Earlier knowledge on the landscape of an optimization problem is highly important and desirable for a subsequent beneficial choice of the appropriate algorithm and the accurate setting of the involved parameters. The primary form of this prior information is represented by the number of existing optima, which must be properly determined together with some indication of their position while at an inexpensive computational cost. It is with these aims that the current paper puts forward a novel technique that estimates the number of peaks and their approximate locations by means of an evolutionary algorithm. The approach embraces a problem-independent topological discrimination among basins of attraction for optima and considers an elitist methodology for the preservation of possible solutions, at a minimum expenditure of calls to the objective criterion. Since knowledge on the multimodality of the problem may be alternatively directly derived from a clustering procedure, the proposed evolutionary technique is put against two well-performing such methods. To show the general nature of the approach and make the comparison more impartial, the test suite comprises of artificial functions and a practical clustering instance, all defined over real-valued domains.