• 2022 (Vol.36)
  • 1990 (Vol.4)
  • 1989 (Vol.3)
  • 1988 (Vol.2)
  • 1987 (Vol.1)

A linear regression method robust to extreme stationary clutter

© 2020 D. A. Bocharov

Institute for Information Transmission Problems RAS, 127051 Moscow, Bolshoy Karetny per., 19, Russia

Received 30 Sep 2019

The study deals with the task of two-dimensional linear regression of noisy data. The data is represented by pairs of coordinates provided by an image-based object detector and can include outliers along with the regular data. As for main outliers model a stationary clutter is considered, i.e. the coordinates of outliers are close to a fixed but undefined point. The paper introduces a linear regression method that is robust to extreme, i.e. the number of outliers exceeds the number of regular data, stationary noise. The proposed method is based on Hough-analysis of input data coordinate histogram and consists of two sequential stages: estimation of angle and shift parameters. Algorithm based on the proposed method is used in the automatic vehicle classifier for wheels trajectory localization. The precision of proposed algorithms and Welsch m-estimator on simulated and real data is investigated. The conducted experiments show that the algorithms based on the proposed method with the same computational complexity have higher precision of wheels trajectory linear model regression in the cases of presence of extreme stationary clutter.

Key words: grobust linear regression, outliers, object detection, automatic vehicle classifier

DOI: 10.31857/S0235009220010059

Cite: Bocharov D. A. Metod lineinoi regressii, ustoichivyi k ekstremalnym statsionarnym pomekham [A linear regression method robust to extreme stationary clutter]. Sensornye sistemy [Sensory systems]. 2020. V. 34(1). P. 44–56 (in Russian). doi: 10.31857/S0235009220010059

References:

  • Grigoryev A., Gladilin S., Khanipov T., Koptelov I., Bocharov D., Matsnev D. Arkhitektura sistemy detekcii i klassifikacii avtomobilei sredstvami tehnicheskogo zreniya v estestvennykh usloviyakh [Architecture of a system for computer vision-based vehicle detection and classisication under natural conditions], Sensornye sistemy [Sensory systems]. 2017. V. 31. № 1. P. 72–84 (in Russian).
  • Ershov E., Asvatov E., Nikolaev D.P. Robastnaya ortogonalnaya lineinaya regressiya dlya malomernykh gistogramm [Robust orthogonal linear regression on histogram in small-dimensional spaces]. Sensornye sistemy [Sensory systems]. 2017. V. 31. № 4. P. 331–342 (in Russian).
  • Koptelov I., Grigoryev A., Khanipov T., Emelyanov S., Nikolaev D. Maket avtomaticheskogo klassifikatora transportnykh sredstv [A model of automatic vehicle classifier]. 2014 ITiS [ITaS] IITP RAS P. 350–356 (in Russian).
  • Amin A., Fischer S. A document skew detection method using the Hough transform Pattern Analysis & Applications. 2000. T. 3. № 3. C. 243–253.
  • Baird H.S. The skew angle of printed documents Proceedings of SPSE 40th Symposium of Hybrid Imaging Systems. 1987. C. 739–743.
  • Bezmaternykh P., Nikolaev D., Arlazarov V. Textual blocks rectification method based on fast Hough transform analysis in identity documents recognition Tenth International Conference on Machine Vision (ICMV 2017). International Society for Optics and Photonics, 2018. T. 10696. C. 1069606.
  • Grigoryev A., Bocharov D., Terekhin A., Nikolaev D. Vision-based vehicle wheel detector and axle counter, Proc. 29th Europ. Conf. Model. Simulat., 2015. C. 521–526.
  • Huber P. J. Robust statistics. Springer Berlin Heidelberg, 2011. C. 1248–1251.
  • Hull J. J. Document image skew detection: Survey and annotated bibliography Document Analysis Systems II. 1998. C. 40–64.
  • Khanipov T., Koptelov I., Grigoryev A., Kuznetsova E., Nikolaev D. Vision-based industrial automatic vehicle classifier. Seventh International Conference on Machine Vision (ICMV 2014). International Society for Optics and Photonics, 2015. T. 9445. C. 944511.
  • Li S., Shen Q., Sun J. Skew detection using wavelet decomposition and projection profile analysis Pattern recognition letters. 2007. T. 28. № 5. C. 555–562.
  • Limonova E., Bezmaternykh P., Nikolaev D., Arlazarov V. Slant rectification in Russian passport OCR system using fast Hough transform Ninth International Conference on Machine Vision (ICMV 2016). International Society for Optics and Photonics, 2017. T. 10341. C. 103410.
  • Nikolaev D., Karpenko S., Nikolaev I., Nikolaev P. Hough transform: Understimated tool in the computer vision field, Proc. 22th Europ. Conf. Model. Simulat., 2008. C. 238–246.
  • Papandreou A., Gatos B. A novel skew detection technique based on vertical projections 2011 International Conference on Document Analysis and Recognition. IEEE, 2011. C. 384–388.
  • Postl W. Method for automatic correction of character skew in the acquisition of a text original in the form of digital scan results. 1988. Patent No. 4,723,297. U.S.A.
  • Rousseeuw P. J. Least median of squares regression Journal of the American statistical association. 1984. T. 79. № 388. C. 871–880.
  • Rousseeuw P. J., Leroy A. M. Robust regression and outlier detection. John wiley & sons, 2005. T. 589.
  • Rousseeuw P. J., Van Driessen K. Computing LTS regression for large data sets Data mining and knowledge discovery. 2006. T. 12. № 1. C. 29–45.
  • Srihari S. N., Govindaraju V. Analysis of textual images using the Hough transform Machine vision and Applications. 1989. T. 2. № 3. C. 141–153.
  • Stahlberg F., Vogel S. Document skew detection based on hough space derivatives 2015 13th International Conference on Document Analysis and Recognition (ICDAR). IEEE, 2015. C. 366–370.