A Survey on NDVI Estimation Techniques Using MODIS Data

Bestin Paul
2019 International Journal for Research in Applied Science and Engineering Technology  
Kerala Flood 2018 causes massive destruction of the vegetation area. For the estimation of this Greenfield changes we can effectively use the NDVI technique. MODIS data is used for calculating NDVI. Processing the high resolution Satellite Images is a challenging task. Several authors proposed different methods to solve this problem. This paper summarizes these different works and the methods used in them. I. INTRODUCTION Remote sensing using satellite images has gained popular attention due to
more » ... ar attention due to its versatility and accuracy. This is very useful for processing large geographical area of interest. Greenfield estimation is one of the most popular applications of remote sensing. NDVI is the most common technique used for this purpose. NDVI which stands for Normalized Difference Vegetation Index is an indicator of vegetation greenness. It is computed from the difference between the reflectance of the near-infrared and the red bands. MODIS data which is an open source provided by NASA are used for calculating NDVI. Many authors tried to analyse MODIS data with the help of NDVI in order to indentify green area, effect of landslides on vegetation, crop identification, yield estimation etc. Also some author tried to implement new algorithms to reduce noises in the MODIS data. This survey is just an evaluation of some of these techniques. II. RELATED WORKS Authors, Jing Huang et al. [1] proposed a Normalized Difference Vegetation Index (NDVI) based estimation for harvest yield calculation. Here they used correlation analysis technique of crop yield and MODIS-NDVI data in order to find the best timing for booting, flowering, and ripening of the crops. Their study concentrates on Chinese three small counties of Province of Yunnan, and they selected corn, winter wheat and paddy rice as the three main crops. This system works comparatively well in small regions. It identifies the crop types efficiently in small regions. This is a 10-year comparison of crop fields. But extra inputs are needed to find out crops other than the mentioned ones. This paper [2] particularly studies the after effects of the 2008 Wenchuan earthquake. The earthquake causes major landslides that led to serious vegetation destruction. NDVI drops in an interval of time can be used to find out landslides. MODIS possess such good quality resolution that is very useful for landslide identification. At first, threshold is determined. For that, MODIS NDVI data is analysed for a short duration of 8 days in a small area. Then it is extended to a larger area. After comparison with the SPOT5 images, 75 % accuracy is found. This method is very useful for land slide that occur rapidly. In this paper [3], the authors mainly focus on land cover change. They used an Extended Kalman Filter to determine per-pixel change of NDVI time-series satellite data. The centre pixel of 3*3 matrix of MODIS pixels were compared spatially with its neighbouring pixel's amplitude and mean parameter to find change factor. The threshold is determined form this change index which classifieds the pixel as change or no change. 89% accuracy is claimed for the change detection. It is only evaluated for the case of settlement development detection. The authors, Bin Tan et al. [4] developed an algorithm which is enhanced technique on TIMESAT algorithm for extracting phenology of vegetation metrics. They incorporated the ancillary information, land surface temperature and snow-cover flag in order to improve the TIMESAT algorithm. This enhanced TIMESAT algorithm has better overall retrieval ratio than the original TIMESAT software. For the validation the study depends on the ground data according to the experience of the data collectors. There occurs the subjective influence of the data collector, which is the main obstacle in this area according to this study. For reconstructing the high resolution NDVI time series data the authors Liying Geng et al. [5] developed a compound technique with eight techniques to avoid the noise. First they detect noisy data using the de-noising algorithms, then to replace the noisy data with the medians of the de-noised values of each technique. The asymmetric Gaussian algorithm, the filter for changing weight, the double logistic algorithm, the data reconstruction interpolation technique, the iteration filter of mean value, modified slope extraction algorithm for best index, the Savitzky-Golay algorithm, and the Whittaker smoother technique are the eight techniques
doi:10.22214/ijraset.2019.5137 fatcat:mebrmjrtqrdefodla7zkecxzru