Andreiev A. Object classification technique on aerial and space imagery under low separability of recognition features

Українська версія

Thesis for the degree of Doctor of Philosophy (PhD)

State registration number

0823U100884

Thesis Registration Form

0823U100884.pdf

Applicant for

Specialization

  • 172 - Електронні комунікації та радіотехніка

29-11-2023

Specialized Academic Board

ДФ 004

State Institution "Scientific Center for Aerospace Research of the Earth of the Institute of Geological Sciences of the National Academy of Sciences of Ukraine"

Essay

The study aimed to increase the accuracy of object classification on aerial and space imagery by developing a technique to increase the separability of recognition features. The dissertation research examines the role and place of the classification procedure in the tasks of remote sensing. An analysis of object classification methods on aerial and space images was conducted. It is substantiated that for most thematic tasks, it is appropriate to use supervised classification methods because they allow setting the properties of the object classes in the form of training samples set. The properties of the training samples set were analyzed. An analysis of approaches to processing the training sample was carried out. It is shown that the common disadvantage of the considered approaches is that they do not consider the factor of the separability. It is proposed to use a geospatial data cube to represent input data for classification. Data templates are presented for the selected thematic tasks. Methods of training sample clustering have been developed based on centroid methods of unsupervised classification. 2 methods have been developed: the method of forming training samples set from clusters of initial classes and the method of forming training samples set from the clusters centres of initial classes. A method for assessing the training sample separability has been developed. The algorithm of the method for assessing the separability of two separate classes of the training sample and the entire set as a whole is described in detail. The principles of forming a geospatial data cube, the methods of the training sample clustering, and the method for assessing the training sample are combined into object classification technique, which has two branches of application: reducing the training samples dimensions and clustering the training sample. The choice of the branch depends on the amount of input data. According to the results of experimental studies, the developed method demonstrated the classification accuracy enhancement in each of the 4 examples. The effectiveness of the method is confirmed in one of the examples by the increase of the overall accuracy indicators by 2% (from 91% to 93%) and the kappa index by 2% (from 87% to 89%); in the second example, an increase in the corresponding indicators by 4% (from 77% to 81%) and by 5% (from 66% to 71%); in the third - by an increase in the Pearson correlation coefficient by 28% (from 54% to 82%); in the fourth - by 20% (from 63% to 83%) and by 21% (from 60% to 81%) of the overall accuracy of classification and kappa index, respectively. Scientific novelty The first developed method for assessing the training sample separability during the supervised object classification on aerial and space images. The assessment considers a specific classification method, input data and its structure. The first developed object classification technique on aerial and space imagery under low separability of recognition features, which includes the application of one of the two developed methods depending on the amount of data. When there is an excess amount of data, the method of reducing the training sample size is used, and when the amount of data is limited, the method of the training sample clustering is used. The basis of the developed methods is the method for assessing the separability of the training sample. The first developed method of training sample clustering is based on the developed method for assessing the separability of the training sample. Unlike the existing ones, this method allows the selection of a number of clusters for each class, in which the separation of the training sample reached the highest value among the other considered options. The first developed method of reducing the training sample size is based on the developed method for assessing the separability of the training sample. Unlike the existing ones, this method allows the selection of such input data layers, in which the corresponding training sample will reach the highest value of separability among other considered variants of input data. In this way, not only the dimensionality of the input data is reduced, but also the separability of the training sample is increased. Keywords: aerial and space imagery, unsupervised classification, supervised classification, clustering, training sample, training sample separability Based on the results of the research, 21 scientific works were published, including 2 publications in monographs (of which 1 is in the Scopus database); in foreign specialized publications – 5 articles (of which 1 is indexed in the Scopus database); articles in scientific publications included on the date of publication in the list of specialized scientific publications of Ukraine category B - 4; 10 in collections and abstracts of reports at Ukrainian and international conferences (of which 3 are foreign) and of which 8 are indexed in the Scopus database.

Read more

Research papers

Popov, M., Stankevich, S., Kozlova, A., Piestova, I., Lubskiy, M., Titarenko, O., Svideniuk, M., Andreiev, A., Lysenko, A., & Singh, S. K. (2021). Long-Term Satellite Data Time Series Analysis for Land Degradation Mapping to Support Sustainable Land Management in Ukraine. Geo-Intelligence for Sustainable Development, 165–189. https://doi.org/10.1007/978-981-16-4768-0_11

Stankevich, S. А., Zaitseva, E., Kozlova, A., & Andreiev, A. (2023). Wildfire risk assessment using earth observation data: A case study of the Eastern Carpathians at the Slovak-Ukrainian frontier. In Studies in systems, decision and control, 131–143. https://doi.org/10.1007/978-3-031-40997-4_9

Popov, M., Michaelides, S., Stankevich, S., Kozlova, A., Piestova, I., Lubskiy, M., Titarenko, O., Svideniuk, M., Andreiev, A., & Ivanov, S. (2021). Assessing long-term land cover changes in watershed by spatiotemporal fusion of classifications based on probability propagation: The case of Dniester river basin. Remote Sensing Applications: Society and Environment, 22, 100477. https://doi.org/10.1016/j.rsase.2021.100477

Андреєв, А. А. (2023). Методика класифікування об’єктів на аеро- та космічних зображеннях в умовах низької розділимості розпізнавальних ознак. Український журнал дистанційного зондування Землі, 10(3), 4-9. https://doi.org/10.36023/ujrs.2023.10.3.244

Lubskyi, M. S., Orlenko, T., Piestova, I., Andreiev, A., & Lysenko, A. (2023). Evaluation of indicators for desertification risk assessment of Oleshky sands desertification based on Landsat data time series. Ukrainian Journal of Remote Sensing, 10(1), 17–28. https://doi.org/10.36023/ujrs.2023.10.1.229

Popov, M., Stankevich, S., Mosov, S., Titarenko, O., Dugin, S., Golubov, S., & Andreiev, A. (2022). Method for Minefields Mapping by Imagery from Unmanned Aerial Vehicle. Advances in Military Technology, 17(2), 211–229. https://doi.org/10.3849/aimt.01722

Stankevich, S., Popov, M., Shklyar, S., Sukhanov, K., Andreiev, A., Lysenko, A., Kun, X., Cao, S., Yupan, S., & Boya, S. (2020). Estimation of mutual subpixel shift between satellite images: software implementation. Ukrainian Journal of Remote Sensing, 24, 9–14. https://doi.org/10.36023/ujrs.2020.24.165

Stankevich, S., Popov, M., Shklyar, S. V., Sukhanov, K. Y., Andreiev, A., Lysenko, A. R., Kun, X., Shixiang, C., Yupa, S., Xing, Z., & Boya, S. (2020). Subpixel-shifted Satellite Images Superresolution: Software Implementation. WSEAS TRANSACTIONS ON COMPUTERS, 19, 31–37. https://doi.org/10.37394/23205.2020.19.5

Andreiev, A. A. (2020). Hybrid approach to classification of remote sensing data. CERes Journal, 6(2), 32–37

Popov, M. O., Тopolnytskyi, A. V., Titarenko, O. V., Stankevich, S., & Аndreiev, R. A. (2020). Forecasting Gas and Oil Potential of Subsoil Plots via Co-analysis of Satellite, Geological, Geophysical and Geochemical Information by Means of Subjective Logic. WSEAS TRANSACTIONS ON COMPUTER RESEARCH, 8, 90–101. https://doi.org/10.37394/232018.2020.8.11

Андреєв, А. А. (2018). Особливості розмежування низькоконтрастних природних середовищ. Зв'язок, (1), 12-14.

Kozlova, A., Stankevich, S., Svideniuk, M., & Andreiev, A. (2022). Quantitative Assessment of Forest Disturbance with C-Band SAR Data for Decision Making Support in Forest Management. Lecture Notes in Computational Intelligence and Decision Making, 548–562. https://doi.org/10.1007/978-3-030-82014-5_37

Piestova, I., Kozlova, A., Andreiev, A., & Rabcan, J. (2021). Local Quality Improvement of Multispectral Imagery Classification with Radiometric-spatial Feedback. Computer Modeling and Intelligent Systems, 2864, 158–168. https://doi.org/10.32782/cmis/2864-14

Stankevich, S. A., Popov, M., Shklyar, S., Lysenko, A., Andreiev, A., Xing, K., Cao, S., & Tao, R. (2023). Satellite imagery superresolution based on optimal frame accumulation. In Springer proceedings in physics (pp. 395–412). https://doi.org/10.1007/978-981-99-4098-1_35

Lubskyi, M., Orlenko, T., Piestova, I., Lysenko, A., & Andreiev, A. (2022). Using Landsat Satellite Time Series for Desertification Processes Mapping: Case Study for Oleshky Sands, Ukraine. 16th International Conference Monitoring of Geological Processes and Ecological Condition of the Environment. https://doi.org/10.3997/2214-4609.2022580194

Andreiev, A., Azimov, O., Shevchuk, O., & Tomchenko, O. (2022). Geoinformation Technology of Temperature Mapping of Dumps based on Remote Sensing of the Earth. 16th International Conference Monitoring of Geological Processes and Ecological Condition of the Environment. https://doi.org/10.3997/2214-4609.2022580117

Andreiev, A., Kozlova, A. (2021, 21–24 September) Enhancement of Land Cover Classification by Training Samples Clustering. Pattern Recognition and Information Processing (PRIP'2021): Proceedings of the 15th International Conference. Minsk: UIIP NASB. P.223-227

Popov, M., Zaitsev, O., & Andreiev, A. (2020). A Method for Combination and Ranking Hypotheses Under Conditions of Partial Uncertainty. 2020 IEEE Ukrainian Microwave Week (UkrMW). https://doi.org/10.1109/ukrmw49653.2020.9252781

Titarenko, O. V., Sedlerova, O. V., +& Andreiev, A. A. (2020). The new approach to forecasting areas with oil and gas prospects by classification method. Geoinformatics: Theoretical and Applied Aspects 2020. https://doi.org/10.3997/2214-4609.2020geo105

Kozlova, A., Khyzhniak, A., Piestova, I., & Andreiev, A. (2018). Synergetic Use of Sentinel-1 and Sentinel-2 Data for Analysis of Urban Development and Green Spaces. Proceedings. https://doi.org/10.3997/2214-4609.201801846

Stankevich S., Zaitseva E., Kozlova A., Andreiev A. (2022, 14-15 November) Wildfire risk assessment using Earth observation data: A case study of the Eastern Carpathians at the Slovak-Ukrainian frontier. The Second International Workshop on Reliability Engineering and Computational Intelligence (RECI 2022): Proceedings of the 2nd International Workshop. Delft

Read more

Files

Andreiev_thesis.pdf

autoreferat-Abstract_Andreiev.pdf

Similar theses

0824U000943

BATOUL HADI EL HAJ SLEIMAN

OPTIMIZATION MODELS OF FAULT-TOLERANT AND SECURE ROUTING IN A TELECOMMUNICATION NETWORK OVER DISJOINT PATHS

0824U000870

Anatolii Мerzlikin

METHODS AND MEANS OF INCREASING THE PERFORMANCE AND AVAILABILITY OF MILLIMETRE WAVE COMMUNICATION CHANNELS

0824U000397

Volodymyr Prokopets

A comprehensive method of planning a cellular radio network

0824U000314

Nikita V. Yezerskyi

DSP module for seismic-magnetosphere correlation research

0824U000288

Serhii Sokolskyi

Acoustic detection of unmanned aerial vehicles