Sudhir Gupta
MS by Research in Computer Science Engineering

Areas of Interest:Image Processing, Remote Sensing-Land Cover Classification, Pattern Recognition




Time series of satellite imagery provide a lot of information about land use which is yet to be exploited. I work on creating new algorithms for land cover classification using time series of satellite imagery. Vegetation dynamics cannot be analyzed using single date images and this is where time series comes to our rescue by providing temporal information. Also quick revisit times of moderate resolution satellite imagery allows us to monitor the same geographic area at fine temporal resolutions. Time series has to be pre processed before we can make sense of it. Some valuable insights have been gained in classifying time series similar to spectral curves using a library. It is being found that Euclidean distance is not a good metric for time series classification and dynamic time warping is a better distance metric for such classifications as these processes may have temporal shifts due to various factors.