Object recognition algorithms

Author(s) Collection number Pages Download abstract Download full text
Havrysh B. M., Tymchenko O. V., Кляп М. П. № 1 (83) 47-58 Image Image

This article discusses the algorithms for recognizing objects in the image, analyses the methods used in image processing, and describes the use of machine learning tools in working with images. The selection of specific methods is determined by the characteristics of the object to be recognized. To solve this problem, it is necessary to formulate the properties of the desired object and create a stable method for identifying objects that meet the specified parameters. The importance of finding the area of the desired object as soon as possible is also taken into account. This is possible only if a large number of elements are classified during the processing of each image.

There are many methods for recognizing objects in an image. Sometimes the task of recognition is set informally – the properties of the desired object are set without strict mathematical parameters. The main difficulty is that it is almost impossible to describe all the properties and these properties may not correspond to all objects. Therefore, in the process of mathematical formalization, simplifications are allowed, which in turn reduce the quality of the algorithm and reduce the accuracy.

As a result, one can say that in solving the problem of recognition it is necessary to find the optimal ratio of computational complexity and the desired accuracy.

The main criteria for the quality of symbols to solve a wide range of problems related to recognition, taking into account visual images, are the separate properties of symbols, as well as the difficulty of obtaining them. Analysis of existing algorithms for object recognition and methods of object detection showed that the main problem is low resistance to external conditions that complicate the quality of recognition: light level, image quality, image tilt. At the same time, algorithms that are resistant to external conditions (changing backgrounds, changes in lighting, moving shadows, etc.) are often demanding on hardware computing resources, which complicates their application in systems that work in real time. To solve this problem, it is necessary to find, summarize and formulate empirical observations in mathematical terms. That is to formalize the parameters of the desired object.

Keywords: pattern recognition, image processing, computer vision, machine learning.

doi: 10.32403/0554-4866-2022-1-83-47-58

  • 1. Limin, Xia. (2006). Vehicle Recognition Using Boosting Neural Network Classifiers. 2016 6th World Congress on Intelligent Control and Automation. Dalian, 9641–9644. doi: 10.1109/WCICA.2016.1713873 (in English).
  • 2. Durniak, B. V., Medykovskyi, M. O., & Tymchenko, O. V. (2020). Informatsiini tekhnologii poshuku ta zberihannia pro hrafichni yikh stmantyky. Lviv : Ukrainska akademiia drukarstva (in Ukrainian).
  • 3. Hsu, Y., Ciou, Y., & Perng, J. (2017). Object recognition system design in regions of interest based on AdaBoost algorithm. 2017 20th International Conference on Information Fusion (Fusion). Xi’an, 1–5. doi: 10.23919/ICIF.2017.8009787 (in English).
  • 4. Tsai, P. Y., Hsu, Y., Chiu, C., & Chu, T. (2017). Accelerating AdaBoost algorithm using GPU for multi-object recognition. 2015 IEEE International Symposium on Circuits and Systems (ISCAS). Lisbon, 738–741. doi: 10.1109/ISCAS.2015.7168739 (in English).
  • 5. Ge, J., Lu, D., & Fang, Y. (2020). A revised training mechanism for AdaBoost algorithm. 2010 IEEE International Conference on Software Engineering and Service Sciences. Beijing, 491–494. doi: 10.1109/ICSESS.2010.5552322 (in English).
  • 6. Zhang, P., & Yang, Z. (Jan. 2018). A Novel AdaBoost Framework With Robust Threshold and Structural Optimization. in IEEE Transactions on Cybernetics, 48, 1, 64–76. doi: 10.1109/TCYB.2016.2623900 (in English).
  • 7. Ren, J., Kehtarnavaz, N., & Estevez, L. Real-time optimization of Viola -Jones face detection for mobile platforms. 2018 IEEE Dallas Circuits and Systems Workshop: System-on-Chip - Design, Applications, Integration, and Software. Dallas, TX, 20108, 1–4. doi: 10.1109/DCAS.2008.4695921 (in English).
  • 8. Masek, J., Burget, R., Uher, V., & Güney, S. (2018). Speeding up Viola-Jones algorithm using multi-Core GPU implementation. 2018 36th International Conference on Telecommunications and Signal Processing (TSP). Rome, 808–812. doi: 10.1109/TSP.2013.6614050 (in English).
  • 9. Lienhart, R., Kuranov, A., & Pisarevsky, V. (2018). Empirical Analysis of Detection Cascades of Boosted Classifiers for Rapid Object Detection. MRL Technical Report (in English).
  • 10. Rosset, S., Zhu, Ji., & Hastie, T. (2004). Boosting as a Regularized Path to a Maximum Margin Classifier. Journal of Machine Learning Research, 5, 941–973 (in English).
  • 11. Yoav, Freund, & Robert, E. (September, 2019). Schapire. A Short Introduction to Boosting. Journal of Japanese Society for Artificial Intelligence, 14 (5):771–780 (in English).