Classification of Remote Sensed Data Using Hybrid Method Based on Ant Colony Optimization with Electromagnetic Metaheuristic

In this study, a hybrid configuration of electromagnetic metaheuristic algorithm (EM) with Pachycondyla apicalis (API) ant algorithm (inspired by the behaviour of real ant colony Pachycondyla apicalis) belonging to ant colony optimization (ACO) called EMAPI algorithm is presented for remote sensing data classification. The traditional per-pixel classification method identifies the classes using spectral variance and ignores the spatial distribution of pixels. It requires training data to be normally distributed in the pixels corresponding to land use/land cover classes and creates a lot of confusion between classes within a remote sensed (RS) data. The proposed algorithm is an integrated strategy structure to achieve advantages of global and local search ability of EM and API algorithms respectively. The objective consists of improving overall accuracy of the classified results of RS data. This method can overcome intermixing with regard to scrub land with cultivated areas and build-up land with palm groves. The proposed algorithm is tested on objective functions well used in the literature and EMAPI is used for supervised land cover classification. Results of EMAPI algorithm over 6 classes showed an improvement of 8% in overall classification accuracy (OCA) for EM technique and improvement of 3% in OCA for API algorithm.