Article

Article title UNIVERSAL ROBUST ALGORITHM FOR DETECTION OF FEATURES ON A SCENE
Authors K.E. Rumyantsev, D.A. Petrov
Section SECTION III. CALCULATION ALGORITHMS AND SYSTEM DESIGN
Month, Year 08, 2015 @en
Index UDC 621.391
DOI
Abstract It is known that acquisition of the movement parameters and the current location of the autonomous mobile robot is possible by means of binocular vision systems usage and in some stages of processing require detection of feature points on scenes in the images of two cameras. For stable detection of feature points on scene image it`s advisable to use a detector that is resistant to changes in the conditions of observation on three-dimensional scenes. The second requirement for feature detector is possibility of various feature configurations detected by image processing algorithm. Finally, the structure of the feature detector must be invariant to the angle of feature rotation on the image. The premise for the synthesis of robust universal detector is an image on the surface of the CCD by size of 3x3 pixels. Each pixel of set contains the information about the sig- nal intensity accumulated during the observation period. Signal and noise reference sample are subsets of analyzed area. The task contain a nuisance parameter – unknown variance of the intensity. Informative parameter in task is the mean value of the signal pixels intensity. Authors proved existence of uniform most powerful invariant algorithm based on the t-test. Decision rule for detection of image features on 3x3 pixel patch was found. Simulation allowed to prove stability of the proposed feature point detection algorithm to change of mean value and standard deviation of background pixels intensity. Study also proved invariance of detector to the angle of the image rotation. Versatility of detection algorithm determined only by the number of pixels in the signal sample. Uniqueness of detected features determined by formation of support and analyzed samples. Article contain formulas for the selection of the threshold constants values that provide a predetermined level of false alarms rate, with regardless of the a priori unknown values of mean and standard deviation of the background pixels intensities. This demonstrates the practically important property of decision rule stability. Authors obtained equations that allow to assess the effectiveness of the robust feature detector. Study shown that the probability of signal detection for a given signal/noise ratio is same for features with the number of pixels of signal 1 and 8, 2 and 7, 3 and 6, 4 and 5. At a given signal/noise ratio and locked threshold constant value allocation of features that contain analyzed sample with 4 (or 5) signal pixel maximizes probability of signal detection for 3x3 analyzed area. Simulation in MATLAB package confirmed resistance of robust detector to change of the mean value and standard deviation of background pixel intensity. Thus, the study allowed to synthesize robust feature detection algorithm for a wide class of image features resistant to varying conditions of observation on three-dimensional scenes.

Download PDF

Keywords Robust detector; computer vision System; image processing; selection of versatile features of the image; universal algorithms; the stability of the decision rule; detection efficiency; change in the conditions of observati
References 1. Kemurdzhian A.L., Gromov V.V., Kazhukalo I.F. i dr. Mobil'nye roboty [Mobile robots], Ed. by A.L. Kemurdzhiana. Moscow: Mashinostroenie, 1993, 400 p.
2. Cheng Y., Goguen J., Johnson A., Leger C., Matthies L., San Martin M., Willson R. The Mars exploration rovers descent image motion estimation system, IEEE Intelligent Systems, 2004, Vol. 19, pp. 13-21.
3. Bailey T. Mobile Robot Localisation and Mapping in Extensive Outdoor Environments, Philosophy, 2002, Vol. 31, No. 8, pp. 212.
4. Nistйr D., Naroditsky O., Bergen J. Visual odometry for ground vehicle applications, J. F. Robot, 2006, Vol. 23, No. 1, pp. 3-20.
5. Roumeliotis S.I., Johnson A.E., Montgomery J.F. Augmenting inertial navigation with image-based motion estimation, Robotics and Automation, 2002. Proceedings. ICRA ’02. IEEE International Conference on, 2002, Vol. 4, pp. 4326-4333.
6. Lowe D.G. Distinctive image features from scale-invariant keypoints, International Journal of ComputerVision, 2004, Vol. 60, No. 2, pp. 91-110.
7. Herbert Bay, Andreas Ess, Tinne Tuytelaars, Luc Van Gool. SURF: Speeded Up Robust Features // Computer Vision and Image Understanding (CVIU). – 2008. – Vol. 110, No. 3. – P. 346-359.
8. Kravtsov S., Rumyantsev K. Positioning Autonomous Mobile Robot Based on Measurements Onboard Digital Stereo Vision System, International Journal of Robotics Applications and Technologies (IJRAT), Jule–December 2014, Vol. 2, No. 2, pp. 37-77.
9. Rumyantsev K.E., Kravtsov S.V. Strategiya pozitsionirovaniya mobil'nogo robota po dannym izmereniy bortovoy sistemy tsifrovogo stereozreniya [The positioning strategy of a mobile robot according to measurements on-Board system digital stereovision], Problemy upravleniya i modelirovaniya v slozhnykh sistemakh: Trudy XV Mezhdunarodnoy konferentsii (25-28 iyunya 2013 g. Samara, Rossiya) [Problems of control and modeling in complex systems: Proceedings of the XV International conference (25-28 June 2013 Samara, Russia)]. Samara: Samarskiy nauchnyy tsentr RAN, 2013, pp. 609-614.
10. Rumyantsev K., & Kravtsov S. Analysis of measuring space digital television stereoscopic systems. Point and interval estimation of point coordinates three–dimensional scenes, Electrical Engineering and Information Complexes and Systems, 2014, pp. 38-48.
11. Kravtsov S.V., Rumyantsev K.E. Spektral'nyy analiz oshibok izmereniy binokulyarnoy sistemy tekhnicheskogo zreniya v zadache obosnovaniya ee parametrov dlya mobil'nogo robota [Spectral analysis of measurement errors of the binocular vision system in the task of justification of its parameters for a mobile robot], Problemy upravleniya i modelirovaniya v slozhnykh
sistemakh (PUMSS-2014). Trudy XVI Mezhdunarodnoy konferentsii (30 iyunya–3 iyulya 2014
g, Samara, Rossiya) [Problems of control and modeling in complex systems (UMSS-2014). Proceedings of the XVI International conference (30 June–3 July 2014, Samara, Russia)]. Samara: Samarskiy nauchnyy tsentr RAN, 2014, pp. 621-626.
12. Rumyantsev K.E., Kravtsov S.V. Povyshenie dostovernosti vydeleniya osobennostey stseny na tsifrovykh izobrazheniyakh v usloviyakh intensivnykh nekorrelirovannykh shumov [Improving the validity of selection features scenes of digital images in conditions of intensive noise uncorrelated], ES-FM-2014-011. Fiziko-matematicheskie metody i informatsionnye tekhnologii v estestvoznanii, tekhnike i gumanitarnykh naukakh: sbornik materialov mezhdunarodnogo nauchnogo e-simpoziuma. Rossiya, g. Moskva, 27-28 dekabrya 2014 g. [ES-FM-2014-011.
Physico-mathematical methods and informational technologies in science, technology and the Humanities: proceedings of the international scientific e-Symposium. Russia, Moscow, 27-28 December 2014]. Kirov: MTsNIP, 2015, pp. 90-100.
13. Kravtsov S.V., Rumyantsev K.E. Issledovanie parametricheskoy svyazi bortovoy binokulyarnoy sistemy tekhnicheskogo zreniya i dinamiki mobil'nogo robota [A parametric study of the connection side of the binocular vision system and the dynamics of the mobile robot], Nauchno-tekhnicheskaya konferentsiya «Tekhnicheskoe zrenie v sistemakh upravleniya – 2014». 18−20 marta 2014 g. Tezisy dokladov [Scientific-technical conference "Technical vi-
sion in control systems – 2014". 18-20 March 2014 Abstracts], pp. 4-6.
14. Rumyantsev K.E., Petrov D.A. Informativnost' polutonovykh izobrazheniy prirodnykh landshaftov [Informative gray-scale images of natural landscapes], Fundamental'nye issledovaniya [Fundamental Research], 2015, No. 5, pp. 329-334.
15. Kravtsov S.V., Rumyantsev K.E. Issledovanie effektivnosti analiza stseny detektorom tochechnykh osobennostey v usloviyakh nekorrelirovannykh statsionarnykh shumov [Study of efficiency analysis of the detector of point features in the conditions of uncorrelated stationary noise], Nauchno-tekhnicheskaya konferentsiya «Tekhnicheskoe zrenie v sistemakh upravleniya
– 2014». 18–20 marta 2014. Tezisy dokladov [Scientific-technical сonference "Technical vision in control systems – 2014". March 18-20, 2014. Abstracts], pp. 16-18.
16. Rumyantsev K.E., Petrov D.A. Vydelenie tochechnykh osobennostey na trekhmernoy stsene pri izmenyayushchikhsya usloviyakh nablyudeniya [The selection of point features on a three-dimensional scene under varying conditions of observation], ES-FM-2014-011. Fiziko-matematicheskie metody i informatsionnye tekhnologii v estestvoznanii, tekhnike i
gumanitarnykh naukakh: sbornik materialov Mezhdunarodnogo nauchnogo e-simpoziuma. Rossiya, g. Moskva, 27-28 dekabrya 2014 g. [ES-FM-2014-011. Physico-mathematical methods and informational technologies in science, technology and the Humanities: proceedings of the International scientific e-Symposium. Russia, Moscow, 27-28 December 2014]. Kirov: MTsNIP, 2015, pp. 114-127.
17. Rumiantsev Konstantin, Petrov Dmitry. Detection of Feature Points On The Three-Dimensional Scene Under Varying Conditions Of Observation, 34th Chinese Control Conference (CCC2015). Hangzhou, China, July 28 to 30, 2015.
18. Prokof'ev V.N., Rumyantsev K.E. Invariantnye algoritmy obnaruzheniya signalov pri apriornoy neopredelennosti pomekhovoy obstanovki [Invariant algorithms of detection of signals under a priori uncertainty jamming environment]. Taganrog: TRTI, 1990, 45 p.
19. Bol'shev L.N., Smirnov N.V. Tablitsy matematicheskoy statistiki [Tables of mathematical statistics]. Moscow: Nauka. Glavnaya redaktsiya fiziko-matematicheskoy literatury, 1983, 416 p.
20. Sravnitel'nyy analiz effektivnosti algoritmov vydeleniya tochechnykh osobennostey pri izmenyayushchikhsya usloviyakh nablyudeniya [Comparative analysis of efficiency of algorithms of allocation of point features under varying conditions of observation], Sbornik materialov mezhdunarodnykh nauchnykh e-simpoziumov: Tekhnicheskie i estestvennye nauki: teoriya i praktika. Rossiya, g. Moskva, 27–28 marta 2015 g. [Proceedings of the international scientific e-workshops: Technical and natural science: theory and practice. Russia, Moscow, 27-28 March 2015], Ed. by K.E. Rumyantseva. Kirov: MTsNIP, 2015, pp. 123-141.

Comments are closed.