Nature Inspired Algorithms in Remote Sensing Image Classification

Remote Sensing involves a wide variety of techniques for Image Classification of land cover features of different terrains. Different traditional image classifiers are present for appropriate use of land. However, these features are not satisfactory and efficient. This paper attempts to use Artificial Intelligent algorithms different from traditional classifiers for image classification in order to improve the proper use of land. The reason behind using Artificial Intelligent algorithms is that they have an extensive search space which increases their efficiency. The algorithms chosen for this purpose are meta-heuristic Cuckoo Search(CS) and Artificial Bee colony(ABC) algorithms. The image used for classification is of Saharanpur region of Uttar Pradesh with a 641 x 641 dimension. Both the algorithms prove to be efficient in image classification by effectively classifying each land cover feature and showing satisfactory Kappa Coefficient value of 0.96(CS) and 0.91(ABC). Various other metrics results like User Accuracy, Producer Accuracy has also been tabulated.