The potential matrix mapping (PMM) represents a novel corrosion monitoring technology, offering high precision and sensitivity. It enables the non-invasive measurement of voltage changes on a metal structure’s surface, thereby facilitating the assessment of corrosion within the inner wall of the structure. However, there are difficulties in the analysis and inversion of pitting corrosion signals, such as the multi value nature of pitting corrosion morphology parameters and low detection accuracy. A novel mathematical model for the analysis of the mapping between pitting topographic parameters and electric field fingerprint feature signals has been developed using a neural network optimised based on a genetic algorithm. Compared with the traditional methods, the model proposed in this paper can effectively improve the computational accuracy of pitting signals in PMM detection techniques.