Evolutionary Converging Functions

High learning rate have been a big challenge in case of CNN which make CNN computationally
expensive and requires huge amount of training data. In our research we have aimed to increase
the learning rate of the Image segmentation process. Image of any object in the universe taken
from any type of camera can broadly be divided into spectral and geometric properties. Both
these properties are prominently visible in satellite image. Hence if a satellite image can be
classified using a method any image can be classified using the same technique. Hence, we
proposed Evolutionary converging functions which converts the spectral and geometric
properties of features in a satellite image into mathematical equations using decision tree and
then converges with a neural network. Different highresolution data have been chosen for
extracting different features out of it. This transformation process, anchored in decision tree
methodology, converges seamlessly with neural networks to yield unparalleled results, all while
eliminating the need for computationally intensive convolutions. In our pioneering study, we
venture beyond traditional boundaries by employing diverse high-resolution datasets. Each
dataset, carefully selected, promises to unlock a treasure trove of distinctive features. As we
embark on this uncharted journey, we are committed to unraveling the latent potential of
Evolutionary Converging Functions, thereby opening up new horizons for automated image
classification. The synergy between spectral and geometric properties emerges as a powerful
combination, endowing our methodology with the ability to extract nuanced and context-rich
information, redefining the landscape of image analysis. The results show a high accuracy i.e.
above 90% in almost all objects of different shape and spectral signature.