585.

LI X, HU W, ZHANG Z, et al. Robust visual tracking based on incremental tensor subspace learning[C]. IEEE International Conference on Computer Vision, Barcelona, Spain, 2012: 1008-1016.

鐘樺, 楊曉鳴, 焦李成. 基于多分辨率共生矩陣的紋理圖像分類[J]. 計(jì)算機(jī)研究與發(fā)展, 2014, 48(11): 1991-1999.

ZHONG Ye, YANG Xiaoming, and JIAO Licheng. Texture classification based on multiresolution co-occurrence matrix [J]. Journal of Computer Research and Development, 2014, 48(11): 1991-1999.

薄華, 馬縛龍, 焦李成. 圖像紋理的灰度共生矩陣計(jì)算問題的分析[J]. 電子學(xué)報(bào), 2006, 34(1): 155-158.

BO Hua, MA Fulong, and JIAO Licheng. Research on computation of GLCM of image texture[J]. Acta Electronica Sinica, 2006, 34(1): 155-158.

CLAUSI D A and DENG H. Design-based texture Feature using Gabor filters and co-occurrence probabilities[J]. IEEE Transactions on Image Processing, 2005, 14(7): 925-936.

吳光文, 張愛軍, 王昌明. 一種用于壓縮感知理論的投影矩陣優(yōu)化算法[J]. 電子與信息學(xué)報(bào)，2015, 37(7): 1681-1687. doi: 10.11999/JEIT141450.

WU Guangwen, ZHANG Aijun, and WANG Changming. Novel optimization method for projection matrix in compress sensing theory[J]. Journal of Electronics Information Technology, 2015, 37(7): 1681-1687. doi: 10.11999/JEIT 141450.

KHAN Z H and GUI Y H. Online domain-shift learning and object tracking based on nonlinear dynamic models and particle filters on Riemannian manifolds[J]. Computer Vision and Image Understanding, 2014, 15(6): 97-114.

齊苑辰, 吳成東, 陳東岳, 等. 基于稀疏表達(dá)的超像素跟蹤算法[J]. 電子與信息學(xué)報(bào), 2015, 37(3): 529-535. doi: 10.11999/ JEIT140374.

QI Yuanchen, WU Chengdong, CHEN Dongyue, et al. Superpixel tracking based on sparse representation[J]. Journal of Electronics Information Technology, 2015, 37(3): 529-535. doi: 10.11999/JEIT140374.

CHENG X, LI N, ZHOU T, et al. Object tracking via collaborative multi-task learning and appearance model updating[J]. Applied Soft Computing, 2015, 31: 81-90.

ZHONG W, LU H, and YANG M H. Robust object tracking via sparse collaborative appearance model[J]. IEEE Transactions on Image Processing, 2014, 23(5): 2356-2368.

WU Y, LIM J, and YANG M H. Online object tracking: A benchmark[C]. IEEE Conference on Computer Vision and Pattern Recognition, Portland, USA, 2013: 2411-2418.

相關(guān)文章

施引文獻(xiàn)

資源附件(0)

訪問統(tǒng)計(jì)