An empirical review on clustering algorithms for image segmentation of satellite images

Different image segmentation algorithms are used for real-time applications like autonomous vehicles, robotics, disaster management, etc. Because of the computational complexity of these algorithms, hardware realizations are cumbersome and complicated. The frames per second achieved are barely sufficient for accurate perception of the problem at hand. The next important challenge is the implementation of evolutionary clustering algorithms like genetic algorithm for improved accuracy, after introducing some simplifying techniques to make the ensuing hardware quicker, less complex with improved power consumption and hardware area/size. Bioinspired algorithms are an excellent candidate solution in this regard. It can be implemented in an FSM based approach to detect faults in the centroid initialization phase like detecting zero data members within a cluster. To further reduce the complexity challenges of existing algorithms while maintaining good accuracy, some new bioinspired algorithms like roller dung beetle clustering have also been tested.