The detection of rectangular objects is an important part of document recognition and workflow automation. At the same
time, outlier noise may appear during the detection process – falsely determined coordinates of the document location,
negatively affecting the overall quality of object detection and further processing. The task of filtering such noise
when detecting a flat rectangular object on frames of a video sequence by constructing a single coordinate system is
posed. The proposed method is based on the coordination of the graph of estimated projective transformations of frames
with templates and frames among themselves. In the experiments used, simulated outlier applied to some frames of the
video sequence. To assess the quality of the algorithm, the results of object location before and after the module
matching the graph with the true location of the object are compared. This assessment is based on the open MIDV-500
database of video filming of identity documents. This data set demonstrates a reduction in the cumulative error compared
to the results before using graph matching in calculating projective transformations.
graph adjustment, projective transform, common coordinate system, video stream, object detection, cumulative error
- Savelyev B.I., Mamay B., Nikolaev D.P., Arlazarov V.L., Bulatov K.B., Skoryukina N.S. Metod soglasovaniy grafa proektivnych preobrazovanii dly zadachi panoramirovaniy obyektov. [A method of projective transformations graph adjustment for panorama stitching problem for images of planar objects]. Proceedings ISA RAS. 2018. V. 68. № S1. P. 124–133 (in Russian)
- Hartshorn R. Osnovu proektivnoy geometrii [Fundamentals of projective geometry]. M.: Mir, 1970. 160 p. (in Russian).
- Babu N., Soumya A. Character recognition in historical handwritten documents. Proceedings of the 2019 IEEE International Conf. Communic. and Signal Processing. 2019. P. 299–304.
- Fan B., Qingqun K., Tomasz T., Zhiheng W. Receptive fields selection for binary feature description. IEEE Transactions on Image Processing. 2014. V. 23. № 6. P. 2583–2595.
- Iglewicz B., Hoaglin D. How to Detect and Handle Outliers. The ASQC Basic References in Quality Control: Statistical Techniques. Eds F. Edward, P. Mykytka. 1993. 77 p.
- Kelson R.T., Andre M.S., Adelardo A.D. Optical Flow Using Color Information. ACM New York. NY, USA. 2008. P. 5–10.
- Lukoyanov A.S., Nikolaev D.P., Konovalenko I.A. Modification of YAPE keypoint detection algorithm for wide local contrast range image. Information technologies and nanotechnology. 2018. P. 1193–1204.
- Martin A.F., Robert C.B. Random sample consensus: A paradigm for model fitting with applications to image analysis and automated cartography. Comm. of the ACM. 1981. V. 24. P. 381–395.
- Moreno-García C.F., Elyan E., Jayne C. New trends on digitisation of complex engineering drawings. Neural Computing and Applications. 2019. V. 31 (6). P. 1695–1712.
- Newman P., Ho K. SLAM-loop closing with visually salient features. IEEE Proc. of International Conference on Robotics and Automation. 2005. P. 635–642.
- Novack G.D., Lim M.C. Retinal Detachment: Patient Perspective and Electronic Health Records. American Journal of Ophthalmology. 2019. V. 208. P. 64–67.
- Olson D. L. Delen D. Advanced Data Mining Techniques. Springer, 1st edition. 2008. 138 p.
- Skoryukina N., Nikolaev D., Arlazarov V. 2D art recognition in uncontrolled conditions using one-shot learning. ICMV. 2018. P. 1–8.
- Skoryukina N., Nikolaev D.P., Sheshkus A., Polevoy D. Real time rectangular document detection on mobile devices. In Seventh International Conference on Machine Vision. 2015. V. 9445. P. 1–6.
- Turcot P., Lowe D.G. Better matching with fewer features: The selection of useful features in large database recognition problems. Computer Vision Workshops (ICCV Workshops). 2009. P. 2109–2116.