A high speed roller dung beetles clustering algorithm and its architecture for real-time image segmentation

Several practical applications like disaster detection, remote surveillance, object recognition using remote sensing satellite images, object monitoring and tracking using radar images etc. essentially require real-time image segmentation. In these applications computational complexity of the algorithm play a vital role along with accuracy. In this work image segmentation is dealt as a clustering problem and a bio-inspired algorithm based on the behavior of ‘Roller Dung Beetles (RDB)’ is proposed to determine effective solutions. The beauty of this proposed RDB Clustering architecture is its lower computational complexity O(N). The software implementation of the proposed algorithm is carried out in MATLAB environment and a hardware architecture is developed in Verilog HDL using Modelsim, Xilinx ISE for FPGA environment. The architecture has a comparison free sorting module, two data storage modules and a parallel threshold comparator unit, all of which use fewer mathematical operations. The performance of the proposed architecture is validated on many synthetic and standard benchmark color images. Further application of the proposed architecture is carried out for real-time segmentation of 8 NASA LANDSAT / ESA satellite images. Performance comparison has been carried out with other existing architectures based on Artificial Immune System (AIS), Genetic, K-means clustering, CNN etc. Simulation results reveal that the proposed RDBC architecture is 42% faster than the K-Means implementation with a clock frequency of 230.52MHz with an increased PSNR of 6.825% and SSIM of 15.55%. Statistical analysis and silhouette index also confirms the superiority of new clustering architecture over existing implementations.Compared to CNN the RDBC architecture is economical both in terms of lower chip-area and power consumption.