DOI QR코드

DOI QR Code

Region of Interest (ROI) Selection of Land Cover Using SVM Cross Validation

SVM 교차검증을 활용한 토지피복 ROI 선정

  • 정종철 (남서울대학교 공간정보공학과) ;
  • 윤형진 (남서울대학교 공간정보공학과)
  • Received : 2020.04.28
  • Accepted : 2020.06.12
  • Published : 2020.06.30

Abstract

This study examines machine learning cross-validation to utilized create ROI for classification of land cover. The study area located in Sejong and one KOMPSAT-3A image was used in this analysis: procedure on October 28, 2019. We used four bands(Red, Green, Blue, Near infra-red) for learning cross validation process. In this study, we used K-fold method in cross validation and used SVM kernel type with cross validation result. In addition, we used 4 kernels of SVM(Linear, Polynomial, RBF, Sigmoid) for supervised classification land cover map using extracted ROI. During the cross validation process, 1,813 data extracted from 3,500 data, and the most of the building, road and grass class data were removed about 60% during cross validation process. Based on this, the supervised SVM linear technique showed the highest classification accuracy of 91.77% compared to other kernel methods. The grass' producer accuracy showed 79.43% and identified a large mis-classification in forests. Depending on the results of the study, extraction ROI using cross validation may be effective in forest, water and agriculture areas, but it is deemed necessary to improve the distinction of built-up, grass and bare-soil area.

본 연구는 토지피복 분류에 사용 가능한 ROI 생성 과정에서 기계학습 기반 교차검증을 활용하였다. 연구지역은 세종시를 포함한 2019년 10월 28일 단시기 KOMPSAT-3A 영상을 활용하였다. 연구 과정에서 4개의 밴드(Red, Green, Blue, Near Infra-red)를 독립변수로 교차검증 과정에서 학습시켰다. 또한 SVM의 4가지 기법(Linear, Polynomial, RBF, Sigmoid)을 활용하여 추출된 ROI를 기반으로 토지피복 분류를 실시하였다. 교차검증 과정에서 훈련된 3,500개의 데이터 중 1,813개의 데이터가 추출되었으며 건물, 도로, 그리고 초지에서 약 60%의 데이터가 제거되었다. 추출된 ROI를 기반으로 다른 SVM기법에 비해 SVM Linear 기법이 91.77%로 가장 높은 분류 정확도를 나타냈다. 분류 클래스 중 초지의 경우 산림과의 오분류가 가장 많이 발생하며 79.43%의 생산자 정확도로 가장 낮은 분류 정확도를 보여주었다. 연구 결과에 따라 교차검증에서 추출된 ROI는 산림, 수역, 그리고 농업지역에 대해서는 90%이상의 분류정확도를 보여주며 효과적인 분류결과를 도출할 수 있었으나, 80%의 분류정확도를 보여주는 건물, 도로, 나대지, 그리고 초지 지역을 분류하는 방법에 대해서는 추가적인 연구가 진행되어야 할 필요성이 존재한다.

Keywords

References

  1. Moon CS, Shim JY, Kim SB, Lee SY. 2010. A Study on the Calculation Methods on the Ratio of Green Coverage Using Satellite Images and Land Cover Maps. Journal of Korean Society of Rural Planning. 16(4):53-60.
  2. Sunwoo WY, Kim DE, Kim SK, Choi MH. 2017. West seacoast wetland monitoring using KOMPSAT series imageries in high spatial resolution. Journal of Korea Water Resource. 50(6):429-440.
  3. Yeom JH, Kim YI. 2014. Automatic Extraction of the Land Readjustment Paddy for High-level Land Cover Classification. Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography. 32(5):443-450. https://doi.org/10.7848/ksgpc.2014.32.5.443
  4. Jo KH, Jeong JC 2019 Reliability Evaluation of KOMPSAT-3A Training Data Automatically Selected Using Iterative Trimming Algorithm. Journal of the Korea Spatial Planning Review. 103:115-129. https://doi.org/10.15793/kspr.2019.103..007
  5. Hong YW, Park WY, Song HS, Jung CH, Eo YD, Kim SJ. 2010. Image Classification for Military Application using Public Landcover Map. Journal of Korea Institute of Military Science and Technology. 13(1):147-155.
  6. Abdi AM. 2020. Land cover and land use classification performance of machine learning algorithm in a boreal landscape using Sentinel-2 data. GISciencce & Remote Sensing. 57(1):1-20. https://doi.org/10.1080/15481603.2019.1650447
  7. Ahmed FYH, Ali YH, Shamsuddin SM. 2018.. Using K-Fold Validation Proposed Models for Spikeprop Learning Enhancements. International Journal of Engineering & Technology. 7: 145-151.
  8. Ramezan CA, Warner TA, Maxwell AE. 2019. Evaluation of Sampling and Cross-Validation Tuning Strategies for Regional-Scale Machine Learning Classifiacation. Remote Sensing. 11(185):rs11020185.
  9. Sharma RC, Hara K, Hirayama H. 2017. A Machine Learning and Cross-Validation Approach for the Discrimination of Vegetation Physiognomic Types Using Satellite Based Multispectral and Multitemporal Data. Hindawi Scientifica. 2017:9806479.
  10. Tennenholtz G, Zahavy T, Mannor S. 2018. Train on Validation:Squeezing the Data Lemon. sata.ML. arXiv:1802.05846v1.
  11. Yang L, Cervone G. 2019. Analysis of Remote Sensing Imagery for disaster assessment using deep learning: a case study of flooding event. Soft Computing. 23(24):13393-13408. https://doi.org/10.1007/s00500-019-03878-8
  12. Zhang J, Okin GS, Zhou B. 2019. Assimilating optical satellite remote sensing images and field data to predict surface indicators in the Western U.S.: Assessing error in satellite predictions based on large geographical datasets with the use of machine learning. Remote Sensing of Environment. 233:111382. https://doi.org/10.1016/j.rse.2019.111382
  13. Zhang K, Liu N, Chen Y, Gao S. 2019. Comparison of different machine learning method for GPP estimation using remote sensing data. Materials Science and Engineering. 490(2019):062010.

Cited by

  1. 머신러닝 자동화를 위한 개발 환경에 관한 연구 vol.15, pp.6, 2020, https://doi.org/10.14372/iemek.2020.15.6.307